AI reads your inbox, not your mind
The most common failure
An automation drafts a reply to a customer and gets the tone wrong, because that customer is touchy about delays in a way everyone in the office just knows. It routes a request to the wrong person, because Dave handles the marine accounts, obviously. The machine was not being dumb. It was working from the written record, and the written record never mentioned any of that.
AI reads what exists: the inbox, the files, the notes. The unwritten layer of a business, which is a thick layer in most small companies, is invisible to it.
The fix is a sentence
The encouraging part: when an automation errs this way, the fix is almost never technical. It is a sentence added to the instructions. Marine accounts go to Dave. This customer gets extra care on timelines. Never quote delivery dates in a first reply.
Each correction takes seconds, and it sticks. Unlike telling a new employee, who may forget, the machine applies the rule every time from then on. The instruction file becomes the business's unwritten rules, finally written.
Feed it like you would a new hire
The pattern that works is treating the first month of an automation like a new hire's first month:
- Expect competent-but-generic output at the start, that is what working from the written record alone produces
- When something is off, name the missing fact instead of declaring the tool bad
- Add the fact to the instructions and move on, the corrections compound fast
A side benefit worth having
Six months of these corrections leaves you with something no consultant could have sold you: the tribal knowledge of your business, written down, current, and attached to running systems. That document outlasts any employee and any tool, and it is half of what makes automation survive turnover.
The audit answers this for your business
Two weeks, $2,500 flat ($1,000 for the first three clients), and you get the map of your own automatable work with dollars on it.