#115 From Experiments to Impact: Building Human-AI Collaboration with Todd McLees

Welcome to the Proteus Leader Show, a podcast for forward-looking leaders driving transformation in their organizations, within their teams, or for themselves. Your host today is a senior leader at Proteus International, a firm that partners with companies around the world to make true transformation easier through change consulting, leadership development, and executive coaching. If you're looking for fresh perspectives and practical insights, can take back to work today, you're in the right place. Enjoy the episode.

Hi, everyone. Welcome to the Proteus Leader Show. Today, we're exploring how AI and human skills can amplify each other. I'm joined by Todd McLeese, co-founder of humanskills.ai, creator of AI Agility, Harvard Business Review contributor, and global speaker whose work focuses on making AI human-first for enterprises and higher education. I recently completed Tod's training and I am excited about our nascent partnership. The AI Agility Challenge was packed with frameworks, prompts, examples that move people from AI anxiety to real capability. We'll dig into how to build practical collaboration patterns with AI, what better machines demand better humans looks like at work, and how to turn tactical experiments into strategic value. Let's get into it. Welcome to the show, Todd.

Marie, thanks for having me. I'm really excited to have a conversation.

Thank you for being here. Let's go straight into it. You talk about moving teams from AI anxiety to AI agility. What does that journey look like for a typical leader, and what mindsets tend to unlock the shift fastest?

Most teams start in a similar place. They've heard the hype about AI, they feel the pressure, they've heard about the upside and the downside, and they're not sure where to begin exactly beyond simple experiments. Anxiety comes from not knowing how to use the technology in a way that actually helps their work or their enterprise in a leadership role. The shift to AI agility happens when they see that they're the ones driving. When you give people a few simple patterns, clarify the outcomes, break work into smaller pieces, and run short feedback loops, it just seems like most of the time something unlocks inside individually and across teams. So they start to feel control and they gain some confidence. To me, the mindsets that move fastest are probably a beginner's mindset, Mm-hmm. Even for experts in certain domains, we need the domain expertise, but bringing a beginner's mindset and being open to what AI can deliver and what it shouldn't deliver really adds. And then I think maybe the most important thing and why we spend so much time on a human-first approach is a sense of agency. You choose what gets automated, what you augment, and what stays human. Once people feel that sense of human agency, The whole room starts to change, the whole effort. So anxiety drops, curiosity goes up, and from there they start to build real capability.

So it's really breaking it down into simple steps. That was the first thing you talked about. And then giving people the choice of what they want to use AI for, that agency. Can you talk about that term? Because we hear it a lot. How do we give agency to people?

Many people think it's innate. We believe it's developed. Unfortunately, less than half of people at work are high agency or identify as high agency. When we roll out of bed in the morning, we feel a sense of control over what's gonna happen that day. We know that we're in charge of our own destiny on a given day, at least most of the time.

Hmm.

We have some skills, so we have self-efficacy to deliver against what we're trying to get done. And we also feel a sense of accountability Over how the day goes. So we own it. Not everybody feels that way. Less than half of us at work, less than a third of college students, community colleges. When you start working your way into underprivileged societies and so forth, agency continues to go down. The lower my sense of self-agency or human agency, the more threat I see in AI. If I came to you and I said, Marie, 10% of your job is changing in the next couple of weeks because we figured out a way to automate some things. What would you say?

Great, I can do other things.

Yeah, exactly. That's what I would say too. Now, when do I start to get concerned? Well, 30% of my job is changing, 50%. Where somewhere there's an individual threshold of pain for change that we all have. It may not be directly correlated to human agency, but If you are low agency and I come in and I say 10%, that feels like you're taking something away from me. And it requires a great deal of empathy because that 10% might be how I've put food on the table. It might be how I built relationships. I may have some self-worth wrapped up in that 10%. We know that jobs are going to change.

Mm-hmm.

Somewhere between 10% and 50% of jobs, depending on the role, during this decade. What you think is painless is not necessarily what somebody else sees as painless. And so, we just have to have that in mind when we're thinking about the decisions that we're making inside of organizations.

Yeah. Change is at the core of what we do at Proteus. And what we see is the leaders, they have usually a lot of agency. They understand why the change is happening. And as you try to take people with you on that change arc, you need a lot of empathy. So, that's a critical skill to understand where people are at, because they will have different curves in terms of coming with you. And we need to meet them where they are and spend the time to explain that change. And it's probably the biggest change we're going through right now is AI.

That's right. And thousands of people having now completed the AI Agility Challenge, I think we can also draw a line to high agency typically translates to, I have a better idea of what I would do with AI to achieve more or get things done faster so I can spend more time on higher-order tasks. If I don't have that sense of agency, it's harder for me to envision this as something that can augment my capabilities and help me go further.

So if leaders had to prioritize a shortlist of human capabilities to thrive with GenAI, like, Clear thinking or strategic framing, ethical judgment, etc., etc. What would make your top 5 and why?

I think about the human skills stack as the things that they separate teams and people from those who get 10% better incrementally and those who get 10x better. When I think about the top 5, I think you're right there with strategic thinking or critical thinking. I want to go a little deeper into the sub-skills or micro-skills related to critical thinking, because I think Depending on the organization, there are different flavors of critical thinking, so we have to think contextually about those things. I think—

Yeah.

Going from tactical to strategic gets me into a place where I can think about how work gets done today and the impact that AI will have on that. So I think strategy and going beyond how do I use AI in a transactional way to get things done faster, to how does it impact workflows is really important. Ethical discernment is huge today, and not only in terms of when should I use AI, my ability to be transparent about its use, but also the outputs, being able to discern the quality of the outputs and the impact that those outputs might have on different people in—

Yes.

My recipients, my audience. I think prompt design as opposed to prompt engineering is really critical today in order to get consistent work product out. I think there's somewhere between intentional use, so being intentional in the way that I'm using it, and then not only reviewing the AI's work and what the output is, but also some metacognitive reflection. So my ability to review my own thinking and the impact that I had on the work that was produced. I don't see those as academic skills. I see them as— Driving the work that needs to get done.

Maybe to illustrate those skills, if you're happy to share, because I loved your frameworks, maybe you can share with our listeners one kind of key way to do a prompt or an example.

In the AI Agility Challenge, we roll out a couple of different prompt frameworks and see them as complementary. And you don't always use every aspect of these frameworks, but the first fundamental one, just to make a real difference in how you're using the technology today, if you're just using it conversationally, which I do, often. But if the stakes are higher and, you know, what I get out of the AI is really important, then I might turn to something like our first framework, which is I give the AI a role so that I can get it thinking along the lines of the expertise that I need to call on. If I were calling on a person to help me, then I would go to the expert that I could find. And so I'm asking the AI to draw on that knowledge. I think the importance of assigning a role has changed a little bit with reasoning models. It's a little different than it was a year ago, but it also gets me in the frame of mind of who I'm communicating with. The next piece is around being clear on the objective.

Yep.

I think audience, you really have to pay some attention to because you're really just trying to hone in on what's the ideal output. And then lastly, context and Being able to find that balance of giving the AI enough context to put it in a position to be successful without overfeeding it, overproviding context where maybe it gets lost a little bit, not because of its limitations, but because you're being a little less clear on what you're looking for.

Yeah, you don't want to overload the tool, but give them enough, so the right parameters, so you get the quality output. And even I taught my son, he's 12, to think about those different elements. Who are you doing that for? What expertise do you need? What's the objective? And in which context is it, like for your homework? So, I think this is fundamental and so clear in the training. I always think about ROARC each time I go to an AI tool. Thank you for sharing.

That's the acronym. Let me just say this, Maria. I think if I hired your 12-year-old son to join our team right now, I wouldn't just say, here's what I need, brilliant 12-year-old. I would say, this is what I need. These are what my expectations are. Here's who it's for. Now let me give you some documents, some knowledge that you might not be familiar with, but will be helpful. In fact, many people are migrating from the thought around prompt engineering to context engineering because that's the most critical piece.

Mm-hmm.

How much do I need to provide to the AI to give it the best chance at providing a successful first output. The whole reason why you invest time on the front end with prompting techniques is so you can spend less time iterating. So there should be a formula there. If I can spend the right amount of time, but not too much on the front end, and give it the right amount of information, but not too much, then I can spend a little less time with quality assurance and elevating the work of AI and so forth.

Yes. And we use less energy too, if we just—

And we use less energy with fewer iterations. That's right.

Also, sometimes you don't know exactly what the tool might need, so you could use the tool as a sounding board and say, what might I be missing, or what else do you need, and make it more fluid and conversational, because it's hard to guess all at once what are the exact parameters.

Yeah, I think if you were working with somebody, you might end that conversation where you say, what else do you need to be successful? And the reality is that AI does a pretty good job of answering that question today.

Yes. I actually find that talking to AI, it's a little bit like being a leader. You need to make clear agreements. You need to be very clear about expectations. You need to validate with the person that they understood the assignment and what support they might need. So I think good leaders would have the core skills to be good prompt engineers Maybe as we're talking, I'm like, those are leadership skills.

No question. And when you listen to the AI technology community talk about the future of AI agents and agentic AI and the reality that we give agents the capabilities, the constraints, and the context, when you've been in business for a few decades, you realize that not everybody would be a good leader or a good manager. And when you listen to the technologists say, well, essentially everybody's going to get a promotion to be a manager of 10 or 12 agents alongside of the people that they're working with. Well, not everybody's going to be great at orchestrating that. They won't have those skills. That's the next piece of skills that we need to teach people are exactly what you mentioned, those leadership skills. Because if we're going to be expected to, and I'll use some AI bubble language, but orchestrate a swarm of agents.

Yep.

That's going to require this extraordinary level of clarity and communication because, you know, in most of the processes that we've designed historically to get work done, we have high performers who will figure it out if something goes awry of the typical path. With agents, we have to be more precise about process. We have to be more clear about our expectations because they don't have the adaptive capability today to really figure out what we meant. Or work through exceptions on their own.

Yeah, I think this is powerful and so connected to that leadership capability. Now, I'd like to go back to ethics. It's a big topic for me. You emphasize designing optimal collaboration patterns with AI and setting clear ethical boundaries. So what good team norms look like in practice for you?

I think the best teams build simple habits, and they keep humans not in the loop, but in the driver's seat. That's one of my least favorite sayings is human in the loop. I think there are certainly processes that we can design where the human has oversight and is dealing with exceptions and so forth. I think there's a lot of that. But I think generally speaking, when we're thinking about generative AI and large language models, it's more about humans driving, humans at the heart, if you will, or humans at the helm, as opposed to in the loop. Oftentimes some habits might look like starting the day with a 10-minute check-in, listing tasks, asking AI to structure the day, and agreeing on what you'll outsource to the machine and what stays human. This is such an important piece. Maybe the best place to start— there was actually a framework recently that was published in Harvard Business Review, not one of mine, but another team of researchers, and they were talking about the 2 axes by which I, I make decisions as to How much AI to apply. And the first piece had to do with, uh, the y-axis was tacit knowledge versus data-driven or structured data.

Right.

And so when the knowledge is experiential, it leans more towards human. And then the second, the x-axis was around the stakes. So does it matter? I suppose it's like Amazon's one-way or two-way door for innovation. How expensive is it if I get it wrong? There are examples of companies who have turned to AI agents too early, and it costs them customer relationships and so forth. AI becoming the customer service team instead of people.

Yes, I've heard the stories. And then they're rehiring the team, right?

That's right. They're rehiring the team. They're asking engineers to work the phones. They're doing all sorts of things to solve the gaps that were caused by being out a little bit too far with the technology. Those are the 2 questions. Is this task based on experiential knowledge? lived experience that AI doesn't have, or is it based on data? And the second is, what happens if we get it wrong? Is it too expensive? Does it represent too much risk?

Mm-hmm.

And so I think that those are pretty good decision points, but oftentimes it does come down to guardrails and creating an environment where we can set ourselves up for alignment. So can we be explicit about what's off limits regardless of which business unit or department we're talking about? Can we require the verification steps that are necessary and make accountability clear because no matter how smart AI gets, the humans, for the foreseeable future, humans are going to own the outcome and the company's going to own the outcome. So teams that treat AI like some sort of magic trick instead of a collaborator, or they hand over too much thinking and suddenly they see skills and decision points beginning to erode. So we often talk about helping establish healthy habits This is true of, you know, young students, and it's true of mid-career professionals and organizations.

Yes. So daily rituals, healthy habits, and clearly defined roles between humans and AI. And that makes me think about change, because often people, they talk about what's changing, but often we ask, what is not changing? And in this case, what is staying human? And I think that would help also to reduce That anxiety, a number of tasks will continue to be human. And here's why. So we only talk about what's changing.

That's right. It's an interesting point. Over the weekend, I was looking at some blogs, like Andrew Yang was just out with a blog post about the jobs that are most likely to be lost and the jobs that are most protected. And oftentimes you'll see, well, where physical manual labor is involved, they're the most protected. And then these low-risk, low-cognitive, but still cognitive work jobs are most at risk. And there's probably some truth to that. What we never see is an analysis, actual research on what are the jobs where AI can play an augmenting role, where the value of those roles when you combine humans and AI explodes. So how do I use AI to augment people? And the starting point for that conversation is What tasks or processes or workflows in our company have to remain human? And then we can work our way down the human agency scale instead of working our way up on the autonomous AI scale. Stanford and Erik Brynjolfsson and a team of researchers at Stanford University about 4 or 5 months ago released the human agency scale. And on one end of the spectrum, it's fully autonomous AI they refer to as H1. And on the H5 end of the scale is purely human work. H3 in the center is shared agency, co-creation, et cetera. If I look at any process today, I think about distilling it into tasks and determining H1 to H5.

Mm-hmm.

How much AI can I apply when? How much do I need to rely on people to do this work? And it's a really useful framework for thinking about how work is going to change, but also to your point, What's going to stay the same?

I think I always have your pyramid that you shared with us, and it's also in the Harvard Business Review that I'll share with the listeners. Many organizations are stuck at pilot stage experiments, one-offs, automation. And I'm really curious about how you help organizations go from creating tactical value to transformational value. Like, what are those high-leverage cases So can you maybe ground our listeners with that pyramid and share your wisdom on that topic?

So when we begin working with an organization, or when an organization begins experimenting on their own, they're collecting novelty wins. And it usually comes down to individual optimization or improvements. And I think that's a great starting place, actually, because we need people to see what's in it for them. Someone tries a tool, someone builds a clever workflow, creates a GPT, whatever the case may be, but nothing changes about how the work is actually done. It's just that certain elements go a little faster or maybe hopefully increase in quality because they're augmented, et cetera. When we work our way up to a shared intelligence across the organization, or we start thinking about transformation and growth, now we're focusing on systems and not one-off tasks. So we start with the work itself, map it out, like I mentioned before, one healthy scale there is H1 to H5, but we map out where time is lost, where judgment is required, where accuracy matters more than other places, and where variation is just killing quality. And if we can bundle those tasks into repeatable workflows, now we can start to codify what works and we can make it easy for somebody else or another team to use the same approach. We worked with a very small manufacturer, less than 50 employees, who had invested the last couple of years in AI language models, and they had some scattered wins that were super interesting, but still isolated. Once we mapped their customer response process as a full workflow, from intake of that customer to analyzing and drafting and verification, they cut cycle time dramatically. The team didn't feel replaced in that case. They felt like their capabilities were amplified. So I think that's the difference between transactional AI and creating transformational value.

And it feels like a necessary step is to map the process and then to categorize the different steps of the process, maybe using that H1 to H5 model. But what parts of the process would really make a difference if they Augmented with AI?

H1 to H5 is my favorite framework. We use it all the time now in that work. You're right, we have to start with mapping process, mapping workflows, and then at a minimum deciding of this work, what is automatable, what is augmentable, and what is purely human. In our core values, in the way that we want to conduct business in our entity, what work is remaining human for now? And then we can work our way to how do we augment people?

Let's start there.

Let's start there.

What can we automate? I'm with you, Todd.

The next step, I think, is augmentation. How do we help our team feel more empowered and capture and create more value? And that can be about augmenting. The automation pieces are certainly going to happen, and I think there's a place where AI-first mindset has a place. Because I want to free people up to do higher-level tasks. But I think if we start there, we're turning more people away from the effort rather than saying, okay, now that all of this is there, what would you automate? What part of your job do you not like that AI can do really well today? Things like that. There's all sorts of opportunities. Marie, where we need to get to as companies is Warren Berger wrote a book called A More Beautiful Question in 2014. And if we could get to the point where we could start asking questions like, what's going so well in our business right now, or our institution of higher education, or whatever the entity, what's going so well that we would 10x it if we had the resources? Just about every process and workflow that we've ever designed is Built on scarcity, access to resources, access to time, access to intelligence. We have designed our businesses around preserving the time of our most expensive and most talented resources. And so in isolated cases where we have access to the type of intelligence that, that we can make use of to multiply and amplify, maybe we can start looking at things like That 10x question. What would we 10x that's going really well today if we had access to more? And then what challenges are we faced with that are so difficult to solve? We can't hire enough people or consultants or whatever the case may be. We don't have enough access to intelligence through technology. We can bring that in and we can start to work on those things too. If we can help people get to that phase, now we're winning and they're gonna feel more empowered and feel like they're winning too.

Definitely more exciting when you frame it this way.

Right.

For sure. Yeah. We love leaving our listeners with practical tips, and it was full of that in this entire conversation. But maybe as a recap, what are 3 to 5 concrete practices that leaders can start doing this week? You can go into starter prompts, but how can we raise the quality of AI-augmented work?

I think the first thing is to realize that this is not a time that calls for us to offload our thinking to a fault. We actually need more precision. We need to be a little sharper in the way that we're working with AI. I guess my first practical tip is write sharper inputs, invest a little more time on the front end of the process and save significant time on the back end and energy.

Yeah.

So just think of in terms of 2 clean sentences of context that can transform the output.

Mm-hmm.

I think the second thing to maybe think about is how do I add a second brain step? So before making a decision, can I ask AI to review my thinking? This is an underutilized use case is around simulating. So not for every meeting, certainly, because it would just increase workload, but If I'm going into a high-stakes conversation with multiple stakeholders or a single stakeholder, I can simulate that discussion before it ever happens. If I'm going to launch a new webpage or product page, I can form today a digital customer advisory board using the personas of my key customers and getting their input, not to replace actual human input, but to prepare me to increase the quality of that input from the people I actually lean on. I can use them a little less often. So I think just building the discipline to use AI as a second brain to check my work before I put it out into the wild. And then the last practice I would say, and this is something that I'm really working on my own, is to build a daily microhabit. Each day I end with, what should I prepare or refine tonight so tomorrow runs smoother? And that's focused on not only my calendar and what's coming up, but the work that I did today that will continue tomorrow. And how can I improve the way I work today? Now, unlike 6 months ago, AI didn't have working memory, a long-term memory. So every session felt like a new session. with some preferences built in. Now it's remembering things from 18 months ago that were in ChatGPT or Claude or Gemini. Sometimes that can actually lead to an issue where it's recalling something from 18 months ago that you've evolved past, and you need to catch that too. But taking advantage of the memory to say, how can I get a little bit better tomorrow? How can I make tomorrow Run more smoothly is important.

I love that daily practice. And probably you need to give a bit of context, so you could share how you felt about certain moments of the day, and I didn't feel efficient at that moment, so you get better advice for the following day. So context.

I think that's right.

Daily practice.

I think that's right.

I've built my personal advisory board as you recommended in the training. So connected to Obviously, your different roles. And it's, again, not to replace humans, but to be more prepared for the conversations.

That's right. You can literally say to any of the major AI tools today, here are the 4 meetings I have coming up tomorrow. Based on the work we've done in the last few days, what would you have me focus on? Or what do I need to refine before I go into those meetings tomorrow in terms of my thinking? What did I leave undone today? So that I'm not articulating something I haven't fully processed, things like that. In manufacturing and other environments, we've been talking about digital twins for years now, several years, and using them effectively to simulate ways to automate manufacturing processes and so forth.

Yeah.

Now there's an opportunity to simulate a discussion that can lead to a deeper exploration with a colleague or a customer and increase the quality of those human connections. So we think of that as human to the power of AI. So how do I build my relationships not in a human plus AI way or human times AI, but with AI as an exponential for my own human capabilities?

Todd, thank you so much for such a clear human-first lens. That's important for me on working with AI, shifting mindsets, building practical collaboration patterns, and investing in that distinctly human skills that make the tech valuable. So if you want to know more about humanskills.ai, check Todd's website and follow Todd. If you'd like more resources on change, leadership, and transformation that integrate AI, visit produce-international.com. Thank you for tuning in, and until next time, here's to co-creating the human-centered AI partnerships You truly want. Thank you, Todd.

Thank you, Marie. Really enjoyed the conversation. Thanks, everyone.

Thanks for listening to the Proteus Leader Show. We hope you're leaving with ideas you can put to use on your team, in your organization, or in your own career. To learn more about how we support true transformation through change consulting, leadership development, and executive coaching, visit us at proteus-international.com. Until next time.