What an Audit Is For
An AI workflow audit turns vague ideas into a short list of projects you can actually build. Instead of asking where AI might be impressive, it asks where your business repeatedly receives an input, applies known rules, produces an output, and loses time or money when that is slow.
A good audit produces two lists: what to automate, and what not to.
Step 1: Interview the People Who Do the Work
Managers describe processes as they should work. The people doing the work know how they actually work. Ask:
- What do you do most often in a typical day?
- What do you type into more than one system?
- Where do you wait on someone or something?
- Which questions from customers do you answer over and over?
- What goes wrong, and how do you fix it?
Step 2: Map Each Workflow
For each candidate, write down:
| Element | Example: new lead workflow |
|---|---|
| Trigger | Web form, phone call, or ad lead |
| Inputs | Name, phone, need, location |
| Systems | Website form, CRM, calendar, phone system |
| Decisions | Is the lead in our service area? Is it urgent? |
| Actions | Call back, qualify, book, assign to a salesperson |
| Waits | Lead sits in inbox until someone checks |
| Exceptions | Existing customers, duplicates, spam |
| Output | Booked appointment and CRM record |
Mapping often shows that the real problem is a missing integration or a slow handoff, not something that needs AI at all.
Step 3: Measure Volume and Cost
For each workflow, estimate:
- How often it happens per week or month
- Time spent each time
- Who does it and their hourly cost
- Revenue at risk when it is slow or missed
- Error and rework rate
Step 4: Score the Opportunities
Score each from 1 to 5:
- Value: revenue protected or time saved
- Volume: how often it happens
- Data readiness: whether the needed information is in systems you can reach
- Rule clarity: whether the steps can be written down
- Risk: how safe it is if something goes wrong (higher is safer)
- Oversight: how easily a person can review or correct it
Rank by total score. Our what to automate first guide includes a worked example.
Step 5: Decide What Each Problem Needs
Not every problem needs AI:
- A simple rule or integration when the logic is fixed, such as routing by ZIP code
- AI when inputs are messy or conversational, such as phone calls, emails, or documents
- A process change when the workflow is unclear or inconsistent
- Nothing yet when the volume is too low to justify the effort
Step 6: Build the No-Go List
Mark workflows that should not be automated now, and say why: too risky, too rare, unclear rules, regulated decisions, or missing data. This list prevents expensive detours later.
Step 7: Write a One-Page Brief for the Top Opportunity
- The problem and its cost
- The trigger, steps, and systems
- What the AI may and may not do
- When a person takes over
- The owner
- The success metric and baseline
This brief is what an internal team or an AI automation agency needs to give you an accurate estimate.
How to Measure Results After Launch
- Cycle time before and after
- Exceptions and their causes
- Cost or revenue impact
- Readiness of the next workflow on your list
Bottom Line
An audit takes a few days and saves months of building the wrong thing. Interview the team, map the work, score the opportunities, and write a clear brief for the winner. Our AI automation readiness audit gives you a template, or we can run it with you on a free discovery call.