In 2025, it will take you longer to read something and comprehend it than the amount of time it took to create it. So, if you think finding the signal amidst all that noise is only going to get worse with all the AI-generated regurgitation of posts on LinkedIn and all the SEO-optimized blog posts that don't actually contain anything valuable—well, you're right. We have to move up the knowledge work value chain, and luckily, we can do it with AI as well.
Let's come back to the fundamentals of knowledge work and find where we can gain leverage. Knowledge work can be broken down into four steps. First, we have our input—data, knowledge, and information that we gather.
Then, we process it to get some sort of output. From the feedback of the output, we then improve our input. Over time, we're going to spiral upwards and improve the quality of our knowledge work.
Now, most people spend the majority of their time thinking about how they can outsource their thinking process to AI. They are using tools like ChatGPT and saying, "Okay, you write this for me; you do that for me. " However, we realize the results are not particularly fantastic.
Why is that? It's because the input that large language models (LLMs) have is pretty average. We're averaging down to the level of the Internet; there's some gold in it, but the majority is pretty mediocre.
Instead of focusing on the process, we want to focus on using AI to improve our inputs. Even if nothing changes in the way we process information, our outputs will be better because we won't have this "garbage in, garbage out" problem. Let's concentrate our efforts on using AI to help us improve our inputs.
The addition we are going to make here is to incorporate academic papers and research into our input. Now, you might think, "First of all, that sounds boring; second of all, this is practical work. " This is not some sort of academic research that we're conducting; we need practical outcomes.
But think about it: Do you read the Harvard Business Review, McKinsey Quarterly, or the MIT Sloan Review? All of these aim to take boring, hard, complex academic papers and make them relevant to working professionals, delivering insights in a way that allows us to extract something actionable from them. We need to realize that we don’t want to be readers of HBR.
We read so slowly. If you just feed the entire archive to ChatGPT, it will immediately know so much more than you. We don’t even want to be the writers for HBR because people spend their entire lives writing a few insights.
Instead, you can leverage those insights. Why spend the time researching? What you want to be is HBR itself for your team or organization.
This means going through trusted, complex, and nuanced resources, translating them into the context you’re in, creating frameworks that turn theoretical knowledge and insights into something actionable, and delivering it in a way that busy people can understand. We need to move up the value chain and look at things that used to be reserved for academics, scientists, and authors who had the resources, time, and brainpower to sift through long-winded papers. Now you can too.
Let’s take an example, and let me show you the three tools that I use to ensure I become the HBR for my team and organization. Let’s use an example of one of the biggest challenges that managers face: the problem of a team underperforming. There are bright people, but they seem like they don’t want to do anything; they’re always complaining about stuff.
How do I improve team performance? Usually, if you get average input, you might go to ChatGPT and ask it this question, which will give you some average results. You might go to Google, and it might give you some average inputs.
Let’s give it a try. “How to improve team performance at work? ” Common question, right?
Oh gosh, it tells you to focus on goals and open communication. These are just such empty, broad recommendations that are not going to lead to great output. What I like to do instead is to use the first tool of today: Elicit.
This tool helps you find academic papers related to the question you ask. So let’s copy and paste this and see what kind of results we get in Elicit. It's free; you just have to sign up, and you’ll get something like this: a summary of the top four papers for free, along with a list of relevant papers.
Now, I like to ensure that what I’m reading is well-cited and peer-reviewed, not just some random paper. I sort by the most cited. For example, "team design features and team performance" seems to show a relationship.
Let’s take a look at this one. What you can do is click into Semantic Scholar, which often leads you to the actual paper itself. In this case, it is locked, so if you don’t have access to a university subscription or something, you might not be able to see it.
No problem; we’ll just copy and paste the paper name and go into Google to see if we can find a PDF version. Here we go: here’s the PDF published by Sage. Let’s double-check—it seems to be the right one.
Okay, we’ll download this, and then we’ll use the second tool of the day: Notebook LM by Google. Unlike ChatGPT or Claude, this tool isn’t using AI on knowledge from the internet; it only uses sources that you upload. So, you have a lower likelihood of hallucination.
Sign up with your Google account, and you’ll see something like this. You want to create a notebook, which essentially just houses your sources. Since we’ve downloaded a PDF, we’re going to upload it here.
Now, we haven’t even read the paper; we don’t really know what’s in there. So, what I like to do is use it as a table of contents first, which is automatically generated by Notebook LM. Now we have a table of contents.
"Groups" is important; there’s classification theory. It looks like we have to look at characteristics, some team size, team level, task design, and leadership. Okay, interesting.
Now, I don’t want to read this yet; it’s a lot of text and really poorly formatted. But you know, Notebook LM is still pretty experimental, so we’ll give them some time to figure this out. Now that we have a general idea, let’s click back on the Notebook guide and look at the suggested questions.
Since we haven’t read it, but AI has processed it, I usually like to use one of the questions just to get a basic gist. Here we go: “How do team design features impact overall team performance? ” Let’s see what it says.
Looks like there are about three components, and we saw that in the table of contents as well: team composition, task design, and leadership. I love this about Notebook LM. You can click into this to find the actual place in the source where it discusses this idea.
So, even though this giant paper is probably tens of thousands of words, I can identify which parts are most relevant to me and actually find the citation so I can better understand what they say. Let’s say, for example, that I can’t change the team composition—I’m not going to fire anyone or hire anyone new—but I want to focus on task design and leadership. I can prioritize these cited areas first.
That’s the first cool part about the Elicit to Notebook LM workflow, but this is only one paper. What I would do again is go back to Elicit and now see if there are any other interesting papers. “Leaders can learn about teamwork by developing high-performance teams.
” That seems interesting. Let’s take a look and see if we can find the paper. Oh, here’s the PDF—fantastic!
Now we don’t have to. . .
oh no, okay, bad request. That’s all right; it happens. Let’s go back and check the publisher.
Oh, it’s a Wiley paper. Let’s see if we can get the PDF. All right, we can’t.
If you’re at work, surely your workplace will help subsidize getting this paper. But even if they don’t, let’s see if we can find it somewhere else. Let’s see if this is the right one.
Fantastic! Here’s the second PDF we got. We’ll head back to Notebook LM, and this time we’re going to press the plus sign and again upload it.
Once that’s done processing, you’ll see down here that now you can chat with both sources. But before we do that, I want to uncheck the previous one because I just want to understand what this new paper is about. So let me generate a table of contents for this one source, and we’ll see that it’s generating a new note.
Let’s wait for that. All right, here we go: there are five essentials to building high-performance teams: capable and committed members, norms, a team structure that has a clear mission, responsibility, goals, and a framework to get results. See, this is already so much clearer and more specific compared to the Google search results, which just got diluted over time by everyone—maybe one person starting from the source, but eventually secondary sources dilute the message.
Okay, a way to manage team processes, and that’s the fifth one. Great! Now let’s select both sources and ask, “What are the key changes a leader can make in a team to improve performance?
” Now it will look through both sources and give us an answer. All right, it says: “Focus on member selection. ” Let’s say that’s not relevant to us.
“Promote task meaningfulness”—this was mentioned in the first source. Great! “Empower the team with autonomy.
” I don’t know if we’ve seen this. Okay, this was in the first document. Great!
It gives us the exact things we should be doing: “Cultivate team coordination. ” All right, “Provide effective leadership,” which means guiding, motivating, and supporting the team. Let’s see the details.
Now, let’s say out of all of this, I think for my team the most sensible action is to promote task meaningfulness. I want to empower the team and facilitate team development, and of course, provide effective leadership. So, give me a detailed plan with three changes we can implement to improve team performance.
Now you have a high-level plan for doing this, but it still feels somewhat vague. Even though we know the input is good and we’re focusing on what academics have found out, we want to do the next part. We’re going to copy this and go into Claude and say, “Help me create a detailed plan to improve team performance using…” and I’m going to paste in what Notebook LM has shared with me.
In just two seconds, we got a 12-month, three-phase, four-module plan to improve team performance. Just think about it—if you were in HR or you were a team lead and you shared this plan with management, they would be blown away. But that’s not all we need.
Let’s say, out of all of this, it still sounds somewhat complicated. So let’s ask, “What are the 80/20 changes I can make to my team to drive performance that depend on leaders and not on the team members to make a change? ” I’ll ask it to refer to these two studies, which are called a “meta-analytic review” by Greg and by War.
Now Claude has a better idea of where I’m getting these insights from. Based on the details it knows about these two sources, it gives me the core leadership action plan. Okay, “Task design optimization” seems to be one of them, “Authority distribution” is the second highest impact, and “Contextual support.
” So now we’ve gone from a 12-month plan to a 3-month plan, and it gives us some expected results. Whether that’s true or not, we can wait and see. But it helps us list out things like key success factors and measurement frameworks.
Let’s say, “Help me create a matrix to better design tasks with clear goals and interdependence structures. ” Here we go! Now we’ve got a framework that we can use to assess goal clarity levels.
We can use interdependence levels, see where they intersect, and even show the team, “Hey, this is what I’m thinking about for changing how we think about tasks. What do you think? ” and get their feedback on it.
Now, if we simply ask Claude, “Help me with what I need to improve my team’s performance,” let’s see what we get. All right, it’s asking me a few questions. That’s actually already an improvement since previous versions of GPT.
Let’s say, “Work with teams. . .
" “That are underperforming and unmotivated. ” All right, and see, the plan it gives me is so much more vague. Right?
It’s “Communicate and provide feedback. ” I know this! “Having clear goals.
” I know this; I know all of this. But it doesn’t help. Help me create a framework to set goals.
Yes, SMART is good, and monitoring is good, but as you can see, because the input quality is not there, we’re stuck with these buzzwords that we hear all the time. I am excited and optimistic about the changes coming to knowledge work. We can critically think more; we can actually be more creative and associative.
Generalists will have a great time in this era of AI. Let me know how you feel about it. Check out this video on how I use AI to create McKinsey-style visuals, and I’ll see you in the next video.
Bye!
