One-page process audit
Before you evaluate a single vendor, pick your highest-pain workflow and map it honestly. This worksheet tells you whether a process is a candidate for AI — whether it is defined enough and the data clean enough that automating it would scale a good thing rather than a bad one. It also gives you the definition of "right" you will need later to know whether the AI is actually working.
Part 1 — Identify the workflow
| Field | Your answer |
|---|---|
| Workflow name | ______ |
| Where does it start? (trigger event) | ______ |
| Where does it end? (definition of done) | ______ |
| Who owns it? | ______ |
| How many times per week does it run? | ______ |
| What does "done right" look like — in a number? | ______ |
Part 2 — Map the actual steps
Do not describe how the process is supposed to run. Watch someone do it, or read the system logs. Record what actually happens, including loops, hand-offs, and exceptions.
| Step # | What happens | Who does it | Time (min) | Cost per run ($) | Error or rework rate | AI could help here? Y/N |
|---|---|---|---|---|---|---|
| 1 | ||||||
| 2 | ||||||
| 3 | ||||||
| 4 | ||||||
| 5 | ||||||
| 6 | ||||||
| 7 | ||||||
| 8 |
Part 3 — Tally the waste
| Metric | Current state | What "good" would look like |
|---|---|---|
| Total time per run (sum of Step column) | ______ min | ______ min |
| Total cost per run | $______ | $______ |
| Error / rework rate | ______% | ______% |
| Volume per week × cost per run = weekly cost | $______ | $______ |
| Where does it wait longest? (bottleneck step #) | Step ______ | —— |
| Where do exceptions go? (step # and who handles them) | ______ | —— |
Part 4 — Data readiness check
AI runs on the data this process produces and consumes. Answer honestly.
- ☐ The data is clean — records agree with themselves across systems
- ☐ The data is accessible — it can be read by a tool without a six-figure integration project
- ☐ The data agrees with itself — the same entity is named consistently across systems
- ☐ The data is current — it reflects reality, not a snapshot from last month
Number of boxes checked: ______ / 4
If fewer than 3 boxes are checked, data cleanup is the real first project. The AI is the easy last mile. The data pipeline is the road.
Part 5 — AI-readiness verdict
| Question | Answer |
|---|---|
| Is the process documented well enough that you could write down what "right" looks like? | Y / N |
| Are the steps identified as AI candidates (Y above) lookup, comparison, drafting, formatting, extraction, classification, or routing — not judgment? | Y / N |
| Is the volume high enough that small per-task savings add up to real money? | Y / N |
| Do you have a baseline metric you'd use to measure improvement? | Y / N |
| Is data readiness ≥ 3/4? | Y / N |
If all five are Y: This is an AI candidate. Define a 90-day pilot metric and proceed to vendor evaluation.
If any are N: Resolve the gap first. Automating a broken or data-poor process scales the problem, not the solution.
My next action
Gap to resolve first (if any): ______
Pilot metric: ______
90-day target: ______
Owner: ______
Want a second set of eyes on this in your firm? The no-sell promise applies — if it isn't a fit, I'll tell you in the first ten minutes.
Book a 30-Minute Call →