The frontier labs are blazing the trail. They’re the ones out ahead, clearing ground toward something none of us can see the end of. That’s hard work and I’m glad somebody’s doing it.
The rest of us are on the trail behind them. Following, mostly. And the trail isn’t a highway — it’s unmarked in places, dark in others, hard going. It’s easy to veer off it and not notice for a while.
So there’s a role in between. A guide. A scout. Somebody who runs up ahead, talks to the trailblazers, gets a look at what they’re building — and then comes back down the trail to everyone else and says: here’s what’s up there, here’s what it means for you, here’s where the footing is bad.
Then does it again. And again. Back and forth along the same stretch, until the path is worn in and marked.
That’s what the stack of stones in our logo is. A cairn — what a hiker leaves at a turn so the next person doesn’t get lost. Somebody already walked this. That’s the whole idea.
I want to be clear about what a scout is not. We’re not inventing the future — that’s a different job, and it’s happening somewhere else. But we’re in it. We see the promise, because we’re using this stuff on real work every day. And we see the risk, because you can’t walk a trail this often without finding the places it drops off.
Which is also why the job doesn’t expire. The frontier keeps moving. The trail gets longer. Somebody has to keep going up and coming back.
And it’s why the right way to do this is to speak from having actually walked it, and to share what you found — freely. Not as an authority on a technology nobody is yet an authority on. As someone who’s been a little further up the path than you have, this month, and came back to say so.
Jonathan Levav, who teaches AI strategy at Stanford’s Graduate School of Business, has the evidence: AI adoption stalls on human barriers, not technical ones. Diagnose the psychology before you touch the tooling.
I taught that finding for a year before I noticed what was missing from it. “It’s psychological” tells you where to look. It doesn’t tell you what you’ll find when you get there.
What I needed was a map.
Trust has three legs. When it breaks, it almost never breaks everywhere. It breaks on one leg — and that leg has a name.
Frances Frei teaches at Harvard Business School, and was brought into Uber in 2017 to help rebuild its culture. With Anne Morriss she argues that trust isn’t a mood or a chemistry — it’s built on three drivers. People trust you when they believe three things: your reasoning is sound and they can follow it (logic), you care about them and not just the outcome (empathy), and they’re getting the real you (authenticity).
Frei is the first to say the idea is old — it’s Aristotle’s ethos, logos, pathos, rebuilt for people who don’t think in Greek. What’s new is the diagnostic. When trust breaks, one leg is wobbling. She calls it your wobble.
That’s what turns the framework from decorative into something you can use on a Tuesday. “They don’t trust me” is a condition — you can brood on it, you can’t act on it. “They doubt my logic” is a diagnosis, and a to-do list.
And each wobble has its own repair. Candor for authenticity. Shown work for logic. Demonstrated stakes for empathy. The discipline is working the broken leg — because the instinct, and I catch this in myself as much as in my clients, is to push harder on the legs that are already strong. The caring leader cares harder. The brilliant leader explains again, louder. The broken leg goes untouched.
She wrote it about people. I use it for AI — because when someone resists an AI system, they distrust it the way they’d distrust a person. And it wobbles one leg at a time.
The Trust Triangle and the wobble diagnostic are Frances Frei and Anne Morriss’s — “Begin with Trust,” Harvard Business Review, May–June 2020, and Unleashed (Harvard Business Review Press, 2020), building on Aristotle. We didn’t invent it. We just think it’s the clearest map anyone has drawn for this.
This is where most of the doubt sits, and it deserves a better answer than a demo.
Executives who’d never cite an unverified number in a board meeting are being asked to rely on a system that occasionally invents one with total confidence. That hesitation isn’t technophobia. It’s the same judgment that made them good operators.
And notice what’s actually wobbling. Not the model’s average accuracy — which is often better than the analyst it supplements. What’s wobbling is the legibility of the reasoning. People extend logic-trust when they can follow how a conclusion was reached and check it. A black box that’s usually right earns less trust than a colleague who’s sometimes wrong and shows their work.
So the repair isn’t a better model. It’s grounding the system in sources your people already trust — their data, their documents, their numbers — and making verification cheap.
Once your AI is secure and its output is reliable, the work moves faster. Speed shows up as cost. Capacity shows up as more clients in front of you. And once you trust it enough to build on it, you start creating things that weren't on the table before — new revenue, not just the same work done quicker.
Your team is asking a version of this quietly. Is the owner trying to replace me? Do the people who built this care about businesses like ours, or are they making billions and telling us it's for our own good? Those are fair questions, and they don't go away because somebody ran a training.
AI systems fail this test in ways people feel before they can articulate: the tool that optimizes for engagement over usefulness, the vendor whose incentives sit inside the product, the rollout that plainly serves the budget rather than the people using it.
The repair is to show whose interests are inside the system’s calculus — in how it’s configured, not just in how it’s announced. No capability fixes an empathy wobble, because the resistance was never about capability.
Push on an AI’s conclusion and watch it fold. Suggest the opposite and watch it agree with that too.
Executives pick this up faster than any benchmark would predict, because they’ve spent careers around the human version — the advisor with no point of view, the deputy who echoes the last voice in the room. We have a word for that, and we don’t trust it. A system with no resistance reads as a system with no self.
The repair is to build the point of view in. Instruct the system to challenge assumptions. To make the case against. The first time an AI pushes back on a CEO’s plan — usefully, specifically — is the moment I watch the skepticism crack.
I used to file this under empathy. I’ve stopped — because it isn’t a judgment about the tool at all. It’s the person’s own identity under threat, and no amount of trust-building in the system touches it.
It deserves its own answer, and the only honest one runs through the leader: name the tasks the machine takes, name the judgment it doesn’t, and name what people are being freed up to do. If some roles change, say that too. A hard truth builds more trust than a soft evasion — which is itself an authenticity wobble wearing a kind face.
We take it seriously because we had to answer it ourselves. A lot of the consulting work I have done for years could be done with AI now. So who am I as a consultant in an agentic world? My own answer is that I’m an AI coach — an AI-enabled consultant. I wasn’t replaced by this. I changed because of it, and the work got better.
That’s the answer we help people find. Not reassurance — a real one, specific to what they actually do.
There’s a second layer here, and it’s the one I watch leaders miss most often.
During an AI rollout, your people aren’t only deciding whether to trust the tool. They’re deciding whether to trust you. Your judgment in choosing it — that’s logic. Your candor about why — authenticity. Whose interests the change serves — empathy.
Same three legs. And in my experience the leader’s wobble, not the technology’s, is usually the binding constraint.
Which is why this work is coaching and not installation. You can buy a better model. You can’t buy the conversation you haven’t had with your team.
Look back at the three questions this page is built on. Does it make things up. Does it just agree with me. Whose side is it on.
Those are the questions you ask about a person — which is what Frei and Morriss wrote the triangle about in the first place.
So treat it like one. What you have is a new hire with no onboarding. You wouldn’t hand a chief of staff your calendar on day one and walk away. You’d interview them. Set expectations. Give them context. Correct them specifically, more than once. Review how it’s going.
It starts with a discovery call. Bring the thing you keep meaning to fix. We'll work out together whether coaching is the right answer — and if it isn't, we'll tell you what is.