How we're creating AI Implementation Strategies for our clients

today we're going to talk about kind of the way that we're maturing our service overall you know it kind of started out as like we're building single use case agents stuff like customer success agent or customer support agent or voice receptionist you we're getting all the meeting transcripts from sales calls and we are putting them in a vector database and doing rag on it creating reports whatever it is like single use case agents that like make a lot of sense and you can kind of point to a specific role in a company and say like

yeah that agent is supporting that role but obviously like as we've been working on these systems and you know some of my past videos kind of Point towards this I think that there's a top to bottom transformation that a lot of companies are going to one I think it's inevitable that they're going to do it and two they're interested in doing it right like these are conversations that we're having with prospects and clients about okay yeah it's cool like we can have immediate help right we can solve the paino like right away we're having more

and more conversations where it's like okay yeah like how do we just build a system where it's top to bottom AI agent first and we have that Foundation built and then we can go and hire more people or scale that system overall or whatever it is this is the direction we're going AI implementation across the entire organization rather than single use case for a single role conquering a single kind of like workflow the reason for this by the way is because you can kind of get this stuff off the shelf now there's tons of like

agent Builders but very easy platforms to kind of sign up for and solve that problem right like you can go and solve the AIS SDR problem one because it's easy to build but then two like realistically it's just about integrating in prompts right like integrating your email and CRM and and maybe calendar and then you know what's the what's the actual prompt behind it like little little micro SAS products that people are basically it's like an agent and they're building a whole product around it like those are are popping up everywhere now and we're aware

of it which is why like we're you know there's there's kind of seeing the forest from the trees there's a deeper approach here that I think a lot of businesses entrepreneurs executives are interested in exploring and kind of those single use case agents like while they might get the job done it's more of like do we really want to just bolt on a bunch of things to the business and you know a couple years down the road untangle it all because you know we didn't build it from the bottom up first right we didn't set

the foundation of like the data is coming in properly we're adding proper context to that data if we want to change something it's just a matter of changing the prompt rather than changing the entire way that it's kind of configured we're having these conversations and because of that we're now maturing our service and part of that process is upfront providing Consulting strategy development and being very clear about like you know what are your workflows how much time are you spending on these workflows how much money is that what's what are kind of like the bottlenecks

that are causing you not to scale overall what initiatives in the future do you want we we've been talking to one client who they want to franchise their business and it's really hard when all of your systems are kind of ad hoc and the knowledge is stuck in people's heads you know there's a lot of dependencies there and the data isn't streamlined in a way where you can kind of easily triage things out and so it's it's we have to take a bottom up approach there we can't just like bolt on an agent and expect

it to work what we're doing now as part of our sales process but then also just kind of like our agency in general is providing AI implementation strategy docks and wireframes up front you have an intro call with us we learned a bit about your business we learn about kind of like your approach the challenges your goals we qualify you you qualify us it's like hey you know this would be a good partnership we should definitely work together over the long term okay cool next is like a second call which is an hour depending on

how much we're talking about and how much we're talking but it's about an hour and we really dig into the details on okay what exactly is this workflow how many people are involved here what systems do we have to integrate with what are we expecting like the AI agent to do and where do we see this going like what future things can we build you know maybe we start with one Department one single use case and expand across the rest of the business because sometimes there's a butterfly effect you do one thing at the beginning

during the sales process and it it affects everything all the way down down the line and honestly it should right it should anyway we've been doing these AI implementation strategy documents I have a template here based on a real document that we actually cre created and let me just kind of walk through it so I'll set the scene this client came in first of all there's a big mess right the leads come in it takes a while to kind of figure out which account executive needs to talk to the lead there's no real firm process

around triaging effectively leads end up talking to the wrong people sometimes and so the first kind of problem to solve is like okay we need to just make sure that the leads who come in they're getting assigned to the right account executive so that a count executive can take action now the next step is okay instead of the count executive sending an email right trying to schedule the call or calling right away we'll have an agent do that instant calling instant trying to schedule the meeting so that way the county executive just sees the meeting

show up on their calendar and it's ready to go so this is the problem we're trying to solve so the overview is we're going to develop an AI driven solution for the client to improve lead engagement customer onboarding production and Order management so the customer onboarding production order management that will come later as part of the road map but first we're going to tackle lead engagement right and then we're going to work all the way down the client Journey custom AI studio will build maintain and Implement multiple AI agents to optimize these processes starting with

lead engagement this AI implementation strategy will cover our proposed solution for lead engagement and so for each new kind of like phase of this we we'll have a new kind of document for it but for now lead engagement and then we break down the actual agents and what they're going to do so inbound lead routing agent right how do we clean that that system up well we're going to Route the inbound leads to the appropriate account executive obviously in the actions it's it's going to analyze the lead data so analyze the form and then determine

like which AE should be assigned which account executive and then go ahead and update the record in CRM and then the Integrations our HubSpot for kind of the lead tracking right obviously and then the sales assistant agent is kind of the key agent here you know almost it's almost the case where this could be just a regular automation but there's a reason why we made it an agent uh and I'll get into that in a second but sales assistant agent so this one is maintain consistent engagement with new leads with the goal of scheduling sales

meetings with the appropriate AE okay it's going to send an instant engagement email to the new lead with the purpose of scheduling a meeting it'll send follow-ups let's say that they sent the instant engagement lead an answer we need to follow up it's going to reply to emails with relevant responses to schedule the meetings when appropriate right like hey yeah do you have time to meet on Monday no I don't really have time to meet on Monday what what are your what's your availability like on Tuesday it needs to be able to respond to that

kind of thing send follow-up emails to leads attempting to reschedule if a meeting is canel or or if or the email conversation asks the agent to do so so like let's say that you know one the lead canel the meeting we want to follow up right away hey why' you cancel do you need to reschedule something going on whatever A lot of times that's that's what happens that's why people cancel the second scenario is like okay what if what if the AE or the lead is like hey you know I know we had this meeting

scheduled on Tuesday but can we reschedule to like Wednesday well we want the agent to be able to go hey yeah let me check the calendar checks the calendar for availability and then comes back back to the lead saying hey yeah I found a couple times to do these work for you and then obvious Integrations here are going to be like HubSpot Gmail Google Calendar assuming that this is their teex stack which it isn't but I'm putting that there for now and then we put together the development plan now this is the tough part because

there's a lot of unknowns and you know we built a a calling receptionist that like schedules to call this was an accounting firm someone calls in they schedule a meeting with the accountant if it's a client and they like actually need to talk to a person or if it's anybody and they want to talk to one of the one of the guys who works there it transfers a call to them and then later it's going to cross check against CRM and then it's going to do a bunch of things but I'm telling you this because

we thought that that would take Max maybe 30 40 hours ended up taking like 80 85 90 hours and the bulk of the hours I would say 30% of the time was on the prompt and it makes sense right because the prompt is is kind of like the new way of programming where that 30% of the time would have been like actually writing code but now it's writing prompt and testing prompts and rewriting and testing again so you know you never really know how long the prompt engineering piece of it is going to is going

to take but yeah I mean so I basically say this to say the timelines on these is really really difficult to get right uh and the more that we build the more tools we have you know the more knowledge organizational knowledge we have and the faster we can go so now we're actually starting to pick up speed on some of these builds because we have a lot of stuff that we can kind of pull off the shelf and we have like stand processes around how we actually build these agents and you'll see in the wireframe

like kind of the way the framework that we we go at it from proposed development plan we break it down in like phases uh so first thing we're going to do is attack the lead routing and then here we just have points on like we we try to get as granular as possible about exactly what is happening so for for example set up HubSpot proper property for contact status we need a way for it to tell or at least identify like this is a lead this is somebody who uh needs need to get follow-up emails

you know there's a certain kind of cadence or way to engage with this person because they are marked as lead as opposed to like customer or engage lead most crms have this right but it's it's usually better if we can kind of just create our own properties for the agents to work off of rather than using pre-existing properties simply because like most people don't actually keep their CRM very clean and that's okay but you know if it is very clean and you want to work off your properties that's obviously better but you know a lot

of cases we kind of have to set up the kind of data on your platform first in order for the agents to like work effectively right and we're estimating like 5 to 10 hours it could be less could be a little bit more but it's usually not going to be more we try to be conservative about the estimates sales assistant agent so this is like okay once we once we crack this we know that the right account executives are being assigned to the right leads like we know that's good 100% awesome uh now we need

to actually like do something right so next thing we're going to do is set up the HubSpot trigger when a lead is assigned as an AE so like immediately right when the inbound lead routing agent marks it as a lead and has assigned it to the AE so those two kind of like checkpoints criteria points this agent will be triggered and it'll send an automatic email hey you you have time to to meet on Monday whatever it is you know part of this is we need to develop the prompt and we need to develop some

of the tooling for it to actually like send an email or schedule a calendar event and then we're going to build out more capability for the agent so this is kind of going to be the longer term process I feel like you know if we solve these two issues and we're we're pretty quick about it we're solving you know at least like 60% of the problem that they have which is the lead routing and just engaging with leads faster but how do we make this more robust how do we make it actually useful so we're

going to set up triggers like twice a day anybody who hasn't responded in the last 48 hours is going to get a follow-up email responded I mean like responded to like the initial email to try to set up the call uh we need to develop a prompt for that we need to use the email agent tool we need to develop more to the prompt for kind of like coordinating back and forth scheduling right calendar tool calendar actions re-engagement email so like this is for the scenario of like someone canceled how do we re-engage try to

get on the calendar and then we have some success metrics here right and these are going to get better over time so this document is kind of our you know we we've done it a few times now and and it works great and it's kind of a great Foundation but I want to make this better and a big piece of making it better is tying it to actual numbers so like how are we going to track lead conversion rate well we need to know that at the beginning and then CRM data management like are the

leads organized in the CRM like very effectively part of that is kind of like you ask the team at the end of it but lead conversion we we should be able to literally track the numbers on that to see the Improvement so this is what we provide after the strategy sessions right after that second call this is what we're doing we're working on this right here and then we're also working on this and right now I'm the one who's primarily doing this cuz you kind of need like you need like business workflow experience and knowledge

and then also some development knowledge as well to just know exactly like what's possible and when to do what but anyway so as you can see like we have the breakdown here that kind of goes through the capability set the triggers stuff like that and then we also have kind of this little Ledger notes section over here notes key Integrations The Ledger on like what exactly are you looking at when you're looking at these diagrams and then we have this broken out into different phases phase one 2 3 so I I'll go over this really

quick so inbound lead routing agent very similar content to what's on the document analyzes lead intake form identify which account executive should be assigned based on the predefined criteria and update the HubSpot record by adding the assigned AE the trigger is going to be like a new lead is added to HubSpot via specific sources so you know whenever someone comes from a specific form we'll know that right and then we'll be able to kind of autoark them as lead and HubSpot and then that will trigger the inbound routing agent and then I like to add

like future capabilities like how can we improve this system because realistically like if you're an automation guy you know that you could probably set up just a regular automation here and it doesn't need to go it doesn't need to be anything crazy why would we build it in the form of an agent rather than an automation well it's because we might be able to add future capabilities to it right so new lead comes in maybe we conduct research maybe the type of leads are different and we want to conduct research in different ways based on

the type of lead it is right so we're asking the agent to do a little bit more there maybe there's another capability that we could add but this is kind of the main one where you know how do we just get get it to do more and then we have the sales assistant agent so the capabilities here are pretty much the same right send instant like same as on the document send instant engagement email naturally coordinate scheduling by providing the A's available time asking the Le what time makes sense for them so instead of sending

kind of like cly link or a scheduling link we want it to be a natural coordination conversation send follow-ups to unresponsive leads create calendar events update calendar events there's probably a little bit more that I could add here but this gets the idea across on like what is the capability of this agent and then the triggers here are going to be varied so a new lead is assigned to an AE and HubSpot that should trigger it to send that instant email get HubSpot context so this is like scheduled right so like twice per day we're

going to check the last contacted property in hpot and see when was the last time that they were reached out to and did they respond essentially and like if it's more than 48 hours then we're going to send a followup right hey you know did you see my email so like if a new email comes in it's tied to someone who's marked as lead in HubSpot we want it to figure out if it needs to respond or not right like yeah I am available on Wednesday can you book us a time right it needs to

like be able to respond to that so new email and then if an event gets cancelled right this one might be a little bit tricky because sometimes what happens is like the account executive might have a conversation with the lead kind of outside of the agent like perview and they decided to get like cancel the meeting and so it got canceled but then the agent is responding to the lead saying like hey I saw this got canceled should we reschedule at a different time you know we don't want that to be the case so this

one might be a little bit tricky to kind of figure out the the context there which this is kind of like a microcosm of a larger thing about the agent database right the master contextual database that I've been talking about like if we have that set up properly then it doesn't matter then the context is there on like why it got cancelled now and you know it's not just about the database but it's about the data capture right we need to capture all of that data when the account executive is talking to a lead that

should be in the agent database that should be known right it shouldn't be outside of the data that we're able to capture but that's a bigger conversation that that we'll get into in a you know down the road future capabilties so this one we could add many many things to this right like post meeting follow-ups contract and proposal generation re-engaging old leads dedicated assistant per each account executive so each account executive might have their own sales assistant that's doing things for them right almost like an executive assistant but retrieves and sends marketing material or other

documents right so like how do we grab something out of one drive or Google Drive best practice in my opinion is to set up one folder for the agent to access and put all of your marketing material and stuff in there for it to go grab that stuff and use it send it as an attachment to you know whoever you're sending that stuff to you know continually updating HubSpot records and pipeline stages based on conversation history and meetings so have a conversation lead says like hey you know like give me give me until like next

month for us to like figure out budget approvals and whatever right like give me until next month let's push it off let's add a note into the hubs SWOT record for it and then let's change the stage from like you know active to in progress whatever your stages are right so here you can see that we we're essentially breaking down like what we're going to do initially and then outlining where we can go with this if you really wanted to keep going right which part of that qualifying process in the intro is is is trying

to trying to partner with clients who are in it for the long haul who want to grow with us and are really interested in just in implementing AI from like a bottom up approach like those are the people that we want to work with you know for obvious reasons right and then over here we obviously have this kind of like little section here to kind of give context to what you're seeing so in the notes it's saying you know for this one if it's scheduling for this specific AE it needs to have access to that

account Executives calendar it means the agent will need to access everyone's calendar and schedule events update events cancel events and available times for those specific calendars now I solved this issue with an agent that we're building now that's called like mallerie and and I'll show you guys that in in you know a future video but this was a note at the time that I made because wasn't sure exactly how to solve it anyway the agent will need a dedicated email address to send emails from this works better than using the user's email address the agent

will be the owner of the event and add the AE the AE will be able to change and update the event if needed so if the agent is making the events the agent becomes the owner and they can add the AE to it but it also works better when we in this scenario have it be a separate entity with its own like email account you know for obvious reasons because we kind of want it to function as an assistant rather than kind of working on on the account Executives behalf especially if like anything goes wrong

right like it it it says something weird or whatever right it you just want it to kind of be separated and and have it known like this is an AI assistant it's going to help us quite a bit but again it's an AI assistant right and we all know where AI is right now then we have some key Integrations and then we have this Ledger here which breaks down how we conceptualize our framework around like the agents so we kind of call the main agents super agents because you know we can add many capabilities to

it and the idea is that we add those capabilities by changing the prompt right if we just give the agent a series of tools then the only difference between what One agent can do versus another is really the prompt right and you know again I'm going to get into this later but you know had a mini breakthrough this weekend about how do we kind of standardize the process of building multi-agent structures and teams and you know this is kind of pointing in that direction and so like the way that we break it down is really

kind of the super agent as a whole then we have this breakdown here which is kind of the input data or the trigger we have the action which ultimately is going to be the prompt and then we have kind of the execution the tool agent right like how is it actually do the thing how's it actually sending the email well we need an email agent right or at least an API so let's look at the phases so we like to break these down into phases simply because if we get one thing wrong at the beginning

then the rest of it is not going to work so like you kind of have to crack it one step at a time essentially so phase one is how do we get the routing so we build this routing agent the trigger is new contact added to the CRM and marked as a lead and then you know this is what it's told to do right analyze the lead data and assign the correct account executive to it and we just need the HubSpot agent for that so that's phase one shouldn't take too too much time it depends

on the criteria set depends on how many variables there are and most of it most of the time spent on this is going to be on the prompting just making sure it's triaging effectively phase two is when we start building out the sales assistant agent so we're building on top of the previous phase so the sales assistant agent if you remember the first thing we're going to do is like okay now that the inbound routing agent has you know triaged effectively assigned an AE to to the HubSpot record what needs to happen well this agent

will get triggered and it's told to just send an instant email with the purpose of like scheduling the calendar event so it's just sending the email so if somebody responded it wouldn't be able to kind of book it at that time and that's okay because we just want to make sure that first of all that first email that's sent is like generated correctly and you know it works effectively essentially right so like once we do this it should be a pretty quick transition to phase three right and we need the calendar agent and the email

agent for this sales assistant phase three so this is where the rest of the kind of ability or capability set gets built out for this agent right so we have the first one and then we have these other three or other three I guess you could say so okay now we have the the first email being sent out cool now what if they don't respond well how do we solve that we need a trigger to go off twice a day to pull all of the leads that are one marus lead and then two have the

last contacted date which is like a property in h spot it it's more than 48 hours ago so okay cool 10 leads match that criteria what are we doing we're sending a follow-up email trying to schedule on the A's calendar it's just follow-ups right we're sending follow-ups what else is it going to do well if an email is received and we're going to trigger this via HubSpot I think you know for for certain reasons mainly because it's already going to filter by people who are in the CRM and then it's easier to kind of say

like okay you know email came in can we check if they're a lead just right away what's it going to do so it's going to reply with a relevant response so this is going to be heavy heavy prompt like you know we want to make it as simple as possible and straightforward as possible to prompt but realistically the way that you reply to an email or how you handle certain inbound emails it needs to be kind of spelled out and then we're also going to have it schedule or update the calendar event so this is

really giving it the ability to coordinate back and forth with somebody and schedule it right and there's there's some tricks to the trade here in terms of like you know the agent might send one email saying hey do you have some time to meet on Tuesday and then someone responds a day later saying no I I I have time I don't but I have time on Wednesday and the agent needs the context of the conversation to be able to like continue the conversation and to understand like where it's at and what it needs to do

right we can't just say hey you're a coordinating agent like you know handle like meeting coordination I mean maybe you can but like it's not going to be super reliable and then the calendar event is canceled right so what do we do when calendar events are canceled so we're really breaking this down and trying to be as precise as possible about what exactly is happening here you know because it's not a human we can't just say you know do this or that and you know whatever we have to we have to give it the right

tools and give it the right instructions and test those instructions uh so that way it does what we're we're expecting it to do right so this is kind of how we we wireframe it up and it's supposed to give you a visual on what's happening but then also give you the opportunity to give us feedback on hey you know you guys are way off here this is not exactly what I'm picturing or how I see this working it's not it doesn't align with my vision and doing this upfront really saves a lot of time on

the back end because it like I said it helps us align it sets a little bit of a road map on like where we're going and you know this is just kind of one project that we're doing you know this client has onboarding stuff that they want to get taken care of production stuff that they want to get taken care of order management stuff that they want to get taken care of like there are are overarching big systems and workflows that they have that they want to stream mind right and we're simply just starting with

the lead engagement piece right so you know this should be kind of a microcosm you know a micro example of like how we're going to handle the future stuff down the road right like once we've solved this now they know okay this is how they these guys work right this is like the typical timeline how long it takes this is how do I gauge personally right with these people and this is how they kind of like understand and conceptualize and build a framework around like attacking these problems and so it it helps like on the

next project the client understands that and we can kind of like close the gap of misunderstanding between each other um if there is a gap to begin with right so yeah anyway just to recap this is the direction we're going it's more consultative it's more bottom up it's more partnership it's AI transformation it's less single use case it's less onetime project it's more relationship building right over time so I think every business is is going this direction I think there's going to be there's probably going to be a lot of companies who pop up and

compete with us on this right like you know it's my conviction that every single company is going to be using AI I think it's inevitable just like every single company has a website every single company probably uses their phone every person uses their phone so AI adoption is an inevitability especially in the business Arena and the biggest thing that it can do is help with your operations your data flow internally and communication externally and internally right just gluing it all together streamlining that process it it it really should get to the point where information is

instant there is no like Gathering of information or doing the actual research it's humans like reading the information and engaging and building relationships and that's it like realistically or planning strategy and coming up with Sops you know and and you working with agents to help come up with those Sops but like realistically like the moving of data one place to another the pressing of buttons and the you know just updating the CRM and whatever all the things that people don't generally like when it comes to their job agents should be able to take care of

that and a big piece of that is the data side which is why this is an AI transformation not just implementation it's it's a complete transformation at the end of the day so anyway just want to give you guys an idea aide on the direction we're going in terms of Consulting and and service and the agency side of of things and in the next videos I'm going to show like actual agents and the super agent breakdown and one of the agents that I've built that I'm going to add capabilities to over time and one of

the videos I'll also just like live build a super agent that I haven't built yet but I've been meaning to build it because it's a huge paino for me personally and I imagine a paino for a lot of you so yeah that's pretty much it