One of the easiest mistakes to make with AI is to turn every pain point into a new app.
A proposal generator, a sales assistant, a knowledge bot or a scheduling tool might each solve a real problem. But when they touch the same customers, jobs, information and business rules, building them separately can recreate the fragmentation you were trying to remove.
We have been seeing this in a current piece of work. We started with several practical opportunities and built enough to learn where the value might sit, before changing the question from "What should we build next?" to "How should this business work from end to end?"
Don't get me wrong. I think companies should be building, and their ambition should be greater than ever. They can now create systems around their own workflows and judgement in ways that were previously too slow or expensive, provided those systems are part of a wider strategy for how the business should operate.
Cheaper software is still software, though. If a company lacks the resources and disciplines to develop software and manage internal products, it cannot skip building that capability. Otherwise, it ends up with tools that have no clear owner, no place in a roadmap and nobody responsible for support when the business or underlying technology changes.
Quick proofs remain valuable because they replace assumptions with evidence. The important thing is knowing when to zoom out, before those proofs become another collection of disconnected systems.
AI has made building easier, which should encourage greater ambition. It has also made coherent design, product ownership and the discipline to decide what should not be built more important than ever.