My 17 Minute AI Workflow To Stand Out At Work

In 2025, it will take you longer to read something and comprehend it than the amount of time it took to create it. If you think finding the signal within all that noise is only going to get worse with all that AI-generated regurgitation of posts on LinkedIn and all those SEO-optimized blog posts that don't actually contain anything valuable, you're right. Therefore, we have to move up the knowledge work value chain. Luckily, we can do this with AI as well. Let's return to the fundamentals of knowledge work and find where we can gain leverage. Knowledge work can be broken down into four steps.

We have our input— the data, the knowledge, the information that we gather— and then we process it to produce some sort of output, right? From the feedback we receive on that output, we can then improve our input. Over time, we will spiral upwards and enhance the quality of our knowledge work. However, most people spend the majority of their time thinking about how they can outsource their thinking process to AI.

And so, we're using Claude and ChatGPT to say, "Okay, you write this for me; you do that for me." However, we realize that the results are not really fantastic. Why is that? It's because the inputs that Language Learning Models (LLMs) receive are pretty average, reflecting the general quality of the internet. While there is some valuable content, the majority is quite mediocre. Therefore, instead of focusing solely on the process, we want to use AI to improve our inputs. Even if nothing changes in the way we process information, our outputs will be better because we will mitigate the "garbage in, garbage out" problem. So, let's focus our efforts on using AI to enhance our inputs. The 80/20 principle we are going to apply here is to add academic papers and research to our input.

Now, you might think, "Okay, first of all, this is boring. Second of all, this is practical work, right? This isn't some sort of academic research that we're conducting. We need practical things to happen." But think about it.

Do you read Harvard Business Review, McKinsey Quarterly, or the MIT Sloan Review? All of these publications aim to take boring, hard, complex academic papers, make them relevant to working professionals, and deliver the content in a way that allows us to extract actionable insights. We have to realize that we don’t want to be mere readers of HBR, right? We read so slowly, but if you feed the entire archive to ChatGPT, it will immediately know so much more than you. Right?

And we don't even want to be the writer for Harvard Business Review because people spend their entire lives writing a few insights. You know, you can just leverage those insights. If you don't want to write, why spend that time researching? Instead, what you want to be is Harvard Business Review itself for your team and your organization. This means going through trusted resources that are complex and nuanced, translating that into the context you’re in, and creating frameworks that turn theoretical knowledge and insights into something actionable that you can share and deliver in a way that busy people can understand. You need to move up the value chain and examine the 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 do that too. Let's take an example.

Let me show you the three tools that I use to ensure I can be an HR Business Partner (HRBP) for my team and my organization. Alright, let's use an example. One of the biggest challenges managers face is the problem of a team underperforming. They are bright people, but it seems like they don’t want to do anything.

They're always complaining about stuff. How do I improve team performance? Okay, so 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 also go to Google.

It might give you some average inputs. Let's give it a try. So, you know how to improve team performance at work. Common question, right? Oh gosh, it tells you to focus on goals and open communication.

I mean, these are just such empty, broad recommendations that are not going to lead to great output. So what I like to do instead is to use the first tool of today, Elicit. This tool helps you find academic papers on the questions that you ask. So let's just 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. You receive a summary of the top four papers, which is free. Then, you get a list of relevant papers. What I like to do is ensure that what I'm reading is well-cited, peer-reviewed, and not just some random article.

I like to sort by the most cited. I take a look, for example, at team design features and team performance; there seems to be a relationship. Let's take a look at this one. All right. What you can do is then click into Semantic Scholar, which will usually lead you to the actual paper itself.

And 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 this. But no problem. We'll just copy and paste the paper's name into Google and see if we can find a PDF. Okay, here we go. Here's the PDF published by Sage. Let's just double-check.

That seems to be the right one. Okay, so we'll download this. Then we'll use the second tool of today, which is Notebook LM by Google. This tool, unlike ChatGPT and Claude, does not use AI on the internet's knowledge; instead, it relies solely on the sources that you upload.

So you have a lesser likelihood of hallucination. Sign up with your Google account, and you'll see something like this. You want to create a notebook, which essentially houses your sources. Since we downloaded a PDF, we are going to upload it here. Now, we haven't even read the paper yet.

We don't really know what's in there, so what I like to do is to use it as a table of contents first. This is automatically generated by the notebook. Alright, so now we have a table of contents.

Okay. Groups are important. There's classification theory. It looks like we need to consider characteristics such as team size, team-level task design, and leadership.

Interesting. Now I don't want to read this yet. It's a lot of text and really poorly formatted, but you know, Notebook is still pretty much experimental, so we will give them some time to figure this out. But now 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 to give me the basic gist. So here we go. How do team design features impact overall team performance? Let's see what it says. All right, it looks like there are about three components, and we saw that in the table of contents as well, right?

So, there's team composition, task design, and then there's leadership. I love this about Notebook LM: you can click into it to find the actual place in the source where it discusses these ideas. Even though this giant paper is probably tens of thousands of words, I can figure out which parts are most relevant to me and actually find the citations so I can better understand what they are saying. Let's say I find that I can't change the team composition—I'm not going to fire anyone or hire anyone new—but I want to change task design and leadership. I can focus on these cited areas first. Okay, so that's the first cool part about the Elicit to Notebook LM workflow, but this is only one paper, right? What I would do again is go back to Elicit and see if there are any other papers that are interesting. Well, leaders can learn about teamwork and developing high-performance teams.

That seems interesting. Let's take a look and see if we can find the paper. Oh, here is the PDF. Fantastic! So now we don't have to... oh no.

Okay, bad request. That's all right; it happens. Let's go back and see if you have a PDF. Get 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, I mean, surely your work will help you subsidize getting this paper. But even if they don't, let's see if we can find it somewhere else.

Let's see if we can find it somewhere else. Let's see if this is the right one. Fantastic! All right, here's the second PDF we got. We will head back to Notebook LM, and this time we're going to press the plus sign and again upload it here.

Once that's done processing, you'll see down here that you can now chat with both sources. But before we do that, I actually want to uncheck the previous one because I just want to understand what this new paper is about. So let me create a table of contents for this one source, and I will 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: leadership, capable and committed members, norms, a team structure that has a clear mission, responsibility, goals, and a framework to achieve results. See, this is already so much more clear and specific compared to the Google search results, which have become diluted over time. It often starts with one person at the source, but eventually, secondary sources dilute the message. Okay, a way to manage team processes— and that's the fifth one.

Okay, great. Now let's select both and ask, "What are the key character components?" Let's say, "What are the key changes a leader can make in a team to improve performance?" Now it will look through both sources and provide us with an answer. All right, so let's focus on member selection.

Let's say that's not relevant to us. Promote task meaningfulness. This was mentioned in the first source. Great. Impact our team with autonomy.

I don't know if we've seen this. Okay, so this was in the first document. Great! It provides us with the exact things that we should be doing.

Cultivate intra-team coordination. All right. Provide effective leadership, which means guiding, motivating, and supporting the team. Let's see the details. Now, let's say that out of all of this, I think for my team, the thing that makes the most sense is to promote task meaningfulness.

I want to empower the team, facilitate team development, and, of course, provide effective leadership. So, please give me a detailed plan with three changes that we can implement to improve team performance. Now, we have a high-level plan of how to proceed, but it still feels somewhat vague. Even though we know the input is good and we are focusing on what academics have discovered, we need to move on to the next part. I will copy this and go into Claude, asking for help to create a detailed plan for improving team performance using the information from the notebook that LM has shared with me. In just two seconds, we will have a 12-month, three-phase, four-module plan to improve team performance. Just think about it—if you were in HR or a team lead and shared this plan with management, they would be blown away! However, that's not all we need.

Let's say that all of this still sounds like a lot of complications. So, let's ask: what are the 80-20 changes I can make to my team to drive performance that depends on leaders, rather than team members, to implement change? I would like to refer to two studies, a meta-analytic review by Greg A.L. and Warwick. Now Claude has a better idea of where I'm getting these insights from.

And so, based on details that Ian knows about these two sources, it gave me the core leadership action plan. Okay, so task design optimization seems to be one of them, authority distribution is the second highest impact, and contextual support is also included. So now, from 12 months, we're down to three months, and it gives us some expected results. Whether that's true or not, we will have to wait and see.

But it helps us list out things like key success factors and measurement frameworks. So, 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 goals and clarity levels. We can also use interdependence levels.

We can see where they intersect, and we can even show the team, "Hey, this is what I'm thinking about for changing how we approach tasks. What do you think?" We can get feedback on it. Right 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 an improvement compared to previous GPTs. Let's say we are working with teams that are underperforming and unmotivated. The plan it gives me is much more vague, right? As I communicate and give feedback, I already understand this: having clear goals is essential.

I know this. I know all of this, but it doesn't help me. Help me create a framework to set goals, let's say. And yes, SMART is good. 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 think critically and be more creative, associative, and generalist. I believe we 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!