AI Doesn't Have a Trust Problem. We Do.

“The future is already here — it's just not evenly distributed.” — William Gibson

A few weeks ago, I wrote about something that has been bothering a lot of UX people.

'Will AI replace us?'

Someone on my team had asked me the same thing.

It was a fair question. My answer then was simple: 'AI isn't going to replace good UX. But it will expose bad leadership.'

Because when people are frightened, they don't need predictions.

They need leadership.

I still believe that.

But something has changed since I wrote that article. We're not just talking about AI anymore.

We're using it. Every day. And that has led me to a different question.

Can we trust it?

We start with fear, now we're judging.

India is moving quickly. Microsoft's 2026 Work Trend Index found that 32% of India's AI users qualify as "Frontier Professionals": People actively redesigning how work gets done around AI agents. That's twice the global average.

But here's the part I found more interesting: Indian AI users ranked quality control of AI output as their highest-priority skill, at 63%. Critical thinking came next, at 59%.

Think about that for a moment.

The people using AI most aggressively are also saying:

"I still need to check what it gives me."

That's not resistance. That's maturity.

How's this related to UX?!

I've noticed something similar in UX. Give a designer five minutes with an AI tool and you can get ten ideas. Give them another five minutes and you can get twenty.

The problem is no longer:

"Can we produce something?"

We can.

The problem becomes:

"Should we?"

That's a very different question.

A generated user flow can look perfectly reasonable. A research summary can sound convincing. A piece of copy can read beautifully. A strategy document can contain all the right words.

And still be wrong.

Not obviously wrong.

That's the dangerous part.

Plausibly wrong.

One of the things The Design of Everyday Things taught me is that people don't experience systems as collections of features.

They experience them as interactions.

Cause.

Effect.

Feedback.

Expectation.

Recovery.

AI is no different.

If an AI gives me an answer, I need to understand:

Why did it give me this?

What information did it use?

How confident should I be?

What happens if it's wrong?

Can I correct it?

Can I recover?

These aren't just AI questions.

They're UX questions.

And they're leadership questions.

Because someone has to decide what level of trust the product deserves.

A recent Salesforce study found that 38% of Indian workers said their organisation had experienced an unsuccessful AI pilot (and I am not talking about Air India pilots) in the previous year.

The most common reason?

The AI lacked business context.

Other complaints included generic outputs, poor personalisation and insufficient training.

That makes perfect sense to me.

Imagine hiring an incredibly intelligent consultant.

Then refusing to tell them anything about your customers.

Your business.

Your history.

Your constraints.

Your team.

Your goals.

Then becoming disappointed when their advice sounds generic.

We wouldn't call the consultant unintelligent. We'd say: "You don't have enough context."

AI is often dealing with the same problem.

I've spent years managing people with very different strengths.

One person needs autonomy.

Another needs more structure.

One wants the problem and freedom to solve it.

Another wants to talk through the options.

One thrives on ambiguity.

Another does better when the destination is clear.

The mistake is assuming that because everyone has the same job title, they need the same kind of support. AI is no different.

There is no universal "AI strategy." The right question isn't: "How do we get everyone using AI?"

It is: "Where can AI make this person better at what they already do?"

That is individualization.

And I think it is going to become one of the most important management skills of the AI era.

The evolution of a manager

This is probably the biggest change I've felt personally.

Earlier in my career, I spent a lot of time improving the work - Reviewing, editing, critiquing, fixing.

As I became a better manager, I realised something. I didn't want to become the final quality-control checkpoint for everything my team produced.

What I really wanted was to help them develop better judgment. AI makes that distinction even more important.

If a team can generate ten solutions in seconds, I don't need to help them generate an eleventh.

I need to help them decide which of the 10 deserves attention.

That means asking better questions.

What problem are we solving?

What evidence do we have?

What are we assuming?

What could make this wrong?

What would we need to learn before acting?

Those questions haven't become less important because of AI. Matter of factly, they've become more important.

Upskilling is not a popular thing to tell a manager. Prompting matters. Tool fluency matters.

Knowing how to work with agents matters.

But I don't think they are the long-term differentiators. The differentiator is judgment. Knowing when to accept an answer. When to challenge it. When to ask for more evidence. When to throw it away. When to involve another person.

And when the problem itself needs to be reframed. AI can give you an answer. The only problem is it just can't decide how much the answer deserves to be trusted.

At least, not without humans deciding what trust means first.

And this brings me back to leadership

When I wrote the earlier article, I said that the future of UX wouldn't be decided by AI.

It would be decided by how leaders help people navigate it. I still believe that.

But now I'd add something - Leadership isn't just helping people navigate the uncertain.

It's helping them develop judgment inside it. That's a much harder job.

And a much more valuable one. Because eventually, the technology will become better, faster, more autonomous, more convincing, and more human-like.

And when that happens, the ability to say "I'm not sure we should trust this yet" may become one of the most important skills on your team.

AI should be closer to a very intelligent co-driver. It can and should spot things you missed while you were focussed on the road straight ahead. It should suggest routes, calculate options, warn you about obstacles. But you still need someone in the driver's seat. Someone who knows why the journey matters. That's never going to get replaced by AI.

But I digress.

To acquire that ability, we need people who will experiment and question the output. We should be able to talk openly about failures. Share examples of where AI helped. Share examples of where it didn't.

And most importantly: Reward good judgment, not just AI adoption.

Because the person who says,"AI gave me this answer, but I don't think it's right" may be demonstrating more AI maturity than the person who produced fifty AI-generated deliverables before lunch.

The first phase of the AI conversation was about fear. The next phase is about trust. In our own abilities and boundaries of capabilities.

And I don't think trust will come from making AI sound more human. It will come from making humans better at knowing when to trust it.

Today, AI can make us faster. It can make us more productive. It can give us more options than we have ever had. But options aren't judgment. Output isn't insight. And speed isn't wisdom.

The real advantage won't belong to the teams that use the most AI.

It will belong to the teams that know what to do with what AI gives them.

That's a leadership problem.

And I think it's going to be one of the most interesting ones we'll have to solve this decade.