Capability Is Cheap Now. Advantage Is Not.

If ten competent teams with the same models, APIs, open source, and cloud budget can reproduce what you just built, you did not build an advantage. You paid the entry fee, and the fee keeps dropping.

I say this as someone who likes the new tooling. Construction, research, analysis, content, prototyping, and back-office work all got cheaper fast, and that is genuinely good. It also stripped the paint off a lot of differentiation that was decorative to begin with.

The question changed from can we to should we

For most of my career, can we build this was a real constraint. Now the answer is usually yes, which makes it a useless filter. The useful ones are why should we build this, and what do we have that makes building it disproportionately valuable for us?

Cheap execution should make a team more selective. In practice it usually makes them busier. AI has made it very easy to manufacture activity, and a roadmap full of things that were easy to build is not a strategy. We can now ship software nobody needs faster than at any point in my career, and plenty of teams are proving it.

Two lanes, and they deserve different people

Commodity execution is anything a competent team reproduces with AI, SaaS, open source, APIs, and standard infrastructure. Auth, CRUD screens, the fifteenth dashboard, the plumbing between two systems. Buy it, compose it, generate it, rent it, and compress the calendar time it eats.

Differentiated execution encodes something other people do not have: proprietary knowledge, an unusual workflow, decision logic earned over years, data rights, customer understanding that took a decade to accumulate. Put your best people there and give them room.

The sorting exercise is uncomfortable the first time. Teams discover that most of the last two quarters lived in the commodity lane, and that the work everyone was proud of was work anybody could have done. That is not a reason to feel bad. It is a reason to move people.

The advantage does not have to live in code

Engineers hear differentiated and reach for a repository. Sometimes the right home for the advantage is a contract, a data-rights clause, a distribution agreement, an operating process, a service wrapped around the software, or a dataset you are allowed to keep. I have watched teams build beautiful custom systems to protect an advantage that a two-page agreement would have protected better and cheaper.

AI amplifies what you already have

Widely available AI is not an advantage. If anybody can call the same model tomorrow, the model is not the thing. It becomes strategically interesting the moment it touches something asymmetric: a proprietary clinical dataset, a trusted distribution relationship, twenty years of judgment nobody has written down, a workflow position that shows you information competitors never see.

So point AI at discovery before automation. Ask it what proprietary information you already hold, which recurring decisions contain unusual judgment, which datasets are piling up unexploited, and which combinations of assets nobody has examined together. That is a better first project than generating another generic application.

What to do Monday

Take the current roadmap and mark each item commodity or differentiated. Move the commodity work toward whatever gets it done fastest with the least attention. Then look at what is left and ask whether it is genuinely hard for a competitor to copy. If nothing survives, you have learned something important, and you learned it before spending the year.