AI's Cultural Debt Is Compounding. Here Is What to Do About It

A lot of leaders still talk about AI as if it sits outside culture.

It does not.

AI is changing culture whether organisations are intentional about it or not.

It's changing what people think counts as effort. It's changing what ownership feels like. And it's changing whether people trust each other's work, whether they speak up when something feels off, and whether teams still feel like teams or just a collection of people checking machine output.

That is why Deloitte's idea of cultural debt matters so much. Cultural debt is the negative consequence that builds up when an organisation neglects its culture while change is happening all around it. In the AI era, that debt compounds fast, because AI is not just changing tasks. It is changing human-to-human working relationships.

Deloitte's 2026 Global Human Capital Trends report makes that plain: 42% of workers say their organisation rarely evaluates AI's impact on people. That is not a small oversight. It is an indicator that debt is already building.

AI is not just changing work. It's changing the meaning of work

This is the part too many AI conversations miss.

Most AI planning still focuses on tools, use cases, workflows, and productivity gains. Much less attention goes to the social questions people start asking the moment AI becomes part of daily work.

Is it cheating if I use it for this?

What counts as my work now?

If the output is wrong, who owns that?

If someone is using AI heavily and producing more, are they genuinely performing better, or just getting better at looking productive?

Deloitte's report captures that tension clearly.

That one figure tells you culture is already under strain. Once people start questioning whether output reflects effort, judgment, or real contribution, trust starts to thin out.

This is why cultural debt is not a soft issue sitting off to the side of AI strategy. It is a performance issue. If people stop trusting what good work looks like, stop feeling sure about who owns what, or start privately rewriting the norms for themselves, then every AI initiative starts carrying hidden drag.

The warning signs are already here

Deloitte found that 34% of organisations now recognise culture as a direct inhibitor to their AI transformation goals. It also found that while more than half of respondents think AI's impact on culture is important, only 5% say they are making great progress in addressing it.

That is a familiar pattern.

Leaders can feel that something is shifting, but they're still treating culture as something that will settle itself once the tools are in place. It will not.

Culture never sits still during change. It adapts. The question is whether it adapts in a way that strengthens trust and clarity, or in a way that quietly erodes both.

Gallup's 2025 data makes this even more sobering. Only 20% of US workers say they feel strongly connected to their company's culture. Deloitte pairs that with declining employer trust and makes the point directly: culture is already showing strain. AI is not arriving into strong, settled, highly connected workplaces. In many cases, it is arriving into environments where connection is already fragile.

AI does not land on a neutral surface. It lands inside an existing culture. If that culture is already weak, ambiguous, or low on connection, AI often amplifies the problem.

This is what cultural debt looks like in practice

It rarely starts with a dramatic failure.

It starts quietly.

A team stops discussing how they are using AI because it feels awkward.

Managers avoid setting clear norms because they do not want to sound restrictive.

Some people use AI heavily. Others avoid it. Nobody says much.

AI-generated work starts appearing in meetings, documents, and decisions, but no one is clear on what was machine-generated, what was human-checked, or what level of scrutiny is expected.

People begin making private judgments. About fairness. About effort. About quality. About who is pulling their weight.

That is cultural debt.

It is the accumulation of unresolved questions and unspoken norms. And once it builds, it starts to tax everything else. Trust gets thinner. Connection gets weaker. Team debate gets shallower. More time goes into second-guessing and less into real problem-solving.

The reason we think this matters so much is that many organisations still do not count any of this as part of AI implementation. They count adoption. They count output. They count time saved. They do not count what is happening to the human system around the work.

Trust ambiguity explains why this gets so corrosive

In a Harvard Business Review piece co-written by Amy Edmondson and Jayshree Seth, the authors describe this as the gap between believing trust should be warranted and not actually feeling that trust, with that lack of trust becoming hard to discuss.

In AI environments, that happens when the system produces something polished and confident, but people are not genuinely sure whether it should be trusted. They feel they're supposed to trust it. They don't fully trust it. And they often don't know how to talk about that gap.

That is where teams start to wobble.

In Edmondson's framing, people lose confidence not only in the AI output but in their own judgment. They start wondering whether they are being too sceptical, too trusting, too slow, or too old-fashioned. Because AI errors don't work like human errors, teams often cannot metabolise them through normal discussion. They cannot easily ask the machine what it was thinking. They cannot build shared understanding in the usual way. So doubt lingers and spreads.

That is cultural debt in one of its most dangerous forms. The team loses a clear, discussable pathway for making sense of mistakes. And once that happens, trust does not just drop between people and the tool. It starts dropping between people as well.

So what should leaders do about it?

The first step is to stop pretending culture is separate from AI rollout.

If AI is changing effort expectations, ownership, fairness, accountability, and how teams make sense of work, then culture is already part of the implementation whether you have planned for it or not.

That means leaders need to do three things.

First, make the hidden questions discussable. If people are already wondering what counts as good work, when AI use is acceptable, how transparent they need to be, and who owns an AI-assisted decision, those questions need to come out into the open. Unclear norms do not stay neutral. They turn into private workarounds.

Second, treat trust as something that needs active design. If 80% of people are concerned AI is being used to fake productivity, you do not solve that with a vague message about responsible use. You solve it with clearer expectations, better manager conversations, and explicit standards around ownership, review, and contribution.

Third, measure the human impact alongside the business impact. If 42% of workers say their organisation rarely evaluates AI's impact on people, that is the gap to close. You need to know whether AI is improving judgment, strengthening collaboration, and preserving trust, not just whether output has gone up.

Why a survey will not find this

That third one is where most organisations get stuck, and it is worth being honest about why.

Cultural debt is made of things people do not say. Unspoken norms. Private judgments about fairness and effort. Questions that felt too awkward to ask out loud. You cannot survey your way to that, because a survey asks people to report on exactly the things they have already decided not to discuss.

So you have to watch what people do instead.

That is what the AI Readiness Experiment is built for. It puts people on a real work problem for ten days, using AI, under real constraints — and captures what actually happens. Where AI creates genuine value. Where it creates rework and checking burden. Where people struggle to verify an output. Where human judgment quietly overrides the machine, and where it quietly stops trying.

It also produces a second layer that surveys rarely reach. The problems people choose to work on are themselves data. Aggregated across a cohort, those choices reveal recurring patterns in workload, role clarity, support, change management and decision-making — the conditions underneath the culture, surfaced from what people did rather than what they reported.

Don't ask only whether your AI rollout is progressing.

Ask what it's doing to trust.

Ask what it's doing to norms.

Ask what it's doing to connection.

Culture is changing either way. The only real choice is whether you are shaping that change on purpose or paying for it later with interest.

If you want to see where cultural debt is already building under the surface, the AI Readiness Experiment is where to start.