How to Identify the Best Processes to Automate: A Business Automation Checklist
The first automation you ship decides whether there's a second one. Choose a politically visible, high-volume, low-exception process and you'll get funded for three years. Choose the CFO's pet edge case and the programme dies in month four.
Here's the scoring checklist we run in discovery. It takes two weeks and it consistently changes what clients thought their priority was.
Score every candidate out of 25
| Criterion | 1 point | 5 points |
|---|---|---|
| Volume | A few times a month | Hundreds of times a day |
| Rule clarity | Requires expert judgement | Fully documented rules |
| Exception rate | Over 30% of cases | Under 5% of cases |
| Data availability | Paper / unstructured | Structured, in a system with an API |
| Business visibility | Nobody notices it | Named in board reporting |
Anything scoring 20+ is a strong first candidate. Below 12, fix the process before you automate it.
The full checklist
- Is it repeated at least weekly by more than one person?
- Can you state the rules without saying "it depends"?
- Are the inputs digital and consistently formatted?
- Is there a measurable baseline — time, cost or error rate — available today?
- Does the process cross systems that already re-key data manually?
- Is there a named business owner who wants this to happen?
- Would the automation still be needed if a planned system change lands next year?
- Is the compliance and audit position clear?
- Can you define a safe fallback if the automation fails at 2am?
- Would success be visible to someone who controls budget?
Red flags: don't automate these yet
Undocumented, tribal processes
Three people do it three different ways. Standardise first or you'll automate one person's habits.
High exception rates
Above ~30%, you'll spend the whole budget on edge cases and save nothing.
Systems being replaced in 12 months
You'll build an integration to a corpse.
Processes with no owner
Nobody to sign off requirements, nobody to maintain it after go-live.
"We just need AI on it"
A solution looking for a problem. Start from the cost, not the technology.
A two-week discovery plan
- 1. Days 1–3: collect candidates. One short form per team: task, frequency, duration, systems touched.
- 2. Days 4–6: observe. Sit with the people doing the top ten. What they do rarely matches the documented process.
- 3. Days 7–9: score and rank. Apply the 25-point model. Attach hours and rupees to each.
- 4. Days 10–14: define the first build. One process, a written baseline, success criteria, exception path and an owner.
Know what it'll cost before you commit
Run the 60-second Project Requirement & Cost Estimation Audit for an itemised budget range for your first automation.
Prefer to talk? Message us on WhatsApp