How to Build an AI-Ready Culture: A Practical Guide

Welcome to a special operators cut bonus edition of the AI Daily Brief. I recently put out an episode about all of the things that we've learned across thousands of interviews with executives as part of the super intelligent agent readiness and opportunity mapping assessments. Now for background, the way that these surveys work is that we deploy voice agents to interview a very wide cross-section of people, both leadership and those on the front lines, in order to understand how work works today. We then process that through a proprietary LLM-based process that has access to a bunch of custom data sets that we've built in order to provide recommendations around where AI and agents could help create new opportunities and solve problems, as well as providing advice around things like change management initiatives that can help their organization get more ready. For some time, we've been thinking about maybe doing a more educational type series that gets into some of the lessons that we've learned and that could be maybe a little bit more practical for people who are trying to apply this stuff inside your own organizations. And the response to the recent episode about those surveys was so good that it seemed like the time was now. So I'm delighted for this three-part series, which will be coming out over the subsequent Saturdays as bonus episodes to have back to the show New Far Gaspar. New Far is our head of research, as well as a former AI leader at Intel and an enterprise AI consultant who helps companies with all of these types of different issues, drawing on both an incredible wellspring of existing experience and a constantly updated new set of experiences as she helps lead research and put these things into practice. On our first episode, we get into the challenges of culture so without any further ado, let's dive in. All right, New Far, welcome back to the AI Daily Brief. How you doing? - I'm good, how are you? - Good, so we're doing something that I've been, we've been talking about for a while. I think that there's this really interesting and frankly quite large space between information, news, education, podcast style content on the one hand and full-on upskilling coursework on the other hand. I think that space in between has a ton to explore, right? And since so many people are already consuming things like this show and mediums like this, I wanna start experimenting with ways that we can use this podcast feed to nudge towards that sort of more educational, informative type of stuff without going full kind of online course. And so this series is a little bit of that. It's also rooted as you're gonna tell us in some of the other work that we do together. So maybe just kick us off, let's dive in. You're gonna be basically in the sort of instructor seat today, we'll go through this and then we'll have a conversation on the other side of it. - All right, so I'm happy to be your instructor for this next few minutes. And let me share with you a few thoughts on agent readiness, specifically on the culture side. You don't have to trust whatever I have to say. You have to trust kind of the thousands and thousands of people that we interview as part of the super intelligent agent readiness. And the insights are not kind of theoretical. They are sourced from all of these transcripts and conversations and of course also from my experience working with many, many different companies of all shapes and sizes. So that's the source of everything that you will be seeing. And the truth is unfiltered, honest and quite valuable insights that they give us because we send an agent to interview them. The employees of the companies, they are typically the best source of truth whenever we're trying to understand what's happening in the company. And I have to say that whenever I look at a transcript of multiple interviews, I feel like after a while I'm getting a very clear sense of what the company is and isn't. And it's something about being interviewed by an agent that gets people to be really open and honest. So a few representative quotes coming from the CEO, honestly kind of saying that it took him a full year to get everyone up to speed with his view on AI, whether it's employees talking about how they will eventually just shadow AI if the company doesn't let them do that, as well as others. For example, DCIO admits that their employees are too biased towards building and that truly slows them down. So we have so many of these nuggets in these interviews and we kind of bundle them all together to give you a more prescriptive overview of what it takes to be agent ready. For the past year, as we mentioned, we've talked to so many companies that we now believe that we have a clear prescription. And there is one very clear view when you look at all the dimensions by which we measure agent readiness, which is culture and data, technology and use cases. The one reality that is I think the most prominent is that no one is agent ready, 0% of fully agent ready according to the way we measure. And in my opinion, and so does the data show that the biggest reason why it's not the technology and it's not the models they will and continuously get better. The tools will also improve. The single biggest reason is the culture. And that's often a blind spot for leaders, not that they're not investing, but they're not investing enough. And we believe that getting agent ready requires a fundamental shift in how the organization communicates and collaborates and creates value. And it's a culture problem first and a technology problem second and I'm a strong believer in that. So to offer you a means to get it right, you don't need to kind of hope for the best. You have to have a framework. And I wanted to create something fun for us and make it a deliberate change. So that's the framework that I wanna offer you today. It starts with communication, human oversight, attitude, network governance and enablement. And I wanna break them down for you and bring an actionable set of insights. So let's go. So it all starts with the C, the clear communication in the absence of clarity from leadership. Employees typically feel devoid with fear and anxiety about their jobs, whether agents are gonna replace them in a minute. And over this last year, there were several very memorable, at least in my opinion, memos coming from the CEOs that shared publicly. So one example was the Shopify CEO. He was very explicit. He was saying to his employees, AI is no longer an option and he expects employees to be proficient. And that was a very clear message about upskilling and that AI is now a mandate in the company. Another CEO that hopped on this AI memo trend was the Duolingo CEO. He went even further stating publicly that AI will replace contractors but not intend to replace the full-time employees. And these are just like examples, but what I believe is that each company need to create and maintain their own kind of AI manifesto. And it doesn't have to be long and tedious process. It can even be one pager, a very clear communication to the employees that will go a very, very long way and reduce a lot of the confusion and the problems that come from lack of communication. And of course, this has to be done at a company level, but also managers at all levels. I believe that they need to create their own version with of course the new answers that are more appropriate for their teams. So to make it easy for you, regardless of your stack like placing the hierarchy, I created a template of how to set your own AI manifesto. And this is the main thing that I believe needs to be included in any AI communication to the employees. So it starts with what you believe, then you should communicate what you expect them to do or how to behave, what you allow to do, what is permissive, address the elephant in the room. If you don't, they will always fill up the gaps, talk about your intentions with regards to jobs. It's better to be in my opinion candid than disregard and define what they should do. So that's the communication piece. Next is the human oversight. And I want to be direct here. We're not at a point where agents can or should replace humans altogether. They can replace tasks, but you know, not entirely. And this has a few clear implications. The main one is that first, you want to have a very clear playbook based on your company regulatory state. So probably companies that are from a higher regulatory point of view, they have to be a much more strict versus companies that perhaps are from industries where they can take bolder risks. And here you have to define where agents can aim for full autonomy and what are the guardrails, where versus places where you're not aiming there and the guardrails need to be accordingly. The other thing that you need to define is let's say that we already get some, something good out of our agents. If employees get their time freed up, you need to clearly define what you expect them to do with it. Stating that it's not a matter of just replacing them, but rather you want to perhaps grow the business, look for other places where they didn't have enough bandwidth that creates both business value and reduces the fear and avoids employees sabotaging the agent behavior because they're afraid that any time freed up will be just a case for letting them go. And lastly, you have to have very clear goals to measure how the agent will be tracked versus the goals or how will you measure success? And that aligns everyone on the value and the why behind the work and creates typically much better results. When it comes to attitude, we're talking about the attitude of your employees and your managers. And when we look at the audit data of super intelligent, often the attitude is a very good predictor of how ready the company will be regardless of all the other data. So what I want to encourage you here is to be very proactive, to manage the duality of employee sentiment and channeling these enthusiasm that they typically have to eliminate the grant work while still addressing biases and profound fears about the job security or being quite set in the old ways. And often the attitude that comes from the interviews are very surprising. For example, one interesting observation is that often the most talented engineers are the worst adopters of new AI tools. They have a deep seated pride in their own system and they often have not invented here bias and that causes them to view any external tool with a lot of skepticism. And it's not a sign of incompetence, it's just a byproduct of their expertise. And to be even more blunt, this is one of my favorite quotes coming from one of the interviews. This is a strong proud internal engineer basically telling their managers that they will not allow for vendor tools to come in and they will fight it till they die, basically. And when you hear that, you know that that's not a technology problem. That's a culture problem and you have to address that. And you know, top down mandate and great attitudes, they are very, very important, but they are not enough. And beyond just getting the employees on the right place and asking them to do everything and so on, you have to do something that is way more democratized. And this is where the end comes in. This is the network of champions and builders. And you know, a company-wide email from the CIO that has probably a half-life of five minutes give or take. But if my peer shows me an amazing tool that they use day in and day out that gets them to do the job 10x faster, that's gonna last forever 'cause I'm gonna do that. The grass-root adoption on the other hand is also very good, but it's not enough. So I want something in between. I want you to formalize the grass-root and how to be intentional about deciding who gets to help their peers. And I believe that there are two highly effective ways to do that. The first, I want you to nominate and train internal AI champions. These are people that are carefully selected, trained, and then continuously groomed, such that they can be the team's AI advocates. And the role will be to identify use cases and build local AI capabilities in some cases, push for a better usage and help their peers. And that's something that is very near and dear to me 'cause I'm working with many companies specifically on that, so I can talk about it very long for a very long time. But one example that I wanted to say is of a company where I think they trained almost 20% of their employees to become these AI champions. They gave them three full days, of course, a very extensive course to help them understand what it takes to get AI from idea to production. And they've seen an amazing uplift in usage and many other KPIs. So that was a proven success for the champions. And the other thing that I want you to do is not to settle just for champions 'cause that's great. But in many cases, I want you to allocate or hire dedicated AI builders or builder, depending on the company size. And it can be individuals or teams, but they need to have primary role to build AI capabilities. They are professionals, they are people that know how to build complex stuff, and they will work on things that others can. And finally, to gain the most value out of these two populations, you should establish like an internal network that allows you for continued learning and sharing between those champions and these builders often together. And that's the sustainable part of nominating and hiring people that will build for you. All right. Even though we're talking about a lot of distributed and democratized activity, we still have to have a lot of guardrails and a lot of control. And of course, G comes for governance. And I want to be very clear here. It's not about creating like a slow, highly bureaucratic committee. And it's not about letting everyone do whatever they want. Good governance is about balancing the two opposing forces of speed versus the safety. And one thing that I want you to do is to have your business units, they need to have the ability to move fast and experiment, ideally in a very safe sandbox. I also want to have your legal and risk teams ensure that you're being safe, secure, compliant and privacy-mindful. And you also have to have an effective AI steering committee, hopefully, or ideally across the entire company of key stakeholders. And their role is not just to approve projects, it's to manage the tension between these two forces and the ROI. And if your governance is all speed, what will happen is that you will have a lot of waste and a lot of risk to the company data and reputation as a result. And if it's all about safety, like what you guarantee is irrelevance and frustration among your employees. So that's the forces that you need to balance. Lastly, we're talking about enablement, E for enablement. And this is more than just offering a training course. It's about being very deliberate about the way you approach upskilling and create a culture that puts experimentation at the center and fighting for whatever human barriers to adoption that might arise in your company. On the upskilling side, companies need to have a very clear plan that look beyond just training employees on basic prompting and using the tools. Those are important, don't get me wrong, but you need to understand that the way your employees do the work is gonna change significantly. And most of them will need to upskill to become ready to manage agents. And most companies are not even there yet and that's something that is very critical. And one reason where upskilling is not enough is that often you have to fight way more than just skills. There are many, many biases and problems and just to name two interesting observations from the audits. The first we're calling that like the business paradox where everyone is so, so busy that there's always chopping wood and never sharpening their axe basically. And they're so buried in the work that they lack the time to learn the tools that should obviously free up time. And it's a matter of deliberately taking people off their hamster wheel and giving them time and permission to slow down and learn and experiment with the ice so they can eventually speed up, hopefully significantly. And the other interesting paradox is change fatigue. And that's applicable to everyone because all of the teams have gone through many tools and methods change over the last years. But what we're seeing is that in companies where there were either major MNAs or reorgs or leadership changes, the change fatigue of their employees is way more prominent, leading them to be very, very wary of AI. And those need to be addressed differently than others. So taking all of that into consideration, you need to give the employees permission and time and everything that they need in order to do good by themselves and eventually by the company. And it's not like a benefit or something that you need to be doing above and beyond. It needs to be like a basic right of employees these days. So these are all the change element that I wanted to share with you today. Becoming agent ready from a culture perspective, it's not just an AT project or something that happens organically. You have to be very deliberate about changing and doing everything that we just talked about. Communication, human oversight, the right attitude, network, smart governance and true enablement. And the ones that we scored highest in the agent readiness, those are the ones that typically are doing all of this well. And to give you the sense that you are not behind, the best time to start be agent ready was a while ago, but I want to give you compassion saying that the second best is today. And I want to encourage you to do all of the things that we just talked about. And of course, next time we'll talk about the other element of agent readiness and that is tech and data readiness. - All right, awesome stuff. It would not be a presentation for the enterprise without a fun acronym, right? - Right. - It's absolutely essential. - So this is super helpful. I love this framework. There's a couple pieces of this that I think are the ones that stand out most to me that get amplified not only over and over again in the interviews that we do, but also when I'm having individual conversations with companies, be they super intelligent customers or AIDV listeners. And one that I think is really interesting is this leadership question. I find myself very frequently in keynotes and speeches and things like that, bringing it back to this leadership question because it is at once the most inescapable part of this, but also the one that leaders have ultimately the most control over. And I think what's interesting is that leadership can make mistakes as relates to AI strategy in two totally different ways. We see leadership employee misalignment where leaders are not sending clear signals around what employees are expected to do or how they're going to be supported in doing it. But we also see the other end of the spectrum where leaders are dropping emails every day about the latest cool tool, but without context, structure, expectation, frameworks put together. And that really leads to that sort of change fatigue that you've said. And by the way, we see the show up in the numbers as well. I know Ryder did a survey last year where they, in December, they released it, where they looked at interviewed 800 managers and 800 employees. And they found that just this vast difference in their belief set around how their companies were doing with AI. I know they're doing another version of that or an updated version of that right now. So I'll be interested to see what that says. But I think that leadership employee misalignment, be it that leaders aren't engaged enough or that they're too engaged, really is, it's almost an unovercomable hurdle when it comes to this stuff. - Yes, I wanna say yes. And 'cause I'm working with a very large company where there is a misalignment between top leadership and also middle leadership. And that's also a huge pain point 'cause we're seeing like the first line and midline managers, they're stuck between a rock and a hard place because senior management are talking to them, like board and management are talking about efficiencies and cost reductions and very, very bullish on AI and how fast they expect to see results where their employees or their engineers are saying, I'm not seeing the value as promised. I'm under a lot of pressure, regardless of AI. And now you're telling me that I need to do something with AI that I'm still not seeing the value. And then these managers are kind of saying, I don't know what to do. Like you're not giving me enough tools, you're giving me a bunch of expectations and the employees are applying pressure. By the way, from multiple directions, some employees will say we want to get so much more time to build and play with AI. And that's perhaps not according to company policies that lets everyone build. We'll talk about it perhaps in the next session. But that too is a big misalignment that needs to be resolved. And I don't have a bulletproof solution for that. It's still a big issue in my opinion. Yeah, I mean, part of why we decided to put culture up front is that these, unfortunately in some ways, are the issues that cannot be outsourced. You can grab frameworks from the outside, you can find people who are good at change management to support, but ultimately these are internal processes that need to be handled internally in conversation dialogue between the different parts of the company. And there's just no shortcuts for that. I think that your point about jobs is one that I echo all the time. It is super important to me. Everyone I think understands, you know, adults being adults understand that it is basically impossible for their companies to say, nothing's ever gonna change, no one's ever gonna get fired, no one, nothing is going to be impacted by this transformational technology. But what people respond well to is leadership articulating how they're viewing AI on a more fundamental level. I've often introduced the sort of heuristic of efficiency AI versus opportunity AI. To what extent is a company thinking about just trying to do the same stuff, but a little bit faster, a little bit cheaper, a little bit better, versus really uncovering new opportunities. Just understanding where an organization thinks in that way can make a huge difference when it comes to employees. I've actually been recently experimenting with POW, P-A-W, Productivity, Automation and Opportunity that really covers the spectrum of get your employees doing their jobs better with co-pilots, automate tasks that can be automated, because almost any tasks that can be automated, people are usually pretty happy to hand off, but also think in terms of opportunity. So that's another big one that I see. The one small point that I thought was really worth honing in on a little bit more is the discussion of the expectations of what to do with time saved from work. Now you had framed it in terms of, are people going to be worried that if they save a bunch of time, their role is not going to be seen as valuable. Another version that I see is people not wanting to just have their work expectations doubled overnight because these tools can happen. And this is actually pretty critical because, especially as you see media articles and things like that, talking about ROI gaps in AI, a lot of that at core, to the extent that there are real issues there, has to do with the difficulty of translating individual productivity gains on an employee by employee level up to the organization level. And having conversations internally about what the changed expectations around how much output you're trying to do or you're trying to have, and just basically how to use that extra time, I think is hugely significant and really, not something that I see a lot of organizations discussing. - All right, yeah, I think there are two reasons for it. By the way, the gap between the individual productivity versus the company level productivity, which is nascent in many situations. At the employee level, most people will attest that they are getting back at least a few hours. And what happens with that either, they just do whatever because the managers don't tell them what to do with the free time. So the best one will go to learn more or do more, but many will just grab their peers and grab more coffee. And thereby creating even more waste than productivity. So that's one problem that there isn't this clear communication. The other is just a matter of shifting bottlenecks. In code, it's very clear that while people are able to deliver more code quickly, the review of the code, and the time it takes to review becomes the clear bottleneck. And then when you look at the overarching productivity for a given team, it's not that significant. So this is what I'm starting to see. And by the way, the discussions are now starting to ramp up towards 2026 plan, I'm hearing more and more organizations starting to say something along the lines, like no more usage and playing with AI, now let's start tracking. And when they're being serious about tracking, that's when these discussions of personal productivity versus team versus overall organizational benefits, what's the sources of all these gaps? One thing that you didn't mention that is tricky because I don't even know where it would fit in and across this framework is the tool quality problem. So one of the big challenges is that organizations often face is in many cases, the tools that they have access to at home after hours with their personal Gmail's are simply straight up and unequivocally better than the tools they have access to at work. Is this a problem that can be solved in any way by different cultures or different managerial styles? Or are we really just stuck hoping that Microsoft gets its stuff together to keep up to capability? I mean, they've added anthropic coding models, it's certainly clear that they're trying to keep up with that, but this is one thing that I think leads to a lot of shadow IT and shadow AI use. It's just the simple gap between the tools that are available to people. - Yeah, and it's also very expensive. Like the companies that are willing to purchase multiple tools for their employees that become like a significant cost uplift and then they go back to the ROI discussion and wonder whether they're getting it back. One thing that I will say here is that whenever I'm talking to employees, they will always say, yeah, I tried it once and it didn't work, so I didn't try it again. So my point is that often even the co-pilots or the tools that they do have access can do way more than what they try. And I always want to encourage people to go back, like even every other week, just to see how fast these tools evolve. But like there is no way around either buying, like just expanding the budgets or hoping that the major players will improve and they do. So that's why I'm not that concerned. I believe that we're a few months to maybe a year tops where the most relevant tools, even if they are not the state of the art, they are good enough for most of what will yield the value for the employees. So if you're frustrated by your employer, wait a little bit longer and I believe that it will be sorted. - The other piece of this, I guess, that is a spot where in that interim period, people could hone in is governance, right? If you have clearly articulated policies around where people can and cannot use external tools and for what, some of this can be taken down because the concern that most people have isn't that they're using, they prefer a clod over co-pilot for writing emails, it's sensitive information, right? It's things that are covered in governance policies. So that's another potential place to look. The two more things that I wanted to just double click on before we get out of here are one, the idea of hiring dedicated AI builders. So this is something that I think is really interesting. I think the champions network intuitively makes sense to a lot of people. We see lots of versions of this, different names, different kind of organizations, but with kind of a common thread of elevating a certain group of people, investing in their time and investing in their ability to share what they're learning about AI. But when it comes to hiring that discrete role, how do you think about that? When someone was writing up a job description for that or where they're looking for that, what is the type of role for that dedicated AI builder and how are they supposed to function inside an organization? - So first of all, that's something that I have a lot of experience 'cause I led such teams within Intel, so I know a lot about it. By the way, the one thing that I tell managers that are contemplating hiring is be willing to open your wallet 'cause that's a very difficult person to hire these days. But when they do, one thing that is a clear differentiator between whether these builders will be successful or not is how plugged in they are to the business and the relationship with the business team. If it is being perceived like I'm the almighty builder, you have to do whatever I say, then there is an internal clash and then we're talking about not invented here between vendors and company. We're talking about not invented here between groups. And that's probably one of the worst ways that you can leverage your internal builders. So make sure that first of all, that it's very clear what is their scope versus what is the team scope, meaning that there needs to be a clear guardrail that below that the teams get to build for themselves and above that it has to go to the builder and what's the rationale to avoid frustration. And also make sure that the business rationale and the technology rationale are clear to both sides such that there is a higher likelihood of the builder to build something that people will actually use. And surprisingly, companies or employees are often more strict versus something that was built internally versus a vendor. So the expectations are extremely high and it's not easy to succeed as an internal builder within especially large companies. - Super interesting. The last thing that I wanted to hone in on and just ask if you had any advice for, I am very frustrated right now by the market's lack of support around training for agent management skills. I agree entirely that this is, that the mindset shift needs to be when it comes to upskilling and support away from just prompt engineering to more kind of agent management. Do you see anyone doing that well right now? Is it all bespoke internal programs? Is it hiring people to come in and do things individually? Are there, there's more off the shelf resources I guess I would say for that or is it still nascent? - So obviously I don't have the full view but from what I'm seeing there is still a high focus on the basics of AI and chasing the tool type of training. So that's what I'm seeing with most training places. There are always, especially in the large ones like LinkedIn learning and Coursera and others. There are some, if you do a lot of cherry picking you might be able to find some of the relevant content hidden here and there but I've yet to see one that is methodologically structuring the process and basically the syllabus of everything that needs to be incorporated in that. Leaving either companies to build for themselves or trainers occasionally to build for themselves. But I think that in many cases it's just that people are not there like they're not seeing what you and I perhaps are seeing that is coming. They're still trying to grasp what is an agent, what is an automation like they're in the first grade when we're talking about a little bit more advanced materials. So I believe that also your listeners are hearing you again and again someone will do that because there is a huge opportunity here. - Yeah, that is a market waiting for a solution. All right, well this is great. can't wait for the next episode. - Awesome.