Pick the first workflow: pain × low risk

Choosing your first workflow is a decision, and it goes better when you treat it like one. A brainstorm will give you a long list of ideas. What you need at the end is a single workflow you can test, with clear edges around it. The steps below are written for a business, but they work the same way if you’re choosing a task in your own job or at an early-stage startup.

List the work that hurts

For five working days, write down the tasks that ate up the day, using the words your team already uses for them. Include the work that only happens when someone’s out sick or on vacation. That work is usually undocumented, which makes it tempting to automate quickly and badly.

Then sort your notes into workflows. “Copy the address into the text message” is a step. “Confirm tomorrow’s stops with each customer” is a workflow. You want to work at the workflow level, because a pilot needs to finish something your team would recognize as a complete job.

Ask the people who do the work which item they’d hand off tomorrow if they could trust the draft. Their answer should carry more weight than the owner’s theory. Owners tend to feel the strategic problems, while the people at the desk feel the repetitive ones. Your first pilot should take on repetitive work that the owner also cares about, or it’ll get dropped the first time things get busy.

Rate the risk if it goes wrong

Give each candidate one of these risk labels:

• Internal draft. A person always edits it before anyone else sees it.
• Customer-facing. A mistake would change what a customer believes.
• Money or records. A mistake would change an invoice, a scheduled commitment, or your system of record.
• Safety or trust. A mistake could hurt someone or break a promise you can’t easily take back.

Your first pilot should be an internal draft, or customer-facing only if a named person still reviews and sends it. Leave anything involving money, records, or safety alone until your review process has a track record.

If every idea on your list is high risk, keep narrowing the job until a bad draft would be mildly embarrassing at worst and wouldn’t cost you anything.

Choose one

Pick the workflow with the most weekly pain that still has a low risk label. Write its name at the top of the AI Pilot Checklist, and fill in the pain, the risk, and the system the work happens in today.

Then set the rest of the list aside. Those ideas aren’t rejected forever, only for this round. Say so when you share the decision, so nobody on the team walks away thinking their suggestion didn’t matter.

Say no to “while we’re at it”

“While we’re at it” is how a two-week pilot turns into a six-month platform project. Adding a second workflow splits your examples and the owner’s attention, and it makes the final decision meeting more about opinions than results.

One way to check is to picture the end of the week. Could you lay out both workflows’ output on one table and compare each to the old way? If not, you’ve taken on two pilots, and you should drop one of them.

Once you’ve chosen, the next step is the shadow-mode pilot. If you’re still unsure which job to pick, the full roadmap walks through the whole process, from the first assessment to the decision to keep or stop.

If you’d like someone who builds these systems for a living to sanity-check your choice, book a Free AI evaluation and bring the note with your chosen workflow on it.