The expensive way to choose
Most automation initiatives start the same way: someone sees a tool demo, gets excited, and then goes hunting for a process to justify the purchase. The result is predictable — a fragile integration nobody wants to maintain, attached to a process that was never worth automating in the first place.
The order matters. Choose the process first, using criteria you can defend to your team and your accountant, and only then pick the tool. Here is the framework we use when scoping automation work with business owners.
Step 1: Inventory what actually happens
You cannot prioritize what you have not written down. Spend one week listing the recurring work in your business: every task that happens weekly or more often, who performs it, roughly how long it takes, and which systems it touches.
Aim for honest, not perfect. The owner of a clinic knows the front desk re-confirms appointments by hand every afternoon. The founder of an agency knows every proposal is reassembled from an old document. Write down what is true, not what the org chart implies.
Ask better questions
The people doing the work tend to under-report their own busywork, because it feels like admitting inefficiency. Instead of asking what is repetitive, ask:
- "If this task disappeared tomorrow, what would you do with the time?"
- "Which part of your week do you dread?"
- "Where do things fall through the cracks?"
Those answers surface the real candidates.
Step 2: Score each process on five dimensions
For every candidate, score from 1 to 5 on five dimensions:
- Volume. How often does it happen? A task that runs thirty times a week beats a task that runs monthly, even if the monthly one feels more important.
- Rule clarity. Could a trained intern do it from written instructions? Processes with clear rules and recognizable exceptions are far easier to automate than ones requiring taste or negotiation.
- Stability. Has the process worked roughly the same way for six months or more? Automating something that is about to change means rebuilding it immediately.
- Data readiness. Is the information it needs already digital and structured — in a CRM, a form, an inbox — or does it live in phone calls, paper and someone's memory?
- Pain. What breaks when it is done late or wrong? Work with direct customer or cash impact deserves a higher weight than internal convenience.
Add up the scores. You now have a ranked list that is not based on whoever spoke loudest in the meeting.
Step 3: Sort into three lists
The scoring produces three natural groups, and only the first should be automated now:
- Automate. High volume, clear rules, stable, decent data. Typical first projects: quote preparation, appointment confirmations, invoice intake, lead routing, routine reporting.
- Assist. Valuable but judgment-heavy. Here AI drafts and a human approves — draft replies, research summaries, prepared documents. Removing most of the typing while keeping all of the deciding is often the best return in the whole program.
- Leave alone. Low volume, shifting rules, or high stakes with no pattern yet. Revisit in six months; processes change.
The most common mistake at this step is automating a process that should simply be deleted. If nobody would miss it, remove it — automation is not a recycling bin for dead work.
A worked scenario
Consider a 20-person distribution business. The inventory turns up three candidates: sales reps copy web-form leads into the CRM by hand, an office manager converts email quote requests into quotations, and the dispatcher re-types delivery notes into the tracking system.
Scored honestly, the quote workflow wins. It runs daily, follows clear rules, has not changed in a year, starts from structured emails, and every hour of delay costs orders. Lead entry scores well too. The delivery notes matter, but they depend on phone calls and handwriting — poor data readiness — so that one goes on the assist list until intake is cleaned up.
That is the shape of a good first project: one workflow rebuilt end to end and measured against the hours it used to consume — not a company-wide platform migration nobody finishes.
Your first 30 days
- Week 1: build the inventory with your team.
- Week 2: score it, sort the three lists, and pick one project — not five.
- Weeks 3 and 4: document the chosen process precisely, define what done-right looks like, and design the workflow with exception handling: what happens when the system is unsure.
One well-chosen process, automated properly, does more for your operations than a shelf of tools — and it builds the internal trust that makes the second, third and fourth automation easier to land.
If you want a second opinion on your ranked list before investing, that is exactly where our AI automation service starts. You can book a free consultation and walk through your inventory with us.