Managing AI responsibly has never been more important. That’s why Steve Mills, BCG’s chief AI ethics officer, doesn’t just help us assess how we’re using AI; he trains us to think differently about it. While his background is somewhat unconventional, it built his core values of kindness, honesty, and integrity.
Growing up, I was obsessed with the outdoors, but not in the way that most kids are. I spent hours fly fishing, doing macro photography of caterpillars and butterflies, and reading books on animal physiology and insect behavior. I was pretty deeply into it, even for a kid.
And I wasn’t just interested in the outdoors. Although I couldn’t have said it at the time, I was also fascinated by the invisible connections that bind ecosystems together. In an ecosystem, every action has second- and third-order effects. That perspective has never left me. It’s how I approach my work, my family—and now, how I think organizations should approach AI.
Finding My Path
I first earned a degree in wildlife management and went to work with loggers in the woods of West Virginia. Forestry soon turned into a passion for me, but I realized that to build a career in the field, I needed a forestry degree. So I went to Penn State, where I pursued a dual master’s degree in forest management and operations research.
Those studies introduced me to large-scale mathematics, which I loved. They also opened up an entirely different path into software development, then national security, and eventually into AI and analytics.
Looking back, that path seems remarkably coherent. At the time, it was anything but. Yet each turn in my path introduced me to new ways of thinking—and, just as importantly, to people who shaped how I wanted to work and lead.
One of the most influential of those was someone who hired me early in my career and became a great friend and mentor. He had more than 50 years of experience but was still incredibly humble and kind. He gave me some great advice: no matter what, always do the right thing, even if it means turning down work because we aren’t the right people for the job or it’s not consistent with our values.
Seeing him live that way shaped what I began to look for in my own career: work that mattered and people who cared about how it got done. I found both at BCG.
Creating Value at Scale
Prior to joining BCG, I did technical work that I was genuinely proud of—and then watched it get promptly shelved when we were done. But BCG was different and that ability to turn good thinking into real impact, at scale, is rare.
The people surprised me, too. I went through 14 interviews, meeting people who were ten times smarter than me but completely humble about it. Always willing to give candid, truthful feedback. No arrogance. That’s harder to find than it sounds.
The combination—technical depth, real-world impact, and people willing to challenge one another honestly—became especially important as I thought about how to deliver the next level of impact in my career.
The AI Journey
When I first started working on responsible AI in 2014–2015, we didn’t even have a name for it. It was just a bunch of us talking about the impact of data science on individuals. I still remember the conversation with my team when it clicked: algorithms already being deployed for things like news recommendations were quietly shaping how millions of people saw the world, and nobody was talking about it.
We thought it could become a real problem long term, so we started researching and writing about it. That’s what pulled me from the technical side into questions about governance and risk.
For a long time, those concerns about AI’s broader risks remained largely confined to specialists. As long as the handful of companies building these systems understood the risks and took actions to mitigate them, it felt manageable. But ChatGPT changed everything. Overnight, AI went from a bespoke tool in expert hands to something everyone, everywhere, could use. That was the inflection point. Managing AI responsibly would take a completely different kind of effort—and still does.
Maximum Benefit, Minimum Risk
As BCG’s chief AI ethics officer, I find that effort is much more hands-on than people sometimes imagine. I spend a lot of my time working alongside business and functional leaders on the products we’re building—weighing risks, benefits, and regulatory implications and ensuring that the right governance is in place.
My role is less about making decisions at the end of a process and more about helping teams ask the right questions from the beginning. We work together to consider the potential benefits and risks, refine how a tool should be used, and ensure that the appropriate safeguards and governance are built in.
Done well, that kind of governance doesn’t slow innovation; it can actually speed it up. When everyone understands what they can and can’t do, and they know when to pause and have a conversation, they can move quickly and confidently.
But governance can’t scale if it depends on a small group of people making every decision. You have to build that judgment across the entire organization. That’s why I am proud of how we’ve approached building AI fluency across BCG.
We think of it as a ladder. At the first rung, people are using AI to accelerate their everyday work. As they build their skills and demonstrate they can use AI responsibly, they move up a run, beginning to build more sophisticated tools, like automated workflows or advanced applications, that would have demanded a specialist team not long ago.
That step-by-step progress up the ladder is deliberate. If you hand someone access to a powerful tool before they understand how to use it well, they will struggle, and the value will never materialize. But if you build the skills and the understanding of risks and mitigations first and then open the access, the capabilities stick.
Everyone has a path up that ladder, regardless of where they start or how technical their background is. And responsible AI isn’t a separate module bolted onto the training; it’s built into every rung. The approach becomes even more important when you consider how much we still don’t understand about AI’s long-term effects.
The Future of AI
I think a lot about how AI will quietly change people. The biggest effects aren’t the ones dominating headlines. They’re the second- and third-order consequences we’re only beginning to see.
Take a hypothetical example. Public transportation networks can use AI to optimize train schedules—setting departure times, the number of cars in each train, which towns get stops, and so forth. But what if AI optimizes the routes with the heaviest ridership, eliminating lesser-used routes to cut costs?
In the moment, the savings seem like a good outcome. But what if the changes eliminate the only means of transportation for some communities to their places of employment? Cutting off this critical service would make it very hard for the people on these lines to get to work. Or they might be forced to take earlier trains, spending less time with their families.
The problem is never as simple as it appears on the surface. It’s exactly the kind of downstream consequence I learned to look for in nature, where changing one part of a system quietly reshapes everything connected to it.
I have the same unease about how AI might change the way we learn. There are many tasks that AI can easily perform that also help people build sound judgment and the foundational knowledge that becomes important later in their career. So, what happens when AI performs those tasks? If junior staff, for example, never build a work plan themselves, how will they recognize years later, when they are managers, if their team is headed off course?
More broadly, what happens as people become increasingly confident in systems they don’t fully understand?
In many cases, the honest answer is that we don’t know yet. I can’t pretend otherwise. To me, leadership means acknowledging that uncertainty and building organizations that can learn their way through it.
Building Better Judgment
That’s ultimately how I think about responsible AI. It isn’t a final checkpoint at the end of development; it’s a way of thinking that should shape every decision from the start. The goal is to help ask the right questions early and often: What are we optimizing for? Who could be affected? What unintended consequences might emerge over time?
When something does need closer attention here at BCG, it’s never for me alone to decide. I work alongside a broad governance team to refine and strengthen the approach together. More than any policy, what we’re really building is sound judgment across BCG and our clients, embedding responsible thinking so deeply that it becomes instinct rather than instruction.
In a moment when AI is moving faster than our understanding of its human consequences, I believe our greatest responsibility is to build organizations capable of asking better questions as the future unfolds.
