The 4 AI Team Members Execs Should Hire Right Now

Today on the AI Daily Brief, 4 AI employees that you should set up right now. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. Alright friends! Happy Memorial Day for US listeners, and happy operator's bonus episode for everyone else! I am once again joined on this episode by Nufar Gaspar. You've probably seen her on the show before. She's back currently with a new AI executive catch-up program. That week is a 4-week sprint to get up to speed and to get current with AI. And today she's walking through some specific recommendations about where she thinks leaders specifically should start. All right, Nufar, welcome back to another AI operators bonus episode of the AI Daily Brief. How you doing?

I'm good. How are you?

Good. You know, I think we don't set these up as meant to be in some sort of sequence or continuous with one another, but they do end up having the effect of, of sort of mirroring the pattern that we're going through as we learn more about what the market needs relative to education and upskilling. This episode specifically, I think, comes out of both an experience that, you know, you and I have had both separately and together, an observation, as well as sort of some things that have fallen from it. You know, we've been running these agent operating system builder programs called Enterprise Claw, moving maybe to a name Agent Boss. And as everyone across the enterprise has started to recognize the need to move into this agentic way of working, getting people who are ready for it into agentic system building, there are a lot of people who are just racing to catch up and are just a few steps behind that, which led to the creation of the Executive Catch-Up Program that we're now offering. And also just this framework that you're gonna share with us today. So I'd love to hear a little bit more about the context for this and your experience that led you to it. And then I'm gonna turn it over and and let you teach us some interesting ways of looking at the world.

Right. So yeah, I fully agree that there is a gap in the market. Like the most frontier firms and individuals, they talk about agentic fleet and agent orchestration and are worried about token maxing and many different things that get some other people to feel left behind even more so than before. We often call that the capability overhang. And I believe that there are so many things that every executive needs to do, even if they are quite advanced, but let alone if they've been kind of feeling that they need more time to catch up. And today I'll try to frame it in a way that is as tool agnostic as possible and just give executives the core decisions and things that they need to make in order to close some of the gap and be ready for orchestration of agents and so on.

Awesome. All right, well, I can't wait to see what you got for us.

Sure. Good. All right. Today, what I was thinking is to kind of share with you what I'm learning based on training executives across like 30 different countries, and I see very frequently 3 patterns that come up again and again, and I think that most leaders are in at least one of them. So one, I'm calling it the podcast CTO, someone that is deeply informed, that knows every release, every benchmark, but in like many cases, they haven't built a system for their own work at all or not yet. The second one, I'm calling them the weekend tinkerer. They are building stuff, they are impressive for their own private time and their own thing that they're building, but they don't really know yet how to bring it to the day-to-day because of various considerations. And then we have the manifesto writer. Those are the one that has the clear vision, that already funded the transformation committee, but personally they have not yet feel that AI can do the work at their level. And while all three of them are making progress, all three are leaving enormous value on the table unless they are using the tools to the full advantage. Because I think the tools have crossed a threshold in the last few months that makes partial engagement kind of a real missed opportunity. And to be even more bold than that, I think that the leader's quality of AI usage is the single biggest predictor of how well their teams adopt AI. And when you see that the CEO is the best user, the organization looks like the most forward AI company out there that I get to see. And when leaders are kind of the ones that hand off the decisions and just talk the talk, they go wrong in both directions. They are underestimating what the technology can do sometimes, and sometimes they go to the flip side. They set unrealistic expectations for the team. And I think that your AI usage, while having its own flavor, it's dealing with a different usage than your ICs or your employees. Because for the most part, leaders have a very high judgment, decisions, very complex stakeholder dynamics, and undocumented context that lives in their heads and not necessarily in any document. They also deal in many cases with communications at scale, and thereby a lot of the more generic productivity tips, they just don't serve executives as well as they serve the individuals. And thereby I want you to have a deliberate system built for how executives, and specifically you, actually work. So in today's episode, what I want to I'll do is to start with what I believe are kind of non-negotiables for any executives who use AI nowadays. Those would be the operating principles that apply no matter what tool you use or what you are building. And then I'll walk you through the 4 AIs, I'm calling them digital employees or AI team members that you should hire, namely the AI data analyst, the strategic thought partner, the communication expert, and the operational powerhouse. I'm gonna be as always very concrete in the methodology. I'll try to give you as many tips on screen and all the thinking behind each every one of these employees and why I'm saying to use them or prompt them the way I do. So one thing before we dive in, everything that I'm about to cover works on any AI tool, whether it's like a project in ChatGPT, Gemini, the Cloud Chat, Copilot, or similar, or if you're already using the modern agentic tools like Cowork Cloud, Codex, and others, they are all applicable there. The nuances is in the technique, not in the subject matter expertise, which will be where we focus today. So all will give you results as long as you work according to some of these principles. In general, I say don't sweat the tool choice. Pick one and stick with it at least for a while. Sweat what you're building and how you're building it. And I want to go over these 5 operating principles that apply across everything that we will do later on. And these are non-negotiables from my perspective. They separate executives who get mediocre output from those who get exceptional results. So first thing that I wanted to say is it doesn't matter whether you use the full voice mode or dictation tools like Whisper Flow, and I know that NLW often mentioned that, but typing filters your thinking, and when you speak, it lets your spiral, your intuitive thinking, those come through. And with the newest tools and models, you can think about the smartest ones, whether they are Opus models or the latest GPTs or literally all of the latest and greatest thinking models, they handle the unstructured and nonlinear input extremely well. And as such, your messy thinking becomes the most valuable input because AI needs everything that resides within your head, and speaking to AI lets you share as much as possible of that with the tool. The second thing is I wanted you to do a lot of brain dumping, and the reason why I'm saying that is that executives carry enormous amounts of undocumented context. A lot of the relationship dynamics, meeting undercurrents that only they could gauge, half-formed intuitions, things that were said before recording started, Maybe it was a look on somebody's face during the meeting. AI gives you mediocre answers unless it has this context. So I want you to capture this context habitually. Voice notes, meeting reflections, quick brain dumps. Even if you don't know what it is for yet, just start capturing that. And I know that some people really like to use tools like Obsidian for it, but no matter how you do that, just make sure that you form the habit and reduce the friction of capturing thoughts such that you will regularly do that. Next, I want you to have AI interview for you because the more senior you are, the more unknown unknowns you have as part of your interactions. And before any complex task, whether it's research or decision or communication, let AI grill you first. And Codex even created a command for that. What are the assumptions that you are making? What haven't I considered? What should I provide in forms of a context? This will often surface a lot of the blind spots before you even get the results that you're after. As noted, some tools do that already structurally. In other cases, you can just ask AI, interview me until you have everything that covers all these blind spots. The next thing that I wanted to say is the more critical the task, the more you need to separate the planning from the execution. So I don't want you to jump straight into the output when there is some thinking to be done. Have a conversation with AI to help you plan the approach. What information do you need? What is the order that you need to execute? What does success look what does it look like? Only after you have a refined plan, then you execute, often even in a separate conversation or a separate tool. And even better, you can always combine that with the AI interviewing you as part of the process. And the last non-negotiable that I wanted to mention is that I want you to be very intentional about your intervention point because the more judgment-heavy the work, the less likely you should probably fully offload it to AI. Because you are no longer doing everything manually, I want you to Hone the skill of knowing where in the workflow your judgment adds the most value in designing the systems such that AI handles everything else and you step in only at those strategic moments, typically for sure at the beginning and a few intervention points later on. And critically, I wanted to make sure that as a leader, you make sure that primers or initial thoughts are always captured because leaders almost never come to a task with a blank slate. For sure, you have your existing assumptions, decisions, premises, maybe half-formed opinions. Offload everything that you come to the table with, the beginning, even if it's messy and unstructured. This is your advantage because this is where all of your experience is baked, and this is where you being very opinionated will really help steer the AI towards the output that actually reflects your thinking rather than getting generic results that you will reject annoyingly anyway. All right, so with these non-negotiables in mind, let's start talking about which AI digital employees I want you to hire. And it doesn't matter how good your human teams is. Odds are you have a very good team. It might even be an exceptional one, but there are some things that you've always wanted that no human team can realistically deliver. It can be like a strategist that is available at 11 PM when you are processing a board conversation or a researcher who can run 6 parallel analysis before your morning coffee. Or maybe you have in mind like a writer that internalized your 10 years of public speaking and writing, is able to truly express your own voice. Or maybe you need an operator who monitors everything in your organization without getting fatigued or annoyed by your constant request to get more and more and more information. And if you have this well-structured digital workforce, this doesn't replace your human workforce, but it gives you capabilities that you never had basically the bandwidth or the headcount for. And with the 5 principles that we mentioned before, the output won't be generic. It will be very much like yours. So I wanna go over these 4 digital team members that every leader should hire. And for each one, I want to give you my top tips on how to get the best possible output, the things that separate the mediocre results from genuinely useful ones. So let's start with the first one, and that will be hiring a research analyst, the one that you always wanted but never had time to brief properly. It's now available on demand. And what I think is that most people type a question and accept whatever comes back, and that's using AI like a search engine. And as an executive, you always research with a decision in mind, or for the most part, and you come in with a lot of existing assumptions and knowledge and thoughts. So I want you to state that up front and then instruct the AI on how to do the research. You want to constrain many things, namely maybe the time horizon. Maybe you want to define which sources matter in what order, what priority. Maybe you want to tell it what to exclude. For example, never use marketing materials as a primary source. Maybe you want to specify what constitute as reliable data in your domain, and you know it probably much better than any AI tool. And this applies equally to internal data, maybe your P&L or your team metrics or your customer data. Same principles of being opinionated about the scope and the output. You are briefing the analyst. You're not asking a Google question, and that's the state of mind that you need to be in. And the more opinionated you are about how the research should be conducted, the better the results. So here are two kind of expert craft-specific techniques from working with hundreds of executives that really get you from good to being exceptional in being able to get results with AI research. So the first craft that I want to mention is wisdom of the crowd. I don't want you to ask one AI tool and then trust the results. What I like to do is to send the exact same research across multiple models, sometimes even the same model or the same tool in several different sessions. Then after you had multiple threads researching the same question with the various tools or the same tool in various instances, you aggregate and you aggregate on where they agree and you investigate where they diverge. Then what I would strongly recommend that you do is that you use a separate model or a separate thread to fact-check the aggregated result because AI is much better at verifying than at generating correct research from scratch. So this is a weird phenomenon, but it works. Here as you aggregate the results. And if there is a consensus across tools, it's likely factual. From my experience, if you see 100% consensus, that's probably a true fact. And if only one tool reports something, that's something that requires going deeper and researching again. So that's also a cheat sheet on how not to have to validate each and every data point from AI by actually expanding the scope of the research and then narrow it down in an intelligent way. The second thing that works very well is that before you act on any research or analysis output, whether it's external market research or internal data analysis, run it through these 3 questions. The first one is grounded in real sources or is AI pattern matching? The second one will be what's missing that I didn't think to ask? And the last one, and that's the most important one, are you feeling comfortable putting your name to it if someone was asking for that? And if you, even without good justification, intuitively don't feel comfortable to put your name to this, this means that you need to do some more work. It takes about 30 seconds and it catches the majority of the AI research failures before they become bad decisions or slop that you share with others. And if you really want to take it even further, like a pro tip will be that the output shouldn't default to a wall of text or a boring bar chart. You need to think with the modern tools, what's the easiest way for me or the consumers of this research to actually interact with the data, whether it's an interactive dashboard, an infographic, a reactive page that you can filter. Maybe it's an audio summary that you listen on your commute. You can build all of this with the same tools. You just have to ask and be as creative as possible. So that's your research analyst that you should hire. The next employee that I want you to hire is the advisor that you couldn't hire at any price, the strategic one that is available 24/7 with no ego and infinite patience. If you already have one, enjoy your human one, but for the most part, we all need the strategic advisor in the form of AI. And another point is that the more senior you are, it's probably very lonely to be there up up there because decision-making that you are making for the most part is with yourself. And even with the best mentors and a very strong leadership team, it's never enough for the sheer volume of decisions and often also the sensitivity of the decisions that you need to make. So, things that you cannot run by the board or that you want to get ready for your board or things that are too early to bring to the team, use AI, use it well, and then the AI becomes the sounding board that's always available for you. And it's the one that gets you out of the biases that you often have, that can also challenge you out of your comfort zone, and that can take you out of the isolation that sometimes comes with being a senior leader. But here's what makes it actually work, and for the most part, it's one word that these listeners are probably very, very aware of regardless, and that is context. The more AI knows about you, whether it's your role, your company, your ecosystem, your competitive stance, your recent priorities, history of decisions, and what made the decision that you're already had to work well or fail, the better it advises. So without that context, you get generic strategy advice, and with it, you get something that feels like an advisor who've been working for you for years. I strongly recommend that you build a personal context system. You can refer back to NLW, great episode on building a personal context system. It was a while back, and it feeds into every strategic conversation, and that's the foundation. Everything else is built upon. So how do you do it well? One way that works very well for strategic discussions is to hire not just one strategic advisor, but a board of advisors. You can think about all the mentors that you ever had or wished you had. You should build them as personas. You can give them a name. You can even select to have as a mentor someone who is very famous and that you really like their thought leadership, You can give them an archetype or decision-making style. You want to have them debate a decision between themselves before they present it to you. Then you will get a very diverse perspective from all of them, not just one AI voice that is pretending to see multiple angles. And what you definitely need to do here is to calibrate the pushback. Don't just create a board of advisors that is a devil's advocate for Spot, but rather create one that not a sycophantic. So I want you to calibrate the pushback, not a devil's advocate for the spot and not a sycophantic agreement. I want you to instruct the board of advisors to challenge and then converge. And the goal is overall a better decision, not an endless debate or one that will just take all the wind out of your idea. Second thing that really moves the needle here is to be aware of your decision style because every leader decides differently. Some leaders need 10 enumerated options in order to choose from. Others want just one bottom line after a long debate, and some want AI to push very hard back on them and then give them the space to sit with it. So you know yourself and you know how best you make decisions. You make sure that you instruct the AI to match your decision-making style. That goes a long way to feel like it's an employee working for you rather than you working for it. And then before any major decision, ask AI to also surface the biases. What biases might I as a human have here? What biases might might you, as the AI, have here? And what am I not seeing because of my position or my experience? And because leaders are in many cases surrounded by agreement and AI should not be another yes voice in your vicinity, but rather it also needs to kind of challenge you only when challenge is required. I want you to calibrate it to be the advisor who makes you better, not the one that exhausts you. And if you want a pro tip here, after any decision, run scenario simulation. So something like, given this decision, what happens if the market shifts to X? And if a competitor does Y, what happens? And if the team pushes back on Z? So try to make sure that you stress test according multiple scenarios, and then you can stand behind the decision under multiple futures and not just the one that you're hoping that will happen. Okay, let's talk about the next employee. The next one that I want you to hire is the communication expert. The one that writes in your voice for your audience and not kind of a generic executive prose that sounds smart but is very annoying. And I think the communication is the area with the widest gap between basic and advanced AI usage because it's trivially easy nowadays to generate text that sounds like everyone and basically no one. And generic, pleasant, forgettable AI is very, very common. But if you want AI to actually sound like you and to differentiate between audiences and platforms and goals, this is a real skill that you need to adopt. And it's also one of the first areas where people around you can tell whether you are using AI lazily, because most people can now identify AI-created content quite easily versus people who are using AI well or that are even writing for themselves. So the distance between I can tell this is AI and this sounds exactly like her is entirely about how you steer it. And here, what I want you to do with this digital communication expert are to, first of all, do some style profiling. And here I want to offer two approaches that work together. First, I want you to collect all your best writing across document types. Those can be board updates, team emails, LinkedIn strategies, and so on. Give them to the AI as examples and then ask the AI to analyze your writing style because AI is very good at naming the patterns that you cannot articulate. You just don't have the language yourself. It will talk about your rhythm, your sentence structure, your rhetorical preferences. Most of us are not linguists. And thereby cannot call it out. And if you are in a position where you want to even improve your writing style, combine the examples that you created with the AI-narrated analysis to form a guide that actually captures you, but also supplement that with additional writing of people that you truly appreciate how they write, or things that you cannot exactly express why you like it, but they resonate very well with you. It's not copying, it's just augmenting your style with some of the aspirations of perhaps how you will want to sound. So that will also help you to evolve your voice beyond just getting AI to sound exactly like you if you think that there is room for improvement. If no, just stay with your own profiling. And then another thing that will truly unlock a lot of the goodness of AI as a writing guide or writing companion is to create detailed personas of your actual readers, what they care about, what drives them to action, what they're skeptical of, and have these personas a little bit similar to how I was talking about board of advisors. Have those personas review the relevant draft. So they should answer the question of, is the message clear? Would I take action? What's missing? And what would make me stop reading? If they answer all of these questions in a way that resonates with your message, by all means do that. Otherwise, there is a lot of room for improvement. And unlike human reviewers who get fatigue, AI gives you unlimited iterations with pointed feedback from the people who actually matter or the persona of the people who actually matter. Those are your readers. So definitely take advantage of that. And the last pro tip here will be whenever you want to give an AI feedback, don't just give it a generic feedback like, I don't like this. I want you to score on very clear dimensions. You decide what are the dimensions, but as an example, maybe you give it a 9 out of 10 for clarity, 5 out of 10 for wittiness, conciseness gets a 7 out of 10, and so on. Because AI tools are very goal-driven, they are much better performing when they have clarity of how far they are from where they want to be. And also when you give it feedback that is more qualitative, make it very, very concrete. Not just, "This is overall bad," but say, "What exactly don't you like? Is it the sentence structure? Is it the phrasing? Is it the ideas?" and so on. That way you don't have to iterate endless times only to get frustrated and ending up writing it yourself, but you will actually get to steer the tools in the right direction. Direction. And then we have the very last employee that you need to hire, and this one will be like the operational powerhouse that will help you in the day-to-day stuff, and that will give you the operational visibility and the support that you always wanted but never had typically the headcount or bandwidth to build. And I think that with the prevalence and the accessibility of the connectors within the modern AI tools, it's now much, much easier and more reliable to gather and synthesize across multiple systems in which your data likely resides in your organization, no matter how messy it is. And for many leaders, the operational support that they rely on, whether it's meeting prep, status report, briefing documents, this one is the work that their teams, they do probably diligently, but often totally resent. And also, this is probably one of the first things that your employees will likely automate with AI for themselves. So leaders should probably do the same. Of course, you can rely on your team, but I would advise that you go even further and do it yourself because you You know best or better than everybody else what are your needs in terms of the specific report and the specific information that you're looking to get much better than everybody else who needs to brief you. So one thing that you can do here to really nail this is to not just automate what you already do, but rather think about any visibility and any operations that you always wanted, but never had the bandwidth for. So maybe you always dreamt to have a daily overview across all the departments in your organization without burdening your reports. Maybe you wanted to get like a deep P&L analysis every morning, or maybe you just should or want, have always wanted to have a stakeholder relationship tracker that reminds you of the specific context before every meeting or status synthesis across 10 channels you can physically monitor. And AI makes something like that is previously infeasible operations to a very, very feasible thing. And then the question is not what can I automate? The question is What would I build if I had an unlimited amount of headcount in my company? And those, I think, should be the first operational dashboards and ideas that you should build. And another thing that makes the operational powerhouse digital employee or AI that helps you become even better is to make sure that it's your own and not generic, because a generic meeting preparation like summarize last transcript and emails is for the most part for leaders not good enough. Enough because your own meeting prep has your own specifics. Maybe you wanted to capture what was the undercurrent feel like in the previous meeting. Maybe you have other sources that are very relevant for this specific audience and so on. This is something that you should definitely personalize. And as mentioned at the beginning, the more you document the intangibles as part of the overall usage of AI, the more your, let's call it personal CRM, aka your traction of what happened with this person previous time, the more context you have about the stakeholders that might not be in any official CRM or other systems, the better you will be ready for meetings and you can think about the analogy and other things that you need to have prepared on a day-to-day basis. If you want a pro tip, whenever you're contemplating building a dashboard or a meeting prep automation or a morning brief or anything else that you have in mind, to automate as part of your day-to-day activities, I want you to never automate before you've tested it manually and repeatedly. So maybe you created like a, a morning brief. I want you to run the morning brief on every morning for maybe a week or two weeks before you, and you can do it manually before you commit to having it run automatically, because only after a week or two of seeing the data and seeing how you consume it, you'll be able to say, yes, that's a good use of of this automation or something that I can improve and then commit to that. And that, by the way, is applicable to anything else that you, the rest of the use cases, whether it's research or strategic thinking or anything that you're contemplating automating, it's always better to first test it for a while, sometimes in a stealth mode, meaning in a way that does not truly impact any systems or decisions and only then to automate. So kind of to summarize, it doesn't matter where you started today, maybe build a lot of the systems already, maybe you're quite a beginner, but if you build even one of these systems with the principles that we covered, I believe that you will start closing the gap between knowing about AI or using it in a certain way to actually having it work fully for you and feeling like a digital employee that you hire that drives value in the way that moves the needle in your context. You then become a much better AI user and then obviously you get better results, hopefully a lot of ROI, but also in the process you become a better leader of adoption because you'll make much more informed and inspired decisions about where and how your organization uses AI. And that's why I want you to roll up your sleeves and not rely on others to do everything for you. And I think that the next steps, once you have these 4 digital team members that are working well for you or your version of the digital workforce that you should have, the natural one is to start building a chief of staff that can orchestrate across all of them, not just like a task-by-task assistant, which is where we focused a lot of the energy today, but one that has a cross-view of your decisions and communication and priorities and so on. That's gonna be like a capstone if you are more of a beginner, but you should earn it first by getting a mileage with each individual employee that we just discussed. I want you to spend your time building these digital team members based on the principles, not on the technicalities of one tool or the other, even though the technicalities of the tools are very tempting and a lot of the training out there are focusing a lot on the tools. I actually want you to focus more on the results and your personality and everything that we discussed because I think that the methodology is a much better constant than a lot of the specific features that are available today or tomorrow. Tomorrow. So, kind of what to do next. Pick at least one of these four team members, implement some of what you heard, even if you already have processes in place, then refine them with these principles and with actual usage. Have the time to do this work. You cannot just do AI with 15 minutes a week. You need more time than that. If you want to do that yourself, by all means do that. If you want to do that in a guided way and along other leaders that are undertaking the same accelerated learning, I'll be very happy to have you with us in the Executive Ketchup. But it doesn't matter what you choose, just kind of stop learning to swim from the shore or just splashing around. I want you to get in the water with intention and start swimming because that's the only way to not only learn but also be a much more informed change leader. That's it.

All right, great. I love this. I think one of the things that's interesting to me is that these sort of interaction patterns, although we're not yet talking about full-fledged agents in agentic systems really will carry people from a simpler use to a more advanced use over time, right? That these, this is sort of training a type of muscle when it comes to thinking about how you interact with AI that I think is really important and really valuable. So I'm excited to see if people take you up on this and where they dive in.

Same here.