AI Made Us Faster. So Why Is the Organisation Still Slow?

AI is changing how quickly work can be produced. Code that once took days can sometimes be generated in hours. Documents can be drafted in minutes. Analysis can be accelerated. Ideas can be explored almost instantly.

It is tempting to look at this and conclude that work itself is becoming faster. But there is a problem. Making something faster is not the same as making the whole system faster. And the more AI accelerates execution, the easier it becomes to see where the real constraints have been hiding.

The bottleneck did not disappear. It moved.

For years, many organisations have focused on improving execution. How can we develop faster? How can we automate more? How can we increase productivity? AI changes that equation. If producing an output becomes dramatically cheaper and faster, execution may no longer be the limiting factor. The constraint moves somewhere else.

Perhaps a team can create a solution in a few hours - but still waits three days for a decision. Perhaps AI can generate several possible approaches - but nobody is clear about which problem should actually be solved. Perhaps implementation accelerates, but requirements remain ambiguous. Or perhaps teams simply create more work than the organisation can review, integrate, validate, or learn from.

Suddenly, problems that were previously hidden behind the cost of execution become much easier to see. Slow decisions. Unclear ownership. Long approval chains. Dependencies between teams. Poorly defined outcomes. Weak feedback loops.Too much work in progress.

AI did not create these problems. It exposed them.

Faster execution can create faster chaos

This creates an interesting paradox. The same technology that can dramatically increase productivity can also dramatically increase the speed at which we produce the wrong thing. If the problem is poorly understood, AI can help us solve the wrong problem faster. If requirements are unclear, we can produce more output based on unclear requirements. If decisions are weak, we can multiply the consequences of those decisions. If the organisation already struggles with prioritisation, AI can give it even more things to prioritise.

Speed amplifies the system it enters. A healthy system may become faster. A confused system may simply become confused at higher speed. This is why I increasingly think that one of the most important questions about AI is not: How can we use AI to work faster? It is: Is our way of working ready for what happens when execution becomes faster?

Non-deterministic technology needs a deliberate process

There is another characteristic of generative AI that makes this particularly interesting. AI models are non-deterministic. The same or very similar input can produce different outputs. A result may look convincing without necessarily being correct. An AI system can interpret an instruction differently from what we intended.

Trying to remove that characteristic entirely misses the point. The model does not need to become deterministic. The process around it needs to become more deliberate. That distinction matters far beyond software development.

Whether AI is helping us create code, analyse information, prepare a proposal, explore a product idea, or draft a decision, we still need several things around the generated output: Clear intent. Explicit criteria. Verification. Feedback. Human judgement. What are we trying to achieve? What does a good result actually mean? How will we know whether the output meets that expectation? Where do we need independent verification? And where does a human need to make the decision?

AI changes how work is performed. It does not remove the need to answer those questions. In many cases, it makes them more important.

Human control does not mean human approval everywhere

There is a trap here. Once organisations recognise the uncertainty associated with AI, the natural response may be to introduce more control. More approvals. More reviews. More checkpoints. More people involved in decisions. That can make the output safer. It can also create the next bottleneck.

If AI reduces execution time from several days to several hours but the result then spends three days waiting for four people to approve it, very little has changed at system level. So the challenge is not simply to add human control. It is to decide where human judgement actually creates value. Some decisions need explicit human ownership. Some outputs need independent verification. Some risks justify a strong control point. Others do not. The goal should not be to put a human in every possible loop. The goal is to design the right loops.

Look at the system, not individual productivity

This is where I think conversations about AI and productivity can become misleading. We often measure the improvement locally. A developer produces code faster. A Product Owner drafts requirements faster. A manager prepares a report faster. A Scrum Master summarises information faster. Each individual may genuinely save time. But optimising individual activities does not automatically optimise the flow of work.

Imagine that AI reduces the time required for one activity by 70%. What happens immediately afterwards? Does the work continue flowing? Or does it enter a queue? Does somebody need to approve it? Is another team required? Is there enough information to make the next decision? Can the organisation absorb the increased output? That is where the more interesting questions begin. Because once one part of a system becomes dramatically faster, the constraints elsewhere become more important, not less.

AI may reveal organisational debt

We talk frequently about technical debt. Perhaps AI will make another kind of debt much harder to ignore: organisational debt.

The accumulation of structures, habits and decisions that make work unnecessarily difficult. A meeting nobody is quite sure why they attend. A decision that can only be made by one person. A process with seven handoffs.

Three teams required to complete something that could potentially belong to one. Requirements that remain vague until implementation begins. A review that exists because it has always existed. Information trapped in silos.

AI can make execution cheaper without making any of those things cheaper. And when execution becomes very fast, waiting becomes proportionally much more expensive. That changes what optimisation should mean. The next productivity gain may not come from generating another output 20% faster. It may come from removing two days of unnecessary waiting.

Maybe AI readiness is not primarily about AI

This leads me to a different way of thinking about AI transformation. Organisations understandably invest in tools, models, agents and training. All of that matters.But perhaps an organisation's ability to benefit from AI will increasingly depend on capabilities that do not sound particularly technological: Can we define problems clearly? Can we make decisions quickly enough? Do people know who owns those decisions? Can teams work across boundaries? Can we distinguish necessary control from bureaucracy? Can we create short feedback loops? Can we verify outcomes rather than simply produce outputs?Can we change a process when the assumptions behind it no longer make sense?

Those are questions about how organisations work. And that is why AI transformation may turn out to be as much an organisational design challenge as a technological one.

The interesting question comes after “faster”

AI can make us faster. That part is becoming increasingly easy to demonstrate. The more interesting question is: What happens next? If one part of the system becomes ten times faster, what becomes the new constraint? Where does work start waiting? Which decisions suddenly matter more?

Which processes were designed around limitations that no longer exist? Which controls still protect us - and which simply slow us down? And which organisational problems become visible only because execution is no longer hiding them?

Perhaps that is one of the most useful things AI can give organisations. Not only more speed. A much clearer view of what was slowing us down all along.