The Real Moat Is Knowing What Problem to Solve

The durable advantage in the AI era is not the code alone. It is the judgement to define the right problem and apply AI where it creates real value.

There is a lot of noise right now about AI replacing teams, compressing delivery times and making software easier to build.

Some of that is true. AI does make certain kinds of implementation faster. It can help write code, draft workflows, summarise requirements and accelerate experimentation.

Warren Buffett has often described strong businesses as "economic castles protected by unbreachable moats.", but in the context of AI, speed is not the moat.

The real moat is knowing how to apply that capability to a specific problem inside a real organisation, with real constraints, messy data, competing priorities and imperfect processes.

That kind of judgement is harder to commoditise.

Two teams may have access to the same models, the same tools and roughly the same technical stack. The difference is rarely who can generate the most code. The difference is who understands what actually needs to change, what should stay untouched, and where a small intervention can create disproportionate value.

The ten-minute fix

There is an old consulting joke about the client who complains about being charged a thousand dollars for a fix that took only ten minutes.

The reply goes something like this: ten dollars was for the fix, nine hundred and ninety dollars was for knowing what to fix.

That joke lands because it captures something fundamental about technical work.

The visible act of implementation is only one part of the value. The harder part is diagnosis. It is pattern recognition. It is knowing which symptom matters, which one is distracting noise, and which change will solve the actual issue rather than just moving it somewhere else.

AI does not make that less important. If anything, it makes it more important.

When code becomes cheaper to produce, the cost of solving the wrong problem goes up. You can now build the wrong thing faster, automate the wrong workflow more efficiently, and scale a misunderstanding with impressive speed.

Harness AI, do not outsource judgement to it

Our view is simple: we aim to harness AI, not to let it think for us.

That means using AI as a tool for acceleration, analysis and exploration, while keeping human judgement firmly responsible for framing the problem, testing assumptions and deciding what success actually looks like.

In practice, that often means asking harder questions before building anything:

• Is this really a tooling problem, or a process problem?
• Are we trying to automate a weak workflow instead of redesigning it?
• Is the request itself the best expression of the need?
• What would make this simpler, not just faster?

Those questions matter because AI is very good at producing plausible outputs. It is not automatically good at understanding organisational context, political reality, trade-offs between teams, or the hidden cost of getting a decision wrong.

Advisory that finds the right problem

This is why our advisory work is not just about identifying solutions. It is about identifying the right problem.

That distinction matters. If you define the wrong problem, even an elegant implementation becomes waste. If you define the right problem, the path forward often becomes much clearer, and sometimes much simpler, than expected.

Our aim is to help organisations slow down just enough to see what is really happening, where AI can genuinely help, and where a different kind of redesign is needed instead.

In other words, good advisory work does not just find a problem to solve. It finds the right problem to solve.

For more information about how we can help your organisation improve, reach out via LinkedIn or via our contact form - we’d love to talk.