Run workflow automation discovery in five moves: inventory the repetitive work, have each person record their tasks once, prioritise by ROI (frequency × time saved, weighted toward rule-based tasks), export the top ones to your automation platform, and keep a backlog of the rest. The doer captures the workflow — not an analyst — so it scales.
If you've read what workflow automation discovery is, you know the concept: capture the work before you automate it. This guide is the how — a repeatable process a team can run this week.
Step 1 — Inventory the repetitive work
Start with a simple list, not a tool. Ask each person on the team to write down the tasks they repeat in their browser on a regular cadence — daily, weekly, monthly. Don't aim for perfection; aim for the obvious ten. Good prompts:
- What do you do every Monday morning?
- What report do you rebuild the same way each month?
- Where do you copy data from one tool into another?
- What do new hires always ask you to show them?
That last one is a tell: anything teachable as a checklist is a discovery candidate.
Step 2 — Capture each candidate by recording
This is the step that decides whether discovery scales. The old way — an analyst interviews people and writes specs — is slow and lossy. The fast way is to have the person who does the task record it once. They know the real filter, the real edge case, the real order. Recording captures all of it without anyone writing a document.
The doer is the domain expert. Discovery works when capture happens where the knowledge already lives — not in an interview a week later.
Step 3 — Prioritise by ROI
You can't automate everything at once, so rank. Score each candidate on three factors:
| Factor | What to ask | Why it matters |
|---|---|---|
| Frequency | How many times per month does it run? | Frequent tasks compound the savings. |
| Time saved | How long does one run take by hand? | Sets the per-run payoff. |
| Rule-based | Could a checklist fully describe it? | The more rule-based, the cleaner it automates. |
A quick formula: monthly hours saved ≈ frequency × time per run. Rank by that, then bump the most rule-based tasks up — they ship fastest and break least. Start at the top.
Step 4 — Export and ship the top candidates
Take your highest-ROI workflows and export them to the platform that fits each one — a Make scenario, a Zapier Zap, an n8n workflow, a Workato recipe, or a Claude skill for tasks that need judgement. (See how to pick the export format.) Ship two or three, prove the time savings, and you'll have the buy-in to do more.
Step 5 — Keep a living backlog
Discovery isn't one-and-done. Put everything you didn't ship yet into a backlog, ordered by ROI. Re-record a workflow when the underlying tool changes. Add new candidates as people notice them. Over a quarter, the backlog becomes a map of where your team's time actually goes — which is valuable even before it's all automated.
Common mistake: trying to document workflows in a wiki instead of capturing them by recording. Wikis go stale the moment a button moves; a re-recording takes one minute.
What good discovery output looks like
By the end, each candidate should be a clean, editable workflow: a name, the inputs it needs, the ordered steps, and the decision points. Not a video. Not a vague paragraph. Something precise enough that exporting it to an automation tool is a copy-paste rather than a research project. For worked examples by department, see our use-case library.
How Spion runs discovery for you
Spion is a free Chrome extension that turns this whole process into recordings. Each person records their recurring tasks; Spion reconstructs them into editable workflows and exports ready-to-run automations to Claude, Workato, Make, Zapier or n8n. The doer captures, the AI structures, and your backlog builds itself.