Most automation projects fail at the selection stage, not the build stage. Teams start with what is visible (a chatbot on the website) or what is fashionable, instead of what moves revenue. The result is a demo that impresses nobody and changes nothing.
There is a faster way to choose, and it takes about thirty minutes: follow one real inquiry through your company, end to end, with timestamps.
The audit: follow one inquiry
Take last week's most normal customer inquiry. Write down every station it passed: when it arrived, when someone first read it, when it was answered, when it became a record in a system, when the follow-up happened. Use real timestamps from the inbox and the CRM, not estimates.
Now look at the gaps. In a typical B2B company the message itself is handled in minutes — but it waits hours or days between stations. That waiting is your automation backlog, sorted by cost.
Score the candidates on three axes
Volume: how often does this step run per month? Automating something that happens twice a year is hobbyism.
Wait time: how long does work sit before this step happens? Waiting is where leads cool down and deadlines slip.
Structure: can you describe the rule a sensible employee follows here? If yes, AI can prepare or execute the step reliably. If the step is pure judgement — pricing, negotiation — leave it human and automate the preparation around it.
Why intake usually wins
Run this audit in almost any operational B2B company and the same step comes out on top: intake — the path from 'an inquiry arrived' to 'a qualified, structured record with an owner exists'. It is high-volume, it is where the longest silent waiting happens, and the rules are describable.
That is why our standard first project is an AI lead intake pipeline rather than a chatbot: it removes the most expensive waiting in the funnel without asking anyone to change how they work. Once intake is structured, the next steps — follow-up nudges, document drafting, routing — attach to it naturally.