Leadership
Trust, Clarity, and AI
AI didn't invent ambiguity. It just stopped letting you defer the bill. Trust, clarity, and empowerment matter more now than they ever have.
There is a classic bit from The Simpsons where Homer pours mayonnaise into a jar of vodka and is about to drink it. Marge tells him she doesn’t think that’s such a good idea. Homer shrugs.
That’s a problem for future Homer. Man, I don’t envy that guy.
For most of my career, software teams have been very, very good at creating problems for future Homer.
We’ll do agile and figure it out as we go. We’ll skip design entirely because someone said you should avoid “big design up front.” The sprints and months go by and the ambiguity we didn’t resolve up front gets paid down slowly over time by whoever happens to be around when the bill comes due. The cost of ambiguity never went away. It was simply amortized over time to some future person, who often turned out to be ourselves. And the cost was compounded every time new ambiguity was added on top of the old.
Today the challenge is even more difficult because AI does not eliminate ambiguity. In fact the ambiguity that used to hit you down the road now shows up today and in spades, smacking you upside the head. Your cognitive load can’t carry the weight of this ambiguity machine we have foisted on ourselves. However, some things haven’t changed.
What Hasn’t Changed
In September, 2026 I presented a talk at The Software Leadership Summit in Drammen, Norway titled Trust, Clarity, and AI: Empowering Teams in an Era of Ambiguity. It was the least technical talk of the conference, and that was on purpose. Most of what we discussed over those few days was about tools, methods, and techniques for getting good results out of AI. I wanted to talk about the people.
For years now I’ve framed everything I do with software teams around three core values: trust, clarity, and empowerment. I began operating with these values during my graduate studies in the late 2000’s. I wrote about them in Software Leadership Distilled back in 2020, long before any of our teams were running dozens of agent sessions in parallel.
Before we get to how these values apply in an AI-driven world, let’s take a closer look at the relationship between AI and ambiguity.
The Relationship between AI and Ambiguity
AI Is an Ambiguity Accelerator and Amplifier
It accelerates ambiguity because the unclear thinking you used to pay down slowly over months now shows up immediately. It’s right there in your face, staring you down. AI collapses the deferral window to zero.
It amplifies ambiguity because of scale. Walls of generated text. Unreviewable pull requests awash in green and red. Of course we had these before AI. Developers have always been capable of producing them. But the pace was slower and the scale was nothing like what we see today.
Ambiguity was always the most expensive thing we were dealing with. Now it’s arriving at lightning speed.
Consider a simple three-step process with humans involved in every step. Put AI in step B. What happens? Step A is still slow, so step B sits there waiting. It doesn’t matter how fast it is. Then step B hands step C a mountain of output to review, and step C chokes. The bottlenecks don’t disappear. They move.
A fundamental insight is the recognition that none of this is a tooling problem.
The problem is people, and the solution is people. And that’s the problem.
The tool is not going to save us. Maybe someday the tools will get so good we won’t care anymore. Even then, the questions remain. How will we trust those systems? How will we have clarity that we’re getting the outcomes we want, and that those outcomes are sustainable and maintainable? Those are human questions, and they don’t go away.
Trust: You Don’t Know Until You Take the Step
Trust grows by making and keeping agreements. That has always been true, and it’s still the foundation. But there is a subtler point I’ve come to appreciate, and it came from watching a colleague of mine wrestle with it for more than a decade.
Years ago, this colleague attended an architecture master class. He walked out convinced. This was going to be transformative! And then Monday came. And he went right back to doing what he had always done. For over ten years.
He wasn’t lazy and he wasn’t unintelligent. He agreed with every word. But accepting something intellectually is not the same as trusting it.
If you’ve ever been to the top of the Sears Tower (yes, I know, the Willis Tower) in Chicago, you know there’s a glass box where you can walk out over the edge and stand on a clear floor, a hundred-plus stories above the city. You can believe all day long that the floor will hold you. How do you know if you actually trust it?
You take the step.
My colleague eventually took the step. When he did, he didn’t just practice what he’d learned, he helped me build a community of practice around it so others could take the step too.
That applies to the tools as well. When I first started leading a team with a heavy AI mandate, my very first instinct was to get to determinism as fast as I could. When you’re working with a tool that gives you probabilistic outputs, what you need to get to is a deterministic ground truth. So we used AI to build tools, wrote tests for those tools, and set up guardrails that gave us objective evidence the system was behaving the way we intended.
We weren’t trying to make the AI smarter. That’s somebody else’s job. We were trying to reduce ambiguity for us and for AI agents. We needed some ground to stand on. Something we could trust and verify.
Clarity: It’s Worth Fighting For
Once you have the beginnings of trust, you can start to fight for clarity.
When I form a team or join one, the first and most important agreement I seek is an agreement about what we need to be clear about. And the first place to look is roles. Who owns what? Who is accountable for what? Who is responsible for what? Not nearly enough teams speak the language of ownership, accountability, and responsibility. (I’ve written more about why role clarity is a liberating constraint.)
I’ve been teaching teams for years to fight for clarity. When something is unclear, it is time to take a step back. Maybe that’s a retrospective. Maybe it’s a quick conversation. Whatever the format, when we discover a significant ambiguity, we come together and we fight for clarity.
Notice the preposition. We fight for clarity, not with each other about it.
This has never been more important than it is now. Think about all the effort our industry is pouring into AI right now. All the tools, all the approaches, all the methodologies, all the agent factories and prompt libraries and instruction files. What are we desperately clamoring for?
Clarity.
And it’s hard. It’s hard to get clarity out of a machine, and it’s even harder when the humans directing the machine aren’t clear with each other. That’s why we must fight. The ambiguity is coming right at us every day, and the teams that win are the ones that have learned to stop, step aside, and get clarity before the next wave hits.
Empowerment: The Loop Runs Both Ways
Once there is trust, because agreements have been made and kept, and once you’ve begun to fight for clarity about roles, responsibilities, and how you’ll operate together, something wonderful happens. You can begin to empower each other.
Empowerment means I know what I do, I know what I’m good at, and I know my limits. I know the same about you. I know how to help you be better at what you do, and you know how to help me be better at what I do.
When we empower each other, trust grows. With more trust, we can fight for even more clarity. With more clarity, we can empower each other even more effectively. That’s a positive feedback loop.
But the loop runs in the other direction too. When people get stressed, they run to their corners. They stop empowering each other. They stop connecting. They stop fighting for clarity. Trust erodes. Then there’s less empowerment, and less clarity, and even less trust.
Anyone who has studied systems knows that’s also a positive feedback loop. In systems engineering, positive doesn’t mean good. It means self-reinforcing.
AI pours gasoline on whichever loop your team is already in. A team with trust and clarity can use AI as a tool for empowerment, speed and scale. A team without them will use AI to generate ambiguity faster than ever before.
Replace Each Other, or Supercharge Each Other?
Humans thrive most as value creators. Coding was never the highest-value task. It was always design. It was knowing what to write, where to write it, and why to write it. Whether it was testable, scalable, and maintainable.
So if AI takes the coding off our hands, good. That means we get more time for the part that always mattered most. Designing. Clarifying. Empowering. Scaling. Building trust with the people next to us.
Or we don’t. We can just as easily spend that time jumping from one AI cycle to the next to the next, alone with our agents, and never look up.
So here is the question I’ll leave with you:
Are we going to use AI to replace each other, or are we going to use AI to supercharge each other?
It’s your choice. Trust, clarity, and empowerment help you make the right one.