Four things
- Volume — how many of these per week, and the peak
- Current cost — who does it, how long it takes
- Thirty real examples, including the awkward ones
- The destination — where output must end up, and whether it has an API
The awkward examples matter most. Thirty clean ones tell us the easy case works, which we already assumed.
The destination question moves the price
Whether your accounting system, CRM or job platform has a usable API affects the estimate more than any technical choice on our side.
We ask early because the answer sometimes changes the recommendation entirely, and it is better discovered before quoting than in week six.
What we look for in the examples
- How much variation there actually is
- What proportion are the difficult cases
- Whether the data quality supports the intended output
- The things nobody thought to mention
When we say do not do it
| Signal | Recommendation |
|---|---|
| Volume too low | Not worth automating |
| Process about to change | Wait |
| A product does this | Buy it, build the connection |
| A rule would do it | Write the rule |
| Nobody will own it | Resolve that first |
We would rather lose a project than deliver one that cannot pay for itself.
What you get
A written scope with what is in and out, a fixed price for the first phase, a duration, the assumptions and the risks with how each would be handled.
If you take that document elsewhere, that is fine. A clear scope is worth having regardless of who builds it.