How to Start an AI Business in 2025 | Full Beginner Blueprint

hi there so in this video I want to give you my perspective on how to start an AI business in 2025 why I think the time is perfect to get into the space right now while I think current professionals with domain expertise are ideally positioned to capitalize on this opportunity and with so much happening in the AI space at the moment I'll try to cut through the noise and show you step by step what to learn what not to learn how to actually monetize these skills and how to Market yourself if you don't know me

yet I'm Ben I am not a coder and got into AI lab last year after my first startup went bankrupt and my life has completely changed since then my AI agency has taken off I also run a community with more than 500 professionals business owners and AI builders that are getting a foot into the AI World themselves I've experienced the opportunity in this space first hand and I want to show you why I think there's such a huge opportunity for non-technical professionals like me to transition into AI which I'll get to later in this video

I'll separate this video into three sections which in my opinion are the three fundamental things things to get right in order to be successful in the AI wave in the first section I'll go over the best monetization strategies and business models for different types of people then I'll do an AI crash course where I'll break down the three main AI skills to learn in 2025 and lastly I'll cover how to actually Market yourself and get clients effectively which is often overlooked but is arguably the most important of them all and I'll finish off by giving

you a clear action plan to get started in AI today this is a video I wish someone had shown me before I started my journey in AI now I'm not pretending to be some Guru with all the answers it's just a full breakdown of my own lessons and from what I've seen is working for others around me in my community and in the AI space now this is probably not the only way so take anything that makes sense to you and discard anything that doesn't so firstly what do I mean with getting into AI I

mean getting a skill set in AI that will allow you to ideally position yourself in the market and to monetize that position effectively whether that's as an employee uh transition into an AI agency an AI SAS or even taking your existing non- AI agency or business to the next level by adding AI into your own work workflows so why now why even consider moving into AI as a professional first of all the global AI Market is projected to hit $2 trillion by 2030 and Studies by McKenzie suggest that by the mid 2030s over 50% of

current knowledge worker jobs could be automated by AI whether it fully replaces Your Role or not AI is the next general purpose technology it will impact every industry change how businesses are run and reshape every profession yours included soon Ai and automation will become one of the fundamental departments inside of businesses and in the upcoming years every business will be forced to adapt and adopt Ai and automation into every part of their business and the people who can use and understand the technology behind these systems will become extremely valuable in the marketplace and the ones

that don't will lose their Edge so learning these skills right now isn't just helpful it will Define your entire career trajectory secondly for the first time in history you don't need a technical background to build powerful systems and products for people and businesses with no code tools and AI you can now take your ideas and transform them into prototypes in days and in production ready systems in weeks as Naval ravian said learn to sell learn to build if you can do both you'll be unstoppable Ai and no code tools now let you do both even

if you're not a coder and I can tell you from my perspective having developed this combined skill set is single-handedly the biggest reason for my career growth over the last 2 years and if you're professional with existing domain expertise whether it's marketing Finance sales or any other this shift plays right into your strengths why businesses don't necessarily only need technical people to implement Ai and automation on itself they need people that use these tools and the technology to resolve real business problems and the people who already understand how companies work and what businesses need uh

can now also actually build the systems that solve those needs and that in my opinion is a superpower and I can tell you from my experience if you commit even 3 months to learning these no code tools and really understanding ai's capabilities you will be ahead of 99% of the world and you'll be able to build some incredible things but the time to move really is now while most people are still on the sidelines which brings me to the the first chapter AI business models now before giving you a crash course on the most important

AI skills to learn I want to give you an overview of how you can actually monetize these skills what are the best business models in Ai and Automation and how do you actually offer these kinds of services over the past year I've tested a bunch of different business models and offers myself and I've also had the chance to connect with uh many other successful Founders who've taken different approaches to AI in my community and through my network and from all of those convers ations and my own trial and error I've narrowed things down to what

I believe are the five most effective business models in Ai and automation right now but before I dive into the five business models that I think are most interesting there's one crucial point that most successful Founders in the space have in common they usually don't sell AI on itself AI by itself is just a technology and businesses aren't necessarily paying for a technology they're paying for solutions that actually move the needle the real opportunity is in identifying and solving real business problems using Ai and no code tools this is another reason why I think professionals

have such an edge because they naturally have more experience working on and resolving real business needs so the five opportunities and business models I'll go over are the AI automation agency which you've probably already heard of because it's most talked about and is a great place to start for beginners getting into this phace then we have the AI productized Services model which can be really interesting for existing non- AI agencies or businesses uh but also for Freelancers Consultants or people with specific domain expertise then we have the service as a software model which is sort

of new and popping up in this AI wave not to be confused with traditional software as a service which is the fourth one right AI SAS and lastly we have ai education I think demand will grow exponentially in the upcoming years for all of these business models because again companies will have to start adopting Ai and automation to stay competitive in the marketplace and the majority will lack the in-house expertise to do this by themselves so either companies will have to hire expensive AI Talent which is cars or Outsource these services to third party businesses

and agencies now to be clear some of these business models overlap a bit and the biggest difference uh between the first four at least is the amount of individual customization required for each client now the first one which I think most people watching already know is the AI automation agency model now in this model you essentially offer Automation Services to increase efficiency reduce costs or solve specific business need now this is also the model where the highest degree of customization is required per client now I've separated this model into two types which is the general

AI automation agency and the niched AI automation agency in the general AI automation agency model you're basically very widely focused and Implement almost any type of automation solution for a very wide variety of Industries now this is a model I don't recommend for most people starting out in the space except if you maybe have a very technical background maybe you're a development agency transitioning into AI or if you have access to a big developer team the reason is because the variety of projects is going to be very wide which means you always have to start

from scratch you'll have to learn lots of different Technologies and solutions from the get-go it's hard to understand all the company needs as you can't have domain expertise in all areas of business and because of it you'll spend huge amounts of time scoping out projects understanding what needs to be built and this is the phase where in my experience a lot of deals go boss and each project requires a lot of time and you'll have to price your services at a really high price point which makes it harder to close a deal it's also going

to be harder to Market yourself as you're not addressing a specific need or industry but I still see many people trying to jump into this model uh at the beginning which is sort of understandable if you don't necessarily know where to start or have a specific solution or industry in mind yet but I still think that the second option is the better place to start for most people which I'll go over in a second that being said there are definitely examples out there of successful General AI automation agencies for example my friend Henry at for

a. they have a more technical team and can Implement very custom projects for a very wide variety of Industries and it can also be a great way to start as along the way you can off specific Solutions or industries that have high leverage to Niche down into later then the second one is the niched AI automation agency which I think is best for most people starting out in the space this is what I started when I started in Ai and I also called this the fractional AI model some people call it uh the AI officer

but in this model essentially you focus on automation solutions for either a specific industry like uh real estate or recruitment or for specific Department like sales or marketing or you can Niche down even further by combining these two like sales automations for real estate companies now this makes sense for most non-developer professionals that are transitioning into AI why because you first of all you want to focus on a niche you have some expertise in firstly because most valuable automations require at least some domain expertise to actually drive good results a programmer that has never done

sales before can definitely make an automated outbound email system but most likely it's not going to get very high converting emails A salesperson designing the same system will probably get way better results secondly you have a clearer target market and offer so it's easier to Market your services and present a valuable offer and a solution to a common business problem besides that most businesses are messy they don't have clear Sops or processes in place and a lot of times besides just automating existing processes you're actually establishing or improving the process in the first place before

setting up an Automation and again domain expertise is key here and lastly because you Niche down you can start reusing more and more automations instead of starting from scratch each time a great example is Brandon who founded agent automation.com Unity who focuses on automations for Real Estate businesses because he has a background in real estate and although for all real estate companies automations will requ require customization and each real estate agent or company will have their unique little workflows and processes generally real estate companies will have similar processes and therefore more and more of his

automations can be reusable and therefore his agency can scale now if you want to check out more of Brandon I'll also make sure to put his Link in the description below but essentially you can position yourself as the in-house automation or AI department for a specific business or Department within a business or both most companies are still trying to to wrap their head around what's actually possible and what can be done with AI on Automation in that company so usually in this model you provide a mix of Consulting implementation and coaching so you figure out

what processes can and should be automated with the founders or the department heads right you implement the right tools and automations and keep refining and adding new automations month after month another example is Alex who founded sales automated. who's also a member of my community who has background in sales and noticed many of the work of bdr and SDR roles can be automated with AI so he has positioned himself as a growth partner agency where he helps companies acquire and convert more leads by implementing outbound and inbound sales automation systems I'll also make sure to

put his Link in the description below too if you want to learn more now the key with this model from my experience is to position yourself as a partner instead of a builder and as a partner you're not just automating existing processes you're actively thinking with a business to enhance their current processes and that's again where the domain expertise will really give you an edge now many people might be insecure if they can find enough or new automation opportunities within a company but trust me you will there are so many processes in most businesses that

should be automated most companies have done nothing with automation yet and most business owners will already have quite some ideas when they approach you and if not they'll come up with them very quickly after they see what's possible after implementing one or two of these automations we have yet to see one company that after starting doesn't already have a long backlog of next automations they would like to implement and again as you're positioning yourself as a partner you want to also price your services as being a partner what worked by far the best for us

is a subscription uh style partnership rather than Project based pricing right so you get a paid the retainer and in exchange you handle ongoing Discovery maybe weekly or bi-weekly uh consulting or meetings and automation implement ation I personally scaled this model to $225,000 a month when I was still working alone with just five retainer clients right each paying uh $5,000 I dedicate roughly one day a week per client but this of course can be adapted to any range right we now charge a lot more sometimes too for a more intense commitment but because of the

subscription model you get consistent income you'll have a long-term relationship with clients who rely on your expertise to keep them ahead of the game now the subscription pricing instead of the Project based pricing also keeps the ball on their end so you don't have to endlessly scope out projects make proposals negotiate constantly and go back and forth like we saw a huge spike in conversions and revenue when we changed from implementation based pricing to subscription based pricing some of these agencies especially in the lead gen space can even ask for a ref share or for

Success fees for booked appointments for example now as these kinds of services are very new to most companies and many companies don't know any other company or founder yet who has done this you will get uh some objections especially with this business model for example the most common ones we get is like what do I actually get for my monthly fee or how long does each project take now in my experience it helps a lot if you have a small portfolio of example automations you can show along with typical timelines of implementation so they get

an idea of what they can expect and how long it takes usually clients also have one or two automations in mind already when they approach you giving them a timeline on the delivery of those also usually helps we also have a cancel any time policy and a money back guarantee if they're not happy after the first month which has helped some of them take the leap and if they're still on the fence which again sometimes happens we sometimes charge a separate lower fee for a smaller initial AI automation implementation to show them what we're capable

of and that these systems can actually work and then we get them over onto the subscription model now the amazing thing with this model is that almost all companies have stuck with us even after 6 months of being with them but as said even though you might be able to reuse more and more templates because this model requires a higher degree of customization for each client it is also harder to scale and can mostly scale through hiring more people now if you want to stay alone this can still be scaled to a really decent Mr

with very minimal overhead but it can also naturally transition into a model that is more scalable like the AI productized services model so the second one is the AI productized Services model now like I said this model can be a natural transition from the niched agency model but can also be an ideal model to start for existing agencies Consultants Freelancers or people with specific domain expertise the product eyes Services model is much more like a traditional service delivery company or agency you're delivering a specific repeatable solution to a well-defined business problem like a lead generation

agency or a marketing agency the big difference being by automating and enhancing and systemizing your backand service delivery with AI and automation you can first of all enhance the value of the offer and second scale to levels that traditional agencies could never scale to a great example of this is Brooks also a member of my community who founded go pipeline formula.com he has a lead generation agency that helps SAS Founders generate leads by building their personal brand through content he records one interview or podcast with the founder per month and the AI system he created

automatically creates hundreds of pieces of content out of that one podcast that will be scheduled automatically across this founder social media for the whole month now important here he is not selling AI he's selling lead generation and of course there's still some work involved but because he systemized and automated most of his backhand Service delivery with his AI content repurposing automation he can scale this model to incredible numbers without growing his team I'll also make sure if you want to learn more about Brooks to make put his Link in the description below another example is

T A friend of mine who had an existing gold email agency which he productized even further with AI and automation T does cold email lead generation for m&a companies and m&a Deals and he niched down to only m&a companies so his Outreach processes are very reusable for each new company and he uses AI personalization and automation to do Mass Outreach while maintaining highly relevant emails High open rates and reply rates T also did a full master class uh in my community on called email so if you're interested in that uh make sure to check it

out I'll put his Link in the description below too now again this model can be great for existing agencies Freelancers Consultants or Professionals in for example lead generation email marketing Etc because creating high quality automations and systems that actually deliver results across a variety of clients requires usually a lot of domain expertise and experience and because the service is productized and automated it allows you to really scale to really really interesting numbers and because you're solving a specific business problem clients are willing to pay a premium many have the ability to charge an implementation fee

plus a revenue share and a monthly retainer especially in the in the Le gen space now the biggest challenge with this model is to make sure the solution delivers consistent results across different clients while maintaining minimum customization and while it's way more scalable than a general agency or even a niched agency there's still some ongoing work and customization involved so another natural transition for productized services could be to transition into the next model which needs even less ongoing work the service as a software model now this is sort of new and that is popping up

in this AI wave and is not to be confused with traditional SAS or software as a service but basically it's like a plug-in play outof thee boox automation system that solves a specific business need if you have developed and optimized a producti service or a niched service so much that it requires almost uh no customization can be completely self- served in the long run and requires close to zero maintenance or support this model can be really interesting for almost infinite scaling we have developed one of these over the last months and are starting to roll

this out basically we started with a product eye service for SEO content generation and we've optimized and productized it so much that it can now be plugged in any existing company within a couple of hours it's a completely self-served platform on air table and basically acts like an SEO Agency for a client all the client needs to do is add a topic and the system does the rest it finds the best keywords decides on the content it has to write adds meta titles does competitor uh analytics write multiple SEO optimized articles with images tables and

post them on the client's website while they are both self- sered solution and they can definitely overlap a bit the key difference between traditional SAS and Service as a software is that traditional SAS usually provides tools or software that assists users when working on a specific use case but still requires them to manage the process and make the decisions Service as a software on the other hand handles almost the entire use case and process by itself and delivers a completed outcome requiring very limited uh involvement from the client for example in the SEO space an

example of a traditional SAS could be sem Rush right it assists a user with SE o content generation it provides keyword research content analytics uh competitor Insight but the user must still create the content optimize it and Implement a strategy for themselves for the service as a software it takes over and handles the entire process it generates optimizes and publishes SEO driven content automatically the system handles almost everything from start to finish another example of this is yonas a friend of mine who offers voice agent solutions for hotels in Germany with aura AI he is

productizing his service so far that he just needs to customize the system for a few weeks to meet a specific hotel's requirement and then the system runs for a client in the long run with minimal maintenance he's able to charge an implementation fee plus a usage based pricing model that can of course be scaled up to very interesting numbers uh if you want to learn more about Ori I'll also make sure to put his Link in the description below and as it almost entirely takes over the service delivery of a traditional agency or employee therefore

it's called a Service as a software and again because it can replace expensive agencies or people this can really be sold at a premium implementation price with usage cost or subscription fees because companies will save huge amounts of money in the long term we are planning to develop more of these kinds of solutions but I'll keep them for myself until we have launched them that being said this of course sounds like an amazing business model and it is but if you haven't already been working with product ey services or niched as a niched agency and

really have developed a specific specific solution that can solve a need across different companies then it's probably not the best place to start then the fourth one is AI SAS which is the traditional software as a service model now I've run a SAS before and I hear many people on the internet saying that SAS is dead uh but I think many people talk about the existing uh overpriced Enterprise SAS applications out there because uh with AI and AI agents SAS or even custom software for businesses will now be able to be built at a much

lower cost meaning that the price points for many of those overpriced SAS softwares might be become hard to justify very soon but this opens up an opportunity for new AI SAS softwares I think and if we look just past uh enterprise software I think they're going to be very interesting uh opportunities first of all because with these no Code and low code tools and AI coding agents it's literally never been easier and cheaper to launch a minimum viable product and second of all As with AI a lot more value can be packed into a product

and can be developed at a much lower cost s softwares can now be made for very specific niches that were previously maybe underserved because creating a specific solution for that Niche would previously be too difficult or too costly a great example of this is from another member in my community who preferred to stay Anonymous he had a meta ads marketing agency for uh very small business owners in Germany and he started automating some of his own processes with with AI and basically he saw that a lot of these small businesses lacked the domain expertise to

set up Ma ads themselves but also many didn't have or don't have the marketing budget to hire an agency and as he had a lot of data on what type of AD creatives and AD copy worked best for this types of small businesses he made it into a self-served platform where these small business owners can create a high converting ad creatives and AD copy with AI themselves he launched his MVP in a matter of weeks with a backand system almost entirely set up through NOCO platforms like m.com and grew to $50,000 in Mr in less

than 3 months now note here that the customer is still in charge of the process and therefore it's not a service as a software but a software as a service also why combinator's latest batch with almost exclusively AI SAS has been one of the most successful and fastest growing batches ever but I do think again that for most people that are new to the space and transitioning into AI without an extremely strong goto Market team or experience or an existing audience or distribution leverage it's probably not the right model to start with for most people

it makes more sense to at least start with a niched agency or maybe a product ised service or even a service as a software which gives you that experience on what to build how to systemize it so far that it could potentially become a self-service platform secondly SAS will become a double-edged sword because Tech and product will be more and more democratized right very soon anyone including non-coders will be able to build powerful products with AI coding agents and these no code softwares and therefore the competition will also be much higher and the mode or

the Competitive Edge of businesses I think will shift more and more towards deliverability and distribution instead of the product or Tech and the ones that are best at reaching their target audience and having the distribution channels will win so even if you have a groundbreaking AI tool it won't matter if you don't have a really good goto Market team or existing distribution channels that's why again I usually recommend starting as an agency or product I Service First you'll be able to discover a pressing real world problem that AI can solve you can refine Your solution

with actual customers and then potentially you could consider turning it into a standalone SAS product because the best SAS ideas often emerge when you're repeatedly solving the same problem for different clients and also you will have build some distribution and marketing channels that can actually make your SAS reach your target audience which brings me to the last one on AI education now I'm not going to go in depth in this video because this would be a video on its own but there's a gigantic opportunity coming in the AI reskilling and upskilling Market most people talk

about all the job losses that AI will bring but few people talk about the massive amount of new jobs it's going to create as businesses and the world need to adapt and adopt AI these new jobs will require skills that most people simply don't have yet today this will create a massive scale Gap in the market in the upcoming years literally billions of people will need to be retrained and upscaled to meet the demands of an AI powered economy this presents huge opportunity for both B2B education right helping companies rescale their Workforce but also b2c

education helping individuals stay competitive and capitalize on these new opportunities but even if you're planning to go the educational route I still highly recommend to actually learn how to implement these real world automations as it will really help you understand where the potential lies also in this educational industry now lastly uh I touched upon a few of these but there are a couple of things I'd recommend you avoid doing especially when you're new to the space I think the first trap which I myself fell into a bit is when non-technical people learn these nood tools

or agents and discover that for the first time they can actually build really powerful uh applications and automations usually the first intuition is how can I standardize a product and basically make a SAS out of it charge on usage cost so I can scale this and get recurring Revenue this is generally uh in my opinion a pitfall because of three reasons first of all because most of these AI automations and solutions drive real business value only when customized for each business for example once the solution is integrated into their softwares for example an inbound leaders

research agent or tool might be great for salespeople but only when the research data actually appears directly inside of their CRM if they have to trigger this manually it's an extra step an extra software in their Tex tack that people are not going to use and businesses don't want secondly good products don't sell themselves especially again as Tech is being democratized more and more where non- tech people can build anything with these no code tools and coding agents the mode of businesses will shift more and more towards the deliverability instead of pure product alone and

if you have no existing or direct client Network big distribution Channel or huge goto market experience you'll likely not be able to sell your solution enough and thirdly you learn only what a good product is by working on and implementing this into companies and the best way to do this is by customizing the service first improving the system step by step once you've implemented a service in multiple companies you've likely learned a lot about what a potential SAS could look like therefore the most logical path to AI SAS for most people is to start with

a niched or general agency identify interesting use cases first when you have found these high leverage automations for specific Niche or product I service you can then potentially transform it into a SAS the second thing I would avoid is the payer project model for these automations and AI agents firstly because almost all companies will need lots of automations we have yet to see a company that only needs one and with the papered project model you're constantly adding friction to the process by having to scope out projects get proposals out and negotiate it's hard to get

continuity going which is going to hurt your cash flow and the perceived results for the company with a subscription model the ball is on on the client side they have to take advantage of the subscription you get consistency and continuity and in the end it's a win-win for both you and the client on a similar note charging by the hour I don't think makes sense in the AI space I've never done it uh because again clients shouldn't be paying you for your hours they're paying you for outcomes and hourly billing just caps your earning potential

positions yourself as a builder again instead of a partner and dilutes your perceived value lastly Consulting is another model which I've done and recently quit and I wouldn't necessarily recommend if you're starting out and you don't have a big audience it might sound enticing to just consult without the actual implementation but in my experience you can get so much more leverage and value out of these deals if you actually take charge over the actual implementation so that brings us to the next chapter AI crash course now with this overload of information and AI tools what

should you actually learn to start one of these AI businesses I've broken the fundamental skills I think you should learn that into three skills and I'll give you a crash course of the fundamentals of each of these skills and the three skills in my opinion are prompt engineering AI workflow Automation and AI agents but before I go into them I want to break down two questions I get asked a lot firstly should I actually learn the tech by myself and if so which softwares should I learn now to address the first question should I learn

the tech right I usually tell non-technical professionals right you basically have two options right you either part up with someone with a technical background or a technical skill set or second you learn the no code tools yourself and I know for many non- tech people even these no Cod tools might seem pretty daunting at first and the first intuition might be to just partner up with someone however my recommendation is usually to actually learn at least the fundamentals of these tools even if your plan is to eventually partner up why three reasons the first one

domain expertise is what almost always gives these Solutions value as I said before for many not all but many of these Solutions domain expertise is actually what drives the value and results from these agents and automations it's not just the tech a tech person with zero marketing knowledge automating a newsletter system will most likely result in bad and low converting email copy the real value and strength of these systems comes when a marketer right with domain expertise injects his domain expertise into these systems to actually get high quality outcomes secondly you actually know what's possible

even basic uh familiarity with these no code AI tools will help you recognize which task can be quickly automated what the scope of projects are and will allow you to speak with confidence to clients and you'll also be able to speak the same language as any future technical partner you work with so you can actually challenge ideas right refine them and avoid over promising on features to clients that are simply not feasible and thirdly and most importantly it is not rocket science right these no code AI tools are much less intimidating than they first appear

speaking from my own personal experience I managed to learn most of what I needed in about 3 to 6 months and if you focus on the 20% of the skills that matter most you can start building really highly effective automations without becoming a hardcore developer and I've boiled it down to three fundamental areas you need to know for 20125 which I'll cover in this section but once you nail those Basics we'll realize that your value in the market replace 10 X's because 99% of people will not learn this and I think the real leaders in

the AI wave will be the generalist so to address the second question which softwares should I learn is it n8n make.com relevance AI voice flow flow wise Lindy now take this with a small grain of salt because this is a bit of an oversimplification but I made this canvas to give you a better idea and to help you choose which software to learn in simple terms on the X access we have the ease of use of the tools going from really easy to use on the left to more technical on the right and on the

Y AIS we have platforms focused more on AI agent building at the top and platforms more focused on workflow automation at the bottom now important to keep in mind is that generally the easier a platform is to use it will of course mean you'll be able to master the tool faster build automations faster but it also generally means higher usage costs and less customizable or flexible this can be great for simpler use cases fast deployment of automations and for for not extremely high volume use cases but of course might limit Us in terms of customization

and flexibility for more custom projects and workflows for high volume use cases it also means significantly more usage costs for example a platform like zapier is very easy to use you'll be able to deploy an automation very fast and it works great for simpler workflow automations but will cost significantly more than a platform like make.com or nadn a platform like nadn or pip dream will have significantly lower cost will allow for more compx workflows and automations but the learning curve is a bit steeper and it will take more time to set up automations secondly although

on most platforms you can build both some have a stronger focus on AI agents building and some more on traditional automations and AI workflow automations now both are core skills you should learn in 2025 in my opinion knowing that what softwares should I learn now if you come from a non-technical background or a non- automation background my initial recommendation is to start with a combination of make.com and relevance AI now of course this is a bit biased because this is what I learned at the beginning but both of these platforms are really intuitive when you

just start out have a low barrier to entry and still allow you to be really flexible and to build more advanced complex systems make.com will teach you workflow automation help you learn how to connect softwares and apis initial prompt engineering and data management and data transformation relevance AI will give you an intuitive agent building framework to to start learning how to build AI agents and AI agent teams and once you get the hang of these platforms or if you already have some experience with automation or you're a bit more technical my recommendation is to start

trying to work with n8n uh which again is a bit more Technical and has a steeper learning curve but has the big advantage of having both strong workflow automation building and agent building in one platform lower usage cost and self-hosting capabilities and of course if you're a coder you can also consider going for coding Frameworks like crew AI Lang graph and Lang chain voice flow and flow wise are also popular platforms and shine mostly in my opinion for more chatbot use cases as they were traditionally more chatbot Builders so they can be interesting to explore

for those types of use cases now lastly it is also important to understand that most of these softwares have very similar setups and infrastructures so learning one will mean your skills will be very easily transferable to other platforms if you want to switch the real skill you'll learn when building these systems is how to architect these Solutions and that knowledge is very transferable to any of these platforms my recommendation is try a few check the one you like most out of any of these really and just start building some automations which in the beginning is

the most important step anyway now let's start with the first one the main fundamental skill to deploy reliable functioning and cost-efficient AI systems is prompt engineering this is also the part where your domain expertise really makes the difference in making a mediocre automation system or a highly valuable system injecting your knowledge inside of these prompts is what makes these systems generate valuable outputs now this might not be the sexiest topic and it's really not rocket science but most people still don't really know how to prompt efficiently the main reason being is that most people are

used to prompting inside of chat GPT which is actually very different from prompting inside of these AI systems in chat GPT we use conversational prompting therefore we can go back and forth multiple times with the model and until we get the desired outcome and therefore we don't really need to implement these proven and studied uh prompt engineering techniques when we prompt inside of AI systems we don't have the flexibility of that back and forth tweaking we need to get the system to generate a good reliable and consistent output every time even with the edge cases

and that really does require good prompt engineering and the best way to make sure we use prompts that are efficient reliable and incorporate uh those best practices is to use Frameworks prompting Frameworks structure your prompts in a way that makes it easy to include uh proven and study techniques but maybe more importantly gives you a repeatable framework to make sure you always include everything that you need to include to get consistent and quality outputs I have a full 50-minute prompt engineering video on my YouTube channel too which I highly recommend you watch but I'll give

you a quick uh breakdown of the most important things here so let me go over the three Frameworks that I consistently use in my own AI systems now remember this is is not the only way there are lots of ways to efficiently prompt but this is just what works for me and what I use so firstly I have the long structured prompting framework then I have the short structured prompting framework and lastly the agent prompting framework now the long structured framework I generally use this for more complex tasks and it includes seven sections now the

first one is the role in this section we want to give the L&M or the language model a role and qualities which is studied to lead to an increased performance let's say for example I want to write a prompt to write a job proposal for my company an example of this section could be roll and I use the hash to get the header you're a worldclass HR specialist with particular expertise in writing engaging draw proposals so in this section besides the role we also really want to Hype up the model and tell it how good

it is at something then we have the second one which is the objective now in the objective section we give the model a direct description of what needs to be done you can think of this section as what you would put into chat GPT for example objective your goal is to write a compelling job proposal to attract talented people to join Ben a you think step by steps through the following process one you'll carefully read the job description in the variable section two you'll write a compelling job proposal using the specific structure as laid out

in the examples and three the proposal should explain the role highlight key responsibilities and motivate candidates to apply then the third one is context now in this section we want to give the language model more context on why it's doing this task and how this task fits into the bigger picture for example context Ben AI helps businesses automate processes using Ai and automation Solutions this job proposal is crucial for helping B grow his business and to drive client success now again giving these language models more context on why this task is important can again improve

outcomes then the fourth one is the instruction section now in this section we go into more detail and include all the important rules and specific output formats if you have them a good way to think about this section is to predict what could go wrong and add the these in a form of rules to this part of the section for example instructions you will always use the same format and structure as in the examples below you will never make up facts about the role and it is vital to my career you use no more than

150 words now this it is vital to my career might sound a little strange but this is what they call emotional manipulation tricks which again might be able to to uh actually increase the performance of these models then we have the example section uh which is one of the most studied and most important sections of your prompt giving input and output examples are really key for Reliable outputs because you give the language model an idea of what it should do with a specific input I would always recommend to include at least two examples now a

good and fast way to do this is by first running a language model without this section to generate an output uh example with the best language models you can then tweak the output if you need to and then add them to this section the way I usually do it is with input and output now then we have the variable section the variable section will change of course based on the specific context or variables for this specific task for example this prompt could be part of a tool that could generate me job proposals for different kinds

of jobs so the variables could be job title with the variable description with the variable and salary with the variable and then lastly we have the notes section now this is actually a pretty important section because lnms actually take the beginning and the end of prompts more into account than instructions given in the middle of the prompt so the note section can be really great to double down on important roles specific output formats and can also be great for when you see that your language model struggles with a specific thing while you're uh testing it

you can add that specific rule to correct it to the note section and you usually see that it corrects itself immediately for example you see that the job proposal is too long the first time you could add something like it is vital to my career you are concise and use less than 150 words in the job proposal now as we don't always need really long detailed prompts for very simple tasks we also have a short structured framework that is basically a reduced version of the long framework right so in this one usually what I have

is the role and objective which I put in one then I have the example sections and the variables if we have any variables because they're simpler tasks we don't need all the other things we can use this framework for simple task because sometimes when we have simpler tasks and we use these really long prompts language mons actually get confused and lastly we have the framework for agents as agents usually have more decision-making responsibilities uh and they also have access to sub agents and tools we make a slight adjustment in the framework so we have role

objective sop context instructions sub agent and tools examples notes now as you can see uh we have two new sections in this agent prompting framework which is sop in sop we basically give the agent more instructions on how to execute a specific uh task or query so we give it more context on in which sequence it has to perform actions use tools use sub agents Etc and in the tools and sub agents section we give our agent more context on what tools and sub agents it has access to because it's very important for agents to

understand what these tools can do when to use them and uh what they can expect uh from these tools now when do we use which prompting framework this really depends on the use case of the prompt and the complexity of the task now in very broad terms there are six main types of prompting use cases in AI systems now take this with a grain of salt because there are many nuances to this and many different use cases and complexities uh beneath each of these categories but I'll try to give you a general guideline for which

type of framework to use and which type of language model to use for each of these types of use cases firstly we have extracting data now these are use cases where an nnm should extract data from a larger context or summarize something from a given context for example uh finding a specific URL from a scraped Google search result now in general uh for tasks like these where l&m's have lots of context N&M are usually very good at this so we can usually get away with using the short structured framework and cheaper language models but again

this does depend on the exact use case and the complexity now secondly we have classification or categorization these are use cases where an L&M should categorize for example categorizing an email now l&m's again in general are very good at this so usually we can get away with short structured prompting and cheaper language models but again if we have very specific and subjective categorization we need to give more context to the model and then long structured prompts will probably be more appropriate then we have generation these are use cases where L&M should generate something right like

text image or video for example generating a report uh an image a job description right now these are one of the hardest tasks for language models and where we see the most hallucinations occur and therefore it's best to give language models as much guidance as possible with long structured prompts and to always use the best models and for more important and more complex tasks you can even consider using the system 2 level thinking models like uh gp01 then we have evaluation now these are also common use cases in inside of these systems as sometimes we

try to build in extra steps to evaluate outputs generated by the system to make sure we got the right results out of other language model steps a common use case for example is to evaluate whether a rag output actually matches the original query from the user or answers the original question now again language models are generally pretty good at these tasks but given that these tasks are usually very important for functioning reliable systems we should use the more expensive uh better models and generally longer structured prompts but sometimes we can get away with uh shorter

ones then we have data transformation these are used cases where an L&M should transform or restructure data in a new format or structure like HTML Json or just plain restructuring of a text in a specific format L&M can sometimes struggle with these tasks and with new features like structured outputs this does become easier when these L&M are evolving but in general you'd want to use long structured or at least use the better more expensive models in my experience and then for the last one we have decision- making which of course is the most common use

case for uh AI agents which I'll get to in a second they will have to break down queries plan out actions to be taken orchestrate and delegate tasks between tools and sub agents now this is one of the hardest tasks for L&M and therefore we should always use the agent prompting framework and give it very detailed instructions and as this decision making requires ERS planning and complex tasks these are in my opinion ideal for system 2 level thinking language models and therefore I predict we'll see most agent systems working best with these types of models

right like the gp01 model but even if you don't use those or you don't have access to them they always use agents with the best models possible now besides good prompt engineering for these AI systems to become reliable it is also key we limit the amount of responsibilities for one language model or AI step as much as possible L&M struggle most when we try to give it multiple tasks inside of one prompt so we always want to reduce it down to one specific small task this is one of the most common mistakes I see beginners

make right they try to let one language model do too much and too complex things and the system starts to break so what do we do then when we have complex tasks that's where chain prompting comes into play and with chain prompting we basically break down a larger complex task into smaller subtasks for each language model or AI step and chain them together to perform the entire task we use the output of the one model as an input to the next model and this makes an AI system perform complex tasks in a reliable and consistent

way I think a good phrase that I recently heard which I think can help you when you're designing optimizing these systems or when you don't get the outcomes or the performance out of your system that you want to is the solution to the problems of AI is usually more AI if you don't get the right outputs the solution is usually adding more steps in your chain of prompts or in the case of Agents adding in more tools or sub agents lastly it is very important to understand the language models that are available to you and

to understand the strengths and weaknesses of each one to get the most reliable and cost effective system now we can broadly Define two types of language models we have the system one non- reasoning language models and we have the system two reasoning language models now beneath each of these categories we have two subcategories we have the cheaper mini models and the more expensive models now for the system one level thinking language models uh which are generally the cheaper and faster models these models are generally great for these simpler use cases and to keep the cost

low each of the language model providers has their own types of these language models like in open AI generally uh these are the mini models like GPT 40 mini and then the second subcategory is the more expensive models that generate generally better and more reliable outputs for more complex tasks like GPT 4 o and CLA uh 3.5 and then we have the system 2 reasoning language models these models have reasoning built in and can therefore handle more complex tasks like math coding and advanced planning right these systems do take more time and use significantly more

gpus and are therefore significantly more expensive so we don't really want to start using these models for simple straightforward tasks but these models can become really interesting for these more Dynamic and complex tasks that require High R reliability like for agents for example we have again two different types of these models we have the cheaper models now these models are the Mini version but as they're reasoning models they're still more expensive than many of the main or expensive system one thinking models and then lastly we have the most expensive models that use the highest amount

of gpus models like gpt1 and gpt3 which was recently released by open AI now again these models should only be used when we try to do really complex things or when we need very high precision and I also think again that agent systems might be very interesting for these types of models now the second main scale is AI workflow automation now this is the scale where you really start blowing people's minds and is the most practical scale and maybe the lower barrier to entry way to start really leveraging AI for real business impact now what

do I mean with AI workflow automation this is the more traditional way of Automation and automates tasks and processes based on on logic and predefined rules this differs from AI agents as AI agents don't have deterministic logic when they make decisions but it's actually what I think most people should learn before learning AI agents and is what businesses most of the time actually need to automate their everyday processes in workflow automations we connect different softwares and databases with apis so we can send transform and receive data through those softwares this way of automation has been

used for decades but with language models and AI these systems have become 10 times more powerful as we can now actually analyze transform and take action on the data before outputting it into the next software and this basically allows us to automate and enhance a lot of tasks that humans do and what automation traditionally could not solve for example in the pre-ai era popular automation workflows were aad filled out aite form and that data gets automatically added to the company CRM Now by simply adding an AI step or a language model between we can do

vastly more powerful stuff we can research the lead uh before actually outputting it into the CRM but we can make it even more sophisticated but we add in another AI module to do the lead scoring right and we can then add another one to actually call the lead if they're qualified or send a personalized email and learning how to do this really is not rocket science with these no code platforms like make.com anden pip dream anyone can really learn how to set up these really powerful AI workflow automations in weeks without knowing how to code

but even though it's straightforward there is a little bit of a learning curve involved and most people simply will never go through that and therefore if you learn this you will really be part of the 1% AI experts with only learning this the main practical things you need to know and understand for AI workflow automations are first analyzing and mapping processes right the second is apis and web hooks then we have data types and transformation and lastly testing and error handling just the fun fals of these concepts are usually enough to build your first automations

and the way you learn most is by actually building some automations but I'll give you a very quick breakdown on these five fundamental skills first for analyzing and mapping processes barely anyone ever talks about this but this is arguably the most important skill which is understanding a current process or workflow in a business and then actually mapping it out in a process that can be automated this might seem easy but the reality is most businesses and people are messy most businesses have no detailed Sops and most business owners can't even completely describe a task or

workflow to you in detail themselves so your job is usually to First really understand the task understand the process the desired output and then map and design the automation now this is again another reason why domain expertise and existing business experience will be so valuable in this process because you understand business needs and can inject your knowledge in designing and optimizing these systems and processes now the easiest way to do this is by mapping out how the process should work through a d diagramming software I personally use fig Jam some people like whimsic Co but

there are tons of others out there and I just map out what the steps in the process should be and what the inputs and desired outputs in each of the steps should be now it's really important to do this before actually jumping into the automation because you can save a lot of time like this then the second one is apis and web hooks now this is where many of the non-coders immediately throw in the towel but trust me apis are really easy and with NOCO platforms it becomes even easier and most of the time we

don't even have to set up uh custom API or HTTP requests anymore as many of these softwares have native integration set up with the most popular softwares so very quickly what exactly are apis apis or application programming interfaces are what allow different softwares to share data and work together they make automations possible by enabling information to flow between tools think of it like giving apps a structured way to talk to each other so to understand apis better I'll give you a quick uh example of creating a contact in hopspot right through an API call so

first we have endpoints and endpoints are like specific addresses where the API is accessed and each endpoint represents a specific function or resource for instance to create a contact in hopspot the endpoint could look something like this post hopspot ai.com conts slv1 SL contact now through the endpoint we basically tell hopspot hey I want to create a new contact these endpoints for different actions are usually really easy to find in in any apps documentation page then we have HTTP methods apis use these methods to Define what kind of action you're taking the most common ones

are get and it's usually used to retrieve data from a software for example pulling a contacts details then we have post which is usually used to create create something new in the software like our contact in hotspot in the example then we have put or patch which is usually used to update existing data or to edit you know a contact's details and then we have delete right to remove data now in our example we using post because we're adding a new contact then we have headers headers carry meta data about the request uh think of

them like instructions for the API on how to process the request right so for instance the most common ones you'll see are authorization right and this tells the API who you are right and it's usually done through an API key or token and to check if you're actually allowed to send data to the software right and then the second one you'll see very commonly is content type and this specifies basically the format of the data that's being sent to the software right so in most cases that will be application Json now for hopspot uh for

example the headers might look something like this header one would be authorization and there you'll have Bearer which is commonly used before adding your specific API key and the second header might be content type and where we specify the type of content that's being sent or the structure which is going to be application Json then lastly we have the request body and the request body contains the actual data you're going to send to the software so for creating a context in hopspot this might look something like this so we have the Json format with the

properties we have the property key email uh first name and last name and the value use where we put the data we want to send to those properties this is the information we're telling hotspot to use when creating the contact it's straightforward it's just a list of fields and and their values right now what about web hooks so while apis are about requesting and sending data web hooks are event driven they basically listen for triggers for example someone fails out of form and then the information automatically gets sent to another tool so it's like a

notification system between apps now the great thing about these popular apps like hopspot is that platforms like make.com relevance SII NAD often handle all of this for you already they have already set up a native integration with these softwares so you don't even need to do any of this or even look at the API documentation uh in most cases but if you have an app that uh is not directly integrated it's still not rocket science as you see uh and these Basics will make it really easy to set up your own apis now then we

have data types and transformation so we also need to know the main data types the structures and how to actually transform them now this is usually sort of the first types of Errors people face when starting with automations arrays and Json uh sound intimidating to non-coders but again it's really not rocket science and these no code platforms again have all the features to transform and interpret these data types in an easy way all at Json is a lightweight format for storing and exchanging structure data right it's used in almost every modern API to transfer info

to so all it is is data wrapped in curly brackets right with keys like email and their corresponding values like in example John do at example.com arrays are not a common structure and they're just a way of grouping multiple items together for example picture a list of products or email addresses or task IDs you need to process one by one arrays basically make this really easy for example a list of emails uh might look uh something like this and as you can see arrays are formed with a square brackets and again if you're using these

no code platforms you can easily Loop through these lists or arrays to handle each item automatically whether it's it's uh sending emails updating new records or anything else then we have objects and nested structures Then There are a bit more complex data structures like objects and nested data now these might seem a little trickier at first but they're just combinations usually of Jon and arrays for example as you can see here the items key contains an array of objects now again might look complicated but once you understand the Json and arays the patterns become pretty

clear once you know this you'll know it for life and think of it like this this stuff which really is not rocket science again and can be learned in weeks separates you from the 99% that will always see this as too technical for them then you also need to understand some about database management it's a an important skill because almost every automation or system you create needs a place to store retrieve and update information right so for example tracking leads managing tasks or syncing data across platforms right usually databases are are sort of like the

backbone right of your automations now again not rocket tiance you don't usually have to dive into complex SQL queries or become a database expert because no code tools like air table or or even Google Sheets can handle most most of uh the heavy lifting for example like air table like a supercharged uh spreadsheet that combines the Simplicity of a table with the power of a database right it allows you to link records filter data and Trigger automat automations directly from tables plus air table can really be used as a front-end interface or dashboard uh giving

your clients and users an easy way to interact with the system and Trigger automations Google Sheets work works perfectly fine for many use cases too then lastly we have testing and error handling now again not the one that's most often talked about because it's definitely not the sexiest but it's definitely what you need to learn to actually deploy a automations in businesses no automation is going to be perfect on the first try which is why testing and error handling are essential skills to learn testing basically ensures your workflow behaves as expected under different scenarios right

while error handling makes sure the system doesn't break when something unexpected happens right some tips for test testing are always test with some edge cases right usually people design these systems for perfect scenarios but the real world is messy and error handling uh makes sure the system doesn't break when something unexpected happens right for example API calls fail quite often sometimes you have data formatting issues uh language model hallucinations now most of these nood platforms offer built-in tools for testing and debugging so you can simulate your workflows and troubleshoot any issues you can also add

error handling steps like retrying failed actions sending notification when something goes wrong or logging errors for later review and lastly I want to say that many people when they start out usually get frustrated that mainly because of one of the reasons above or not understanding one of the things above what I usually try to tell people there is that the problems you're facing now means you're basically mastering these fundamentals of Automation and that overcoming these issues is exactly what will get you in the 1% because most people when they see those errors they don't pull

through and once you've seen these errors trust me you have seen them all and will know exactly how to solve them in the future and the last fundamental scale is AI agent building so first of all what are AI agents what is the difference between AI workflow Automation and when would we actually need to use AI agents instead of workflow automation so in contrary to workflow automation where different L&M apis and tools are orchestrated through predefined logic or rules agents or systems where L&M or large language models take charge over the workflow and decide themselves

what to do to accomplish a task they usually consists of four fundamental Parts at the core we have a large language model which is like the brain of our agent then we have tools memory and knowledge now for the first one as I said there's an L&M that is in charge of the workflow and is like the brain and plans out what to do to achieve a task this L&M is usually in the form of an agent prompt which I talked about before right that defines his actions and his goals now they can also have

access to tools which can perform uh actions either in softwares to send or retrieve data or tools can even include prompt chains or even the entire workflow automation now the third one is memory because our agents need memory to remember the past conversations and to know what past actions it has taken and lastly agents can have access to knowledge usually in the form of rag now this is used when our agents need access to a large amount of context uh that doesn't fit within a normal context window or prompt rag stands for retrieval augmented generation

and although it sounds complicated in simpler terms all it really does it breaks down a large amount of context into very small chunks uses uh an embedding model to convert these chunks into Vector embeddings which are stored into a vector database and our agent can then query that Vector database much like a Google search to search for semantic similarity so when our agent queries the database with a question like what's our company's policy on sick leave the vector database will retrieve the most relevant chunks for that query and that information will be sent back to

our agent who can use that information to respond respond to a question of the user for example so how does an agent actually use tools I'll just give you a quick example if you want to make research agents that has access to a tool to do Google Search and a tool to scrape websites we can now ask our agent with a query like in do in-depth research on how AI agents work and write a full research report because our agent knows what tools it has access to it can now plan out how to achieve this

task himself for example he might do a Google search first to find relevant blog articles with more information on the topic then use the second tool to scrape those websites using the web scraper and then write the research report based on what he got back right so instead of predefined rules or logic our agents decide what to do themselves to achieve its goal so when and why would we use multi-agent systems then first of all because at the moment single AI agents or large language models Excel when we give them single and easier tasks right

they become more unreliable and start to struggle when we give them too many responsibilities or very complex workflows when we are designing AI agent automation systems we always want to optimize each N&M or AI agent for single tasks right and simpler workflows so knowing that if we want to automate more complex workflows we can use multi-agent systems where each AI agent is responsible for a single task or outcome each equipped with tools and knowledge to perform their specific task for example have a manager agent who uses their sub agents to ex execute a more complex

workflow much like how a single agent would use his tools so let's say for example that besides research we also want our system uh to be able to write and post blog post on my website and also to be able to write LinkedIn post and add them to my content calendar so our research agent is already optimized to do research so we don't also want him in charge of writing blog post and Linkedin post so we can add a blog post agent to our system who's optimized for writing blog post and a LinkedIn agent that's

optimized for writing LinkedIn uh content and the blog agent we can give a tool to actually post the blog post on my website and the LinkedIn agent we can give a tool to add it to my content calendar now we add a manager agent to connect all of these agents together into a team and now we can handle a more complex workflow for example now we can say something like please write a blog post and a LinkedIn post on how to build an AI agent and add them to my website and content calendar now our

manager agent will divide this task into subtask and delegate each subtask to a sub agent through a prompt so he might say to the research agent first do research on how to build AI agents then the research results will be sent to the blog agent which he prompts to write a blog article around this and post it to my website and the same for the LinkedIn agent this framework can be expanded into more and more layers of Agents being able to create more and more complex systems that can now handle more and more complex workflows

so when should you decide to use a multi-agent system versus a single agent now take this with a grain of salt because it is nuanced and it depends on a specific use case but a general rule of thumb is to First always optimize each agent for a specific task never use more than seven or eight tools inside of One agent never more than seven or eight sub agents per manager agent and manager agent responsibilities should always be limited to team management and not to performing actions themselves a good framework again to think about these systems

is when you see your system struggle the solution to the problems of AI is usually more AI so inside of these systems that usually means more AI agents or more tools so when should we actually use AI agent systems instead of AI workflow automations and what are the advantages and disadvantages of each one now this is another important framework to keep in mind when you're designing automations right so when building automations with L&M or AI you should always find the simplest solution possible and only increase complexity when needed now I am to blame a bit

here because most of my content evolves around AI agents but in practice this means that for most of the business use cases we don't necessarily need AI agents but we can rely on AI workflow automations in my agency I would say that 75% of our automations are workflow automations and 25% are AI agent automations and this also has a lot to do with the current limitations we still face with AI agents they're still developing and improving week by week so what are the advantages and disadvantages of each one now ai agents Excel when you're dealing

with Dynamic complex inputs or when you're solving open-ended problems where the steps aren't always as predictable they're also quicker to deploy usually because they require less logic based setup but they're also more expensive and more prone to errors these systems also often need more human oversight to ensure that they're working as they should be working but on the good side AI agents are usually also very reusable so once you build an agent uh you can reuse that agent inside of new agent teams uh usually very easily now for workflow automations these are very good for

predictable repetitive workflows where inputs and processes are more consistent so it's more reliable in general it has fewer errors and is significantly cheaper to run however setting up a workflow automation takes usually a lot more time because it requires mapping out logic and rules for every step now once it's run running of course it requires a lot less human intervention and making it yeah a very efficient way to do automations for day-to-day business processes so what are the current best use cases for AI agents now besides coding agents which I get to in a second

and computer control agents there are two main use cases for AI agents for businesses right now now the first wave of business applications of AI agents that is already happening right now is customer facing AI agents these are text or voice-based agents that are deployed on behalf of companies to interact with their customers for example for customer service or sales Mark Zuckerberg also predicts that every company will have an AI agent or multiple AI agents that interacts with their customers in the upcoming years in my own agency we've implemented multiple of these types of AI

agents over the last year on either WhatsApp on websites and through voice agents on on the phone now ai agents are already the preferred option here over traditional workflow automations as interacting with customers and humans is inherently Dynamic and requires this memory and decision making every interaction is unpredictable and unique so the agent can decide on the best steps to take for each of the conversations therefore AI agents are generally a better use case for these customer facing applications now the second business use case for AI agents is for internal process automation now we are

still in the early stages of AI agents for these types of processes because most companies people are just starting with automation right most companies have done nothing with it yet and generally the best place to start uh with companies is usually to start with repetitive well-defined processes and as they are well- defined and predictable workflow automations are usually a better option for many of these internal business processes now why do I talk so much about AI agents then it's because I think slowly but surely when these AI models and large language models become smarter they

will slowly start to take over more and more of traditional workflow automation if you've seen my last video about the 20 AI agent team I think soon we can use these simple English sentences to talk to agents to automate entire workflows instead of having to program and map out these really large logic based workflow automations we can have an agent that already has the tools to interact with all of our softwares we can give it a query and it can automate complex workflows without needing to build out this rocket ship automation workflow anymore but again

at the moment right LMS are still developing and therefore I believe workflow automation will still be extremely important and a very important skill to know in the upcoming years because in the end businesses need automation very soon and they need these systems to be as reliable and cost efficient as possible lastly I just want to go over AI coding agents which is definitely interesting to start experimenting with in 2025 AI coding assistants like cursor lovable and bolt are developing extremely fast and now allow everyday people to start building applications and softwares with no coding background

that being said there is definitely a learning curve involved knowing how to use these tools requires time and effort it is a very powerful skill to have right but I wouldn't recommend to jump into this as your first sort of skill why because the most important thing when you're building software as an application is to know what to build and unless you already have an idea of a SAS or application you want to build or you have lots of experience in the AI space or the SAS space access to a lot of capital or a

big distribution Channel launching a successful app is really hard and requires a long and sustained effort which again prevents you also from learning the other essential skills that being said there is definitely a wave of solopreneurs building these micras and small apps that are being really successful however again the biggest reason for those guys success is usually they have business experience a lot of business experience in the past or they have a large social following and distribution Channel meaning they have an immediate audience and target market for the software as they launch which brings me

to the final section and maybe the most important one marketing and client acquisition now most of the people who've done a business over the last year uh will tell you the same biggest bottleneck is usually client acquisition this is the condition for any of these business models to actually work and it is hard and it will only get harder again like I said before the mode or the differentiator of businesses and people has traditionally been a lot more on the product side 20 years ago 30 years ago great product s sold themselves well not anymore

can have the best possible product in the market if you don't have a way to reach your customers it is worthless and as social media and the software world has developed this differentiator or mode has shifted more and more into distribution you don't necessarily have to have the best product on the market anymore as long as you have an audience and a strong distribution Channel and a product that solves a need and as AI is democratizing Tech even more slowly anyone will be able to build these powerful applications and softwares and that's why it will

the mode will shift more and more into deliverability instead of only the product or service so building an acquisition Channel where you can consistently get clients is really key and should at the beginning especially if you don't have a large Network or audience yet be one of your priorities if not the main priority now there are a thousand and one ways of doing marketing and getting clients but I'll just go over the main ones that I think are most interesting and that I've seen working best in the AI wave right now and before I dive

in I come from a marketing background and one thing I learned from a wi combinator talk a few years ago that's stuck with me and I think is good to realize for other people doing this is that most unicorns grow to be a billion dollar Company by being really good at one acquisition Channel and I see a lot of people making a mistake to try to do a little bit of everything it won't work all acquisition channels are hard and you need to be consistent iterate and improve continuously to become one of the top experts

in an acquisition channel to get results it's almost like a business on it on its own that's how I look at it that's why generally it's good to stick and focus on one at least in the beginning so the first one is building a personal brand now I'm biased of course but I think there's a gigantic opportunity and by far the best way to get distribution and acquisition in the AI space right now is by building your personal brand I think there's a window of opportunity to capitalize on building a personal brand around AI because

of three reasons right first of all because AI is new and all these softwares and Technologies are new you can still position yourself as one of the top experts in a specific AI field secondly there's a general Trend to more personal connections over company or brand connections and thirdly AI content at the moment is being pushed really hard in all the algorithms on all the social channels now which channels can you use of course YouTube in my opinion is great because first of all you make the strongest connection with your audience of all the platforms

by having long form content you really build trust first of all by them watching you longer and also you show what you sell if you show AI demos or systems you've built you immediately build trust because you show what you sell then of course like I said the AI content on YouTube is being pushed really hard YouTube is the second largest search engine which a lot of people forget and for these AI systems maybe the first it attracts lots of business owners and professionals my D main demographic for example is male between 35 and 44

in the US and lastly it's cliche but we do live in an attention economy and people value seeing real people more than ever and it will not only get you client acquisition but also for me personally uh I've met some amazing people and Friends through my YouTube channel now I know it's not for everyone and I thought it wasn't for me but let me help you by addressing some common concerns which I had too when I started first of all nobody cares I know there are so many people that want to try this but don't

do it because it might be cringy or embarrassing to family friends or or colleagues first of all when you're just starting you will not get more than a couple of hundred of views maximum and nobody will find you second of all none of these people unless they're actually in AI will actually watch your videos I know that none of my close Circle has actually properly watched one of my videos second you don't need thousands of views or thousands of subscribers to get leads from YouTube right this is one of the biggest lessons I learned and

that really convinced me to double down on this right I thought I would have to invest years and need thousands of views to get clients and I started getting my first lead after a few hundred views on my third video I think so the YouTube algorithm is made to push your content in front of people who are looking for the exact content you bring out right that's the magic of it thirdly you don't need to be the absolute best or the 1% expert before being able to post on YouTube you just have to be one

step ahead of your target audience the majority of people has done absolutely nothing with this yet you will be one step ahead of the majority of people in fact if you're tooo far ahead sometimes you even lose some of your target audience and only talk to people who are deeper inside of the AI space right instead of actual you know potential clients so maybe those Frameworks can help you take the first step now you can also try on a platform like LinkedIn I'm personally trying to expand my brand a bit on LinkedIn and I've seen

some good results but as YouTube is my main focus I am definitely not an expert uh however there are many people in the AI space right now getting amazing results through Linkedin organic or LinkedIn Outreach right or a combination of them Jake a friend of mine for example with an automation agency is getting 10 plus leads a week consistently from LinkedIn organic only and Alex a member of my community also reached 100 lead bookings with LinkedIn Outreach automations now post that at the moment seem to get a lot of engagement on LinkedIn are first of

all demos right of these AI systems you've built it's the same for YouTube by the way then you have this AI workflow and AI agent giveaways uh which with these comments that getting lots of tractions too and then you have these explainers or diagrams that seem to work uh very well too then X is another platform that gets tons of traction right now with AI content prow is a great example also a member of my community he creates MVPs for businesses with coding agents and he made more than $50,000 in less than three months only

with leads he got through X he does basically uh build in public type videos and gets tons of businesses reaching out to build these MVPs for them now you can also try shorts on Instagram Tik Tok and there are even people getting traction through written content sites like medium but most importantly again try to focus on one of these when starting out focus on providing value and stick to it that's the most important then cold email is also still a great Channel if you focus on it and get good in it with AI personalization you

can now get really high reply rates if you do it the right way we recently had a master class in my community with t who was the founder of slight Edge agency and does over seven figures annual revenue doing cold email Outreach right and he explained in detail how to set up uh email campaigns how to do personalization the right way so if you want to learn more make sure to check it out but yeah C email could definitely be an interesting strategy as you can get direct results and the personal brand might take a

little bit longer now many of the AI communities right now including mine also have many business owners and Founders in there that are looking to adopt AI but don't necessarily have the time themselves to do it so we've actually had multiple client relationships build to our community so being active in these communities right showing what you can responding to people's issues and reaching out to DMS on all of these different communities can definitely be a great strategy for initial client acquisition then we have cold calling uh now this makes more sense for the people with

a sales background uh but cold calling still works if done right and if you can stom make the rejection rate I've done some uh some power dialing uh back in the days but it's not really my thing but for example who I mentioned before selling voice agents for hotels right they get great results called calling clients now event speaking or conferences can also be a great way to find leads the same example yonos right with his voice agent solution he also speaks regularly at AI events and actually gets quite a lot of leads from them

then we have the freelancer platforms like Fiverr upwork freelancer there's a bunch more now I think there's still a window of opportunity on this channel to position yourself as an AI automation specialist as we're sort of still early however these platforms definitely do take time and efforts to consistently get clients from right I'm personally not a huge fan of these platforms as you're competing in a big Marketplace and people are constantly trying to undercut you but for some people and the top profiles on these platforms it can become an amazingly consistent acquisition Channel then we

have ads I haven't run ads myself yet as I haven't needed to but I think there's definitely a big opportunity here too right running Instagram LinkedIn YouTube Facebook ads to lead magnets for example can definitely be an interesting strategy we'll soon have another master class in my community with mosche who's an ad specialist who will explain how to do ads the right way however this of course will cost some money up front so I would more recommend it to people who maybe have some experience in the ad space or have some money to invest now

of course there are tons of other ways to find leads and and get clients whatever your channel will be it is vital to make this one of your main priorities this is probably the biggest reason most companies fail so to finish it off knowing all of this how should you actually get started today first of all and most importantly make a commitment this is the moment we are still early and trust me if I say if you jump into this right now you will be one of the top 1% experts still right you can establish

yourself as an AI expert and whatever way you want to go right being an employee starting a business you will become immensely valuable in the marketplace in the upcoming years and the best way to learn in this space is through practice right it's the only way to be honest so step number one is pick one or two automations that are useful to yourself and start building them out maybe get a template from someone on LinkedIn or some community and try to rebuild it according to your your needs in the process you will run into issues

can be painful in the beginning but this is where most people give up and where you separate yourself join maybe some communities uh there are lots of free ones out there too where people can help you out because most people have already run into the issues you will face and you are facing and honestly I can say that if you dedicate yourself to this you will have the fundamentals in a matter of months if not weeks and if you are employed you can try to position yourself inside of your company as an entrepreneur right as

the automation or AI specialist inside of your own company if you have existing clients or you're maybe a consultant an agency owner freelancer try adding AI into your existing Services most companies have done nothing with it yet and you will blow their mind right even with simple automations and you get real client experience and see what it takes to actually deploy these kinds of systems inside of companies you can also try to build some free automations for your network for your friends colleagues Etc and most importantly while doing this share everything you do and learn

on a social channel to build your personal brand and you will see that it picks up again building a personal brand or acquisition channel is just as important if not more important to get successful in this space then once you get your first client it will be scary because the nature of these Solutions is that every company will need a slightly different solution and most likely you have never done that specific automation yet but don't worry just jump in try to make it work and in the worst case scenario you wire them back their money

that's it on a last note I really really hope you jump onto this because opportunities like these only come around every Century I think you are not too late in fact you are still early if you jump into this right now don't be the person that looks back and thinks what if now if you enjoyed this video I highly appreciate it if you can like it and maybe subscribe to my channel if you want to join my community I'll also make sure to put the link in the description below too thank you so much for

watching and I hope to see you in the next one