How to map your workflows and find what AI can automate

You cannot automate what you cannot see. Workflow mapping is the step most companies skip, and the main reason their AI tools end up as expensive search boxes.

Run a two-hour mapping session

Bring together the people who do the work every day, not just their managers. Pick one process and walk through a real recent example from trigger to finish. On a whiteboard, record each step with four details: who does it, what system or document they use, how long it takes, and what happens next.

Look for five patterns

AI and automation tools handle these well:

• Reading: pulling data out of emails, PDFs, invoices or forms.
• Sorting: routing enquiries, classifying documents, tagging tickets.
• Drafting: first versions of replies, reports, proposals and summaries.
• Checking: comparing documents, spotting missing fields, flagging exceptions.
• Moving: copying information from one system to another.

Steps that need judgement, negotiation or relationships should stay with people, with AI preparing the information they need.

Mark the waiting time

Often the biggest delay is not the work but the wait: approvals sitting in inboxes, information requested and not received. Automatic reminders and routing can cut turnaround time more than any clever model.

Turn the map into a backlog

For each automatable step, note the hours it would save and how hard it is to build. Sort by savings divided by effort. The top three items are your next quarter's implementation plan.

Who should run it

Mapping works best when led by someone who knows both the business and the tools. An implementation lead can usually map and prioritise two or three processes in the first month of an engagement.