How to prepare your business for automation (a practical checklist)
Automation projects fail in the preparation, before any software exists. This is the practical checklist we wish every business ran before implementing anything: what to write down, which process to pick first, and who to involve.
By the Terso AI team · Published 2026-07-11
Why does preparation matter more than the tool?
You cannot automate a process nobody can describe. When automation projects go wrong, the diagnosis is usually the same: the process only existed in someone's head, complete with exceptions nobody mentioned until week six.
The good news is that preparation is cheap. A few focused hours of writing things down does more for the outcome than any tool choice you will make later.
What should you write down first?
Three things, in plain language:
- The steps. Walk through the process as it actually happens, including the workarounds. Ten numbered steps in a Google Doc beat a formal flowchart nobody reads.
- The exceptions. What happens when the customer pays late, the form arrives incomplete, the vendor does not answer? Exceptions are where automation breaks, so an honest list of them is the most valuable document you can produce.
- The volumes and time. How many times a week does this happen, and how long does each one take? That arithmetic is what later tells you what the automation is worth.
Which process should you automate first?
Score your candidates on four traits: frequent, rule-based, expensive in staff hours, and low-risk if a mistake slips through. First projects should score high on all four. Invoice chasing, appointment reminders, lead intake, and report assembly are classic first picks. Pricing decisions and upset-customer conversations are classic last picks.
Resist starting with the most painful process. Painful usually means complex, and complex is a bad first project. Win something small, let the team learn to trust the pattern, then climb. Our post on custom software vs off-the-shelf covers how to decide what that first build should even be.
What data and access do you need to line up?
Automation runs on access to your existing tools, so inventory them: the CRM, the accounting system like QuickBooks or Xero, email and calendar, the phone system, the web forms, and the spreadsheets quietly doing load-bearing work.
Then note where the same fact lives twice. Duplicate records and re-keyed data are the potholes an implementation partner needs to see upfront. And plan for least-privilege access from day one: whoever implements should work from scoped, audit-logged credentials, never a shared admin password.
Who needs to be involved?
Three roles. An owner with the authority to change the process. The person who actually does the work today, because they know the real exceptions and their buy-in decides adoption. And whoever manages the tools, even if that is an outside IT contractor.
What you do not need is a big committee or a transformation initiative. The best preparation we see comes from two or three people who know the work, a shared document, and a couple of honest hours. Scale can come later; clarity has to come first.
The person doing the work today deserves the direct conversation about what changes for them. Automation lands best when it removes the part of the job people already resent, and saying that out loud, specifically, is how you get the exceptions list to be honest.
What does the full checklist look like?
Run through these eight items before any discovery call:
- Written steps for one process, as it actually happens
- A written exceptions list, including the embarrassing ones
- Weekly volume and time per instance
- An inventory of the tools the process touches
- Known duplicate data and double entry, flagged
- A named process owner with authority to change it
- The operator who does the work today, briefed and involved
- A definition of done: the result you would call a win in 90 days
If you can hand someone those eight items, you are better prepared than most companies we meet, and the first conversation gets concrete immediately.
Frequently asked questions
How long does preparation take?
A few focused hours per process, usually spread over a week or two of paying attention to how the work really happens. If you stall, start anyway; a good discovery process pulls the rest out of the team.
Do we need clean data before starting?
No. Perfectly clean data is a myth, and cleanup is often part of the first project. What helps is knowing where the mess is, so flag it rather than trying to fix it first.
What if we cannot decide which process goes first?
That ranking is exactly what the AI Opportunity Audit produces: your candidate processes scored by ROI, highest return first. Run the free audit, or book a complimentary discovery call and we will rank them with you.
More from the blog
Bring the checklist. We will bring the ranking.
The free audit turns your list of candidate processes into a roadmap ordered by ROI.