Template / Workflow design / Introductory
AI Workflow Discovery Worksheet
A structured way to find out whether a web-enabled AI workflow is worth building, before estimating it. Work through it with the person who does the task manually today.
For Operations leads, technical founders, and automation builders
01The task as it exists
- Who performs it, how often, and how long it takes each time
- The exact steps they follow, including the tabs they open
- The part they describe as tedious, and the part they describe as judgment
- What happens downstream when the output is late or wrong
02Inputs
- Starting point: a URL, a query, a list, a schedule, or an event
- Sources consulted today, named individually where possible
- Information that exists only in someone's head or a private system
- Anything behind authentication or outside permitted use
03Outputs
- The artefact produced: brief, record set, table, alert, or draft
- Where it lands and who reads it next
- The quality bar the reader applies, stated as a check not an adjective
- How often the current output is wrong, and how that is noticed
04Automation split
- Steps that are mechanical and repeat identically — candidates for automation
- Steps that require judgment — candidates for assistance, not replacement
- Steps that carry risk if wrong — keep a human decision point
- The smallest slice that would save real time this month
05Decision
- Estimated time saved per week, from the numbers above rather than a guess
- What could go wrong and who absorbs the cost
- Whether a checklist or template would capture most of the value without software
- Build, assist, or leave it manual — with a one-line reason recorded
workflowdiscoveryautomationplanning