Process inventory: finding the candidates
Bringing AI into your business · Lesson 2 / 24
Where the list comes from
You can't invent the list in an office. The people doing the work by hand know more about the routine than any manager — and usually nobody has asked them.
Three sources of candidates
- Asking staff. One question: "what do you repeat every day, and which part of it feels pointless". The answers come back specific.
- System data. How many support tickets, how many documents pass through the department, how long a request sits. Numbers reveal volume nobody feels.
- Waiting points. Where work sits in a queue: a customer waiting for a reply, a document waiting for approval, a report waiting to be assembled. Queues are almost always expensive.
Describing a task so it can be assessed
Task: preparing a reply to an incoming customer request Who does it: support manager, 4 people How often: ~600 requests per week Time per unit: 6-8 minutes Input: request text + customer history in the CRM Output: reply text, sent to the customer What counts as a good result: the customer's actual question answered, no errors in the terms, tone matches ours Cost of an error: medium — wrong terms have to be retracted
A description in this form takes five minutes and removes the main risk: discussing a task that everyone understands differently.
What you'll almost certainly find
- Some of the routine exists because the process is broken, not because of volume. That should be removed, not automated.
- Some tasks are already being handled through personal AI accounts — unmanaged and without your knowledge.
- Real volumes differ from management's perception by multiples, in both directions.
Insight. Half the routine you find is a symptom of a bad process, not of volume. Such a task is cheaper to remove than to automate.
Common mistake. Collecting the list only from managers. They know the processes but not what actually consumes hours for the people doing the work.
Pro tip. Ask directly who already uses AI and for what. It usually emerges that use has existed for a long time — just in the shadows, on personal accounts and without rules.
Cheat sheet
- Three sources: staff, system data, waiting points.
- Describe by template: who, volume, input, output, criterion, cost of error.
- A broken process is fixed by process, not AI.
- Shadow use already exists — measure it.
1. Who knows most about routine in a company?
2. What do you do with routine caused by a broken process?
3. Why describe tasks by template?
Task — checked by AI
Describe three tasks from your list using the lesson's template: who does it, how often, time per unit, input, output, criterion for a good result, cost of an error. Use real numbers — from systems or from asking people, not from impressions. Separately state whether you found tasks where the problem is the process rather than the volume, and whether shadow AI use exists in your company.