What Happens When Your AI Starts Seeing What Managers Miss?

"You finally get the visibility you wanted. Don't damage the openness that made it possible."

I was talking with a Managing Director recently who raised a question I think more leaders are going to face as their AI systems mature.

They were starting to get real value from AI. The information was getting richer, partly because people had grown comfortable with it capturing their meetings, actions and decisions honestly. Useful patterns were beginning to appear across the business, including some uncomfortable ones.

The question that emerged? How do you act on what the AI is showing you without destroying the openness that made it possible?

AI producing useful insights is no longer the interesting part. That's already happening. The harder question is what you do once it starts telling you things you cannot ignore.

This is what happens when the system starts working

As information accumulates, patterns start appearing across meetings, decisions and actions that were previously hard for one person to see.

Some are positive. A team consistently closes what it commits to. A manager keeps resolving problems before they escalate.

Then the uncomfortable ones start appearing. A particular type of action repeatedly goes nowhere. One part of the business keeps missing commitments. The same problem surfaces in meeting after meeting. Perhaps an individual appears to be struggling in an area that had not previously been obvious.

Now you have a very different AI problem. You finally have better visibility. What do you do with it?

Insight is not evidence

If AI identifies a pattern, that should give you a reason to look more closely. It should not automatically give you a conclusion.

Suppose a team appears to be repeatedly missing commitments. Maybe there is a performance problem. Maybe there is a workload problem. Maybe another department is creating a bottleneck. Maybe priorities keep changing. Or maybe the system is only seeing part of the picture.

AI can surface the pattern. It cannot automatically know every piece of context around it.

So the first question should not be: “Who is the problem?” It should be: “What is this telling us to look at?”

Use AI to surface the question. Use people to determine the answer.

Do not poison the source

This is where trust matters.

If people have spent months getting comfortable with AI capturing meetings, actions and decisions, then leadership needs to be careful about what happens when the first uncomfortable insight appears.

Turn the system into a blunt performance measure and behaviour will change. People become more guarded. They question what gets recorded. Some will find ways around it.

You may keep the technology running, but the quality of the information underneath it starts to deteriorate.

That is the risk.

You finally get the visibility you wanted, then damage the openness that made it possible.

 

But this is where the bigger opportunity starts

What if the same AI that helped identify the problem could also help improve it?

Businesses have always been heavily dependent on the quality of their managers. A good manager notices when someone is struggling. They work out what is really going on. They give specific feedback. They coach. They follow up. They help the person improve.

Anyone who has run a business knows how inconsistent that is in practice.

Managers are busy. Some are very good coaches. Others are not. Performance conversations get delayed. Feedback becomes vague. “You need to be more proactive.”

“Communication needs to improve.” “You need to take more ownership.”

None of that tells somebody what to do differently on Monday morning.

AI may change that.

From identifying the gap to helping close it

Once a pattern has been identified and a human has understood the context, AI can help the manager become much more specific.

What appears to be happening? Where does it show up? Is it consistent or situational? What does good look like in this role? What skill or behaviour might need to improve? How could somebody practise that skill? What could we review in two or three weeks to see whether things are changing?

That is already a better performance conversation.

Then AI can potentially support the person as well. Preparing for an important meeting.

Practising a difficult conversation. Reviewing whether actions were properly closed.

Helping structure their thinking before a decision.

Giving them a way to rehearse, reflect and improve without needing their manager beside them every time.

The manager still owns the judgement, the conversation and the relationship.

But AI may give an average manager access to some of the scaffolding that previously only very strong managers provided consistently.

Look at the system first

Before pointing at the individual, look at the system. If actions keep getting missed, is the action process any good? If one team is struggling, what is arriving into that team?

If the same issue keeps appearing across several meetings, is ownership clear?

If one person seems to be underperforming, are expectations clear enough that they would know what good looks like?

Sometimes the problem will be the individual. Leadership still has to make those calls.

But better visibility should help us understand the problem properly before deciding where the fault sits. Otherwise we have built a very sophisticated way of jumping to conclusions faster.

Knowledge is power. How you use it matters.

Over the last couple of years, we have consistently said organisations need more transparency around AI.

That is still true.

You cannot improve what you cannot see.

But if people are being asked to work openly inside systems that can increasingly identify patterns across their behaviour, leadership needs to be equally clear about how those insights will be used.

The aim should be improvement, better systems, better coaching, and where genuine performance problems exist - fairer information to deal with them.

The real promise isn't a better dashboard. It's the chance to take what our best managers already do naturally, noticing patterns, giving specific feedback, following through, and make more of that available across the business.

The first phase of AI helps you see more. The next phase is learning what to do with what you see.

And if trust got you this far, do not blow it when the system finally starts telling you something useful.

The real value is not in finding the problem faster. It is in helping people solve it better.

Find out if your business is ready to act on what AI shows you

Most businesses can tell you they want more visibility. Few have thought through what happens to trust the moment AI starts surfacing the uncomfortable patterns.

The 5 Minute AI Readiness Snapshot helps you see where your business stands, before that first uncomfortable insight puts it to the test.

Take the 5-Minute AI Readiness Snapshot

--

About the Author 

Mark Lucas is Co-founder and Managing Partner at Generation AI, where we help New Zealand business leaders build real AI capability.

Information provided is general in nature and does not constitute legal advice