Choose your first AI workflow

Start with a repeated task, a clear owner and a way to compare the result.

Choose a task you can describe.

Start with a request that appears often enough to study. For a service business, this might be an inquiry about a service, a follow-up after a proposal, or a request that needs missing details. Write down the trigger, the information someone looks for, and the next action.

A broad goal such as “automate customer service” is difficult to test. “Prepare a draft for this kind of inquiry, using these approved sources, for this person to review” is a useful starting point.

Define the boundary.

Separate preparing information from making commitments. A draft can ask for a location or summarize a request. Pricing, availability, refunds and unusual situations may need a person. Decide where the workflow stops before you build it.

Gather representative examples.

Include routine requests, missing information, unclear wording and cases that should be escalated. Remove identifying details before sharing examples. Keep a separate set for evaluation so that the prototype is tested beyond the examples used to design it.

Agree on what a good result means.

• Does it correctly identify the request?
• Does it use approved information?
• Does it avoid promises the team has not approved?
• Does it make exceptions visible?
• How much correction does the reviewer need to make?

Record the current process as a baseline. A polished demo is not enough to establish that a workflow helps your team. Compare draft quality, handling time and exceptions under conditions that match your work.

Assign an owner before expanding.

Someone needs to maintain the source information, review failures and decide when the workflow should change. Agree on that responsibility while the scope is still small.

Find a starting point for your team or explore a scoped AI pilot.

Use this guide to prepare your team's discussion. A project's scope depends on your sources, systems, and review requirements.