Designing AI That People Actually Trust

I've been designing AI before it had a name — not adjacent to it, inside it.

Most companies are just dropping AI on people right now — here's a new tool, here's a new workflow, figure it out. And these are people who already have full plates. That's not a rollout problem. It's a design problem.

At ServiceNow, I own OTTO, the centralized hub for Strategic Portfolio Management, replacing the disconnected spreadsheets and email chains enterprise teams used to plan, collaborate, and execute around. I built Collaborative Work Management from the ground up as designer #1 — it became the fastest adopted product in Strategic Portfolio Management history. That work is in production across 100,000+ users.

The real design work in AI isn't the model. It's the surface: how an agent surfaces itself, how it behaves, what it says and doesn't say, and how it earns trust from a user who's already skeptical of every product that just rebranded itself as "AI now." That has to be embedded in the work itself, not painted over an existing workflow as a feature.

The gap most teams miss is between something that feels powerful in a demo and something that's exhausting to use in real life. Closing that gap — and shipping something people actually want to use — is what design leadership is actually for.

A real conversation on AI design

What's wrong with how most companies deploy AI right now?

Honestly? They're just... dropping it on people. Like, here's a new tool, here's a new workflow, figure it out. And these are people who already have full plates. Now they've got one more thing to learn on top of everything else.

So it's a rollout problem?

It's deeper than that. It's a design problem. Nobody stopped to ask — does this actually make someone's day easier? Or are we just checking a box that says we're "doing AI now."

But the technology is genuinely powerful though.

That's the frustrating part — it IS. That's what kills me. The potential is real. Like, imagine actually offloading the stuff that drains you. The repetitive tasks, the back and forth, the stuff that never needed a human in the first place. That's exciting. That's the version I want to build.

So what does good look like?

It feels like backup. Like you showed up to work and someone already handled the tedious stuff so you can get to what actually matters. The judgment calls. The relationships. The creative problems that need a real person in the room.

And that's not what's happening.

Not even close. Most AI right now just adds noise. More dashboards, more prompts, more things to manage. That's not relief — that's just a different kind of burden.

What's the fix?

Design it like you give a damn about the person using it. Start there. Everything else follows.

I prototype these systems in code, because you can't design agent behavior in a static file — the handoff to engineering has to start from proven ground, not a mockup someone has to reinterpret. (More on that in Why I Prototype in Code.)

If you're building a team around this and want someone who's already lived the hard version of this problem, I'd love to talk.