What we learned from 200 automation projects
Most of them weren't about AI at all. They were about one boring task nobody had time to fix.
Sam Carter
·
9 min
Over the last five years we've built more than 200 automations for small businesses: clinics, trades, shops, agencies and law firms. When we sat down to look back at all of them, the pattern surprised us. The projects that changed the most were rarely the clever ones.
The best projects start small
The projects that paid off fastest were never the big 'transform the business' ones. They were a single, boring task that someone did every day and quietly hated: answering the same question, chasing the same invoice, copying the same details between two apps.
Small projects win because they ship. A two-week build gets used, tested and improved. A six-month programme gets debated, delayed and quietly abandoned when the person who championed it moves on.
Most successful first projects took under two weeks to build
Nine in ten saved time in the first month
The biggest wins came from phones and inboxes, not dashboards
AI is rarely the hard part
The hard part is understanding exactly how a business works today. Which spreadsheet is the real one? Who actually answers the phone at lunch? What happens when a customer cancels by text instead of email? We spend more time listening than building, and it shows in the results.
When a project struggled, it was almost never because the technology wasn't good enough. It was because a step in the process only existed in someone's head, and we found out about it after launch.
People care about the handover
Owners don't mind a machine answering routine questions. They mind a customer getting stuck. Every automation we build now has a clear, fast way to reach a person, and we test that path as carefully as the automated one.
The best compliment we get isn't 'this is clever'. It's 'I forgot it was running'.
What we'd tell any owner
Pick the one task that annoys you most this week.
Write down every step, including the awkward exceptions.
Count how often it happens and how long it takes.
Automate the boring middle, and keep a person on the edges.
That list is your first automation. It won't feel ambitious, and that's exactly why it will work.
