Skip to main content
Field notes

Choose the first workflow by the work it leaves behind.

A practical way to turn a broad AI ambition into a useful first implementation.

2026-10-06 · 3 min read · Onyx

The best starting point is a recurring piece of work your team can describe clearly. A crowded inbox, a document lookup, a routine follow-up or a handoff that needs another round of chasing.

Begin with one real example.

Walk through the last time the task happened. What arrived? Which information did someone need? Which tools did they open? What did the next person need in order to continue?

That example gives the implementation a starting point. It also reveals the exceptions that a generic feature list can miss.

Define the useful result.

Describe the output a person would actually use: a prepared reply, a qualified record, an approval packet or an appointment with the right context attached.

Agree how the team will review that output before deciding how much of the workflow should run automatically.

Map the inputs and the handoff.

A workflow needs authorized information and a clear destination. Identify the system that holds the context, the person who owns exceptions and the next step after the work is prepared.

Give the first step a clear review.

Set success criteria and an implementation timeline before the work starts. Review real examples together, then use what you learn to prioritize the next part of the roadmap.

The first workflow is a starting point for a connected operating rhythm across the business.

What keeps your team busy?

Tell us about the repeated task, the unanswered calls or the inbox that never quite clears. We’ll help you find a useful place to start.