Stop building workflows. Start giving objectives.

Look closely at a modern company and it is a collection of integrations. Tools are bought to solve local problems and then wired together, and the wiring quietly becomes the operating model. Nobody designed it; it accreted.

Workflows encode assumptions that expire

Every automated workflow is a frozen belief about how a job should be done. That is fine while the belief holds. It is expensive when the market moves, because the workflow keeps executing confidently against conditions that no longer exist.

The shift worth paying attention to

The change AI introduces is not that steps get faster. It is that a system can be given an objective and a set of constraints rather than a sequence. That moves the design problem from 'what are the steps' to 'what outcome are we accountable for, and how would we know if it went wrong'.

This is a strategy question dressed as a tooling question. Specifying objectives forces a company to say what good actually looks like — which many cannot do, which is why the workflow was comforting.

What I take from this

When I build research and decision tools, I try to make the output a decision, not a document. A brief that ends in a recommendation and a way to check it later is worth more than one that ends in a summary.