Skip to content
FQ AI Summit 2026

Check the Dependencies Before Running a Workflow in Parallel

My notes from attending the FQ AI Summit in New York, November 10, 2026.

Parallel AI Workflows and Human Overrides

Which parts of a workflow can happen at the same time?

Before you run a workflow in parallel, map what each task needs before it can start. Some work depends on an earlier decision, and other work waits only because the process has always passed one document from team to team.

A Power Shift speaker described moving away from a long sequence of product-development steps passed from one team to the next. That was a proposed organizational approach, and it's no evidence that every step in every process can safely run at once.

Two reviewers, for example, may be able to assess the same approved brief independently. Preparing final materials before the core requirement is decided can cost you more time than it saves. Some waits are just habit.

Someone has to put the pieces together

Give everyone a shared input and a clear version, so they know which facts are settled and which questions are open. When a decision changes, identify the outputs that need another look. Pick one person to reconcile the results, because two reasonable outputs can conflict once they're combined.

Test the design on a limited workflow and count rework along with elapsed time. You're after less unnecessary waiting while keeping the dependencies that make the final outcome correct.

What to try this week

List the steps in one workflow and mark which ones truly need an earlier decision before they can begin.