Quick answer: Audit the work before the software. In half an hour, you can usually identify which repeated task has clean inputs, a clear owner and tolerable failure costs.
Minutes 0–10: follow the actual work
Ask a staff member to walk through yesterday’s work rather than describing an ideal process. Write down what triggered each task, where information arrived, which systems were opened, where decisions were made and where the work stalled. Count handoffs; they are often more significant than typing time.
For a Bristol-based cleaning service, the visible task might be “answer enquiries.” The underlying work is retrieve postcode → check coverage → read service type → consult rates → ask for missing details → propose a callback. Automating the whole process is much harder than automatically identifying enquiries that need a coverage check.
Minutes 10–20: score opportunities
Give each candidate a 1–5 score for repetition, input quality, reversibility and clarity of success. Subtract a risk penalty for financial impact, sensitive data, misleading output and legal consequence. Keep the scoring worksheet next to the decision; don’t convert it into a mysterious AI opportunity score.
| Process | Repeated? | Can reverse? | Suggested first step |
|---|---|---|---|
| Categorize incoming emails | Often | Yes | Classify and label |
| Change invoice payment details | Sometimes | No | Human-only |
| Draft meeting follow-up | Often | Yes, before sending | Draft and review |
| Decide employee eligibility | Rare/variable | High impact | Expert-led manual process |
Minutes 20–30: define the pilot
Write a one-sentence outcome: “Cut manual sorting time on support messages without increasing incorrectly routed cases.” Establish a baseline from recent data; do not invent hours saved. Select a test owner, a stop condition and a date for reviewing the pilot.
Copyable worksheet
- Trigger and volume per week:
- Current steps and systems:
- Manual time for a typical case:
- Known exception types:
- Personal or confidential fields involved:
- Worst plausible mistake:
- Reviewer who can approve or reject:
- Measurement plan and rollback owner:
Why the audit matters for search-fueled hype
Trending terms like “agents” can make every operation look ripe for automation. Your objective is not to deploy an agent. It is to remove an annoying bottleneck without introducing a new source of mistakes. The audit prevents buyers from spending their budget on the wrong task.
References and further reading
These references provide context and official guidance. The sample workflows above are illustrative designs, not vendor performance tests.
Published by: FlowHarbor Editorial Desk. Read our sourcing and review policy.