Insight14 October 20264 min read
Automate your SOPs, not your business: why we start by listening
Most automation projects fail because they force a business into a tool's idea of how work should be done. We do the opposite.
By Michael PresecanAIgenica
There is a quiet reason why so many automation projects disappoint. They start with a tool. The tool has an opinion about how a booking, an invoice or a support ticket should flow, and the business is asked to adapt.
The business usually resists, for good reason. The way a clinic handles a nervous patient, the way an accounting firm chases a late client, the way a hotel answers a complaint, was shaped by years of experience. It works. Replacing it with a generic flow makes the work worse, even if it makes it faster.
Your procedures are already the hard part
When we meet a business, the operations are almost always solved. The knowledge just lives in people's heads and in a hundred small habits: which emails are urgent, which customer gets a call instead of a message, which exception needs the owner's approval.
That is exactly what an automation needs to know. So we start there.
How we listen
- We watch real cases. Screen-shares and shadowing, not slides about how it should work.
- We write it down. Every step, decision and exception as a plain-language procedure your team approves.
- We mark the boundaries. What an agent may do alone, what needs approval, what stays with people.
Only then do we connect the tools. The agent follows your procedure, step for step, and hands over whatever falls outside it.
Why this works better
Automations built this way break less, because the exceptions were captured up front. Teams trust them sooner, because they recognise their own way of working. And the documentation stays with you: even if you never automate a thing, you now have procedures you can train new people with.
The goal is not to change how you work. It is to give your team back the hours they spend on the repetitive part of it.
If you have a process that eats your team's time, the first step is a conversation, not a demo.




