Once a firm has mapped its work, it usually has a dozen candidates and one partner who already knows which one to do. That instinct is sometimes right. It is worth checking against three questions before anything is built, because the first workflow sets the tone for everything after it.
Three questions, one scale
For each candidate, score one to five on each of these.
- Effort. How much has to change for a system to take the routine part of this work? Does the information it needs already exist in a tool, or does it live in someone’s head? Five means the data is there and the steps are clear.
- Risk. What happens if the system is wrong and nobody catches it? Who sees the result: a colleague, a client, a regulator? Five means a mistake is caught internally before it costs anything.
- Payback. If the routine part ran on its own, what would the firm gain each month, in hours, dollars or output? Five means a number everyone in the room recognises as real.
Add them up. The candidates that score twelve or more are the shortlist. Anything under nine goes back on the list for another quarter, with the reason written beside it.
The so-what test
Before building the top candidate, ask one more question: if this worked perfectly, so what? If the honest answer is “a partner saves forty minutes a week”, that may be fine, but it should be chosen knowingly. If the answer is “new enquiries get a reply inside the hour instead of the next afternoon”, that is a different order of result, and it probably deserves to go first even if it scored a point lower.
The test exists because the most visible workflow is rarely the most valuable one. Content and marketing feel urgent and are easy to demonstrate. Intake and meeting documentation are quiet and compound. We would usually build the quiet one first.
Why the first one matters more than the others
The first redesigned workflow is where the firm writes its first Operating Standard, trains its first people, and learns what an approval point feels like in practice. If it goes well, the second workflow takes half the time. If it is chosen badly, the firm concludes that AI does not work here, which is the wrong lesson from the right evidence.
So the first workflow should be one where a person can check the system’s work easily, where the result is visible to the team within a month, and where a mistake is embarrassing rather than expensive. Meeting-to-record is often that workflow. Research briefs, for firms that do a lot of them, are another.
Write down the ones you did not choose
A list of candidates that were weighed and set aside is one of the more useful documents a firm can own. It stops the same idea being re-proposed every quarter, it gives a vendor something to be measured against, and when the first workflow has proven itself it tells you exactly where to go next.