AI in Practice: How a better prep process created a new problem

This week we spoke to the head of sales at a mid-market SaaS firm selling compliance and scheduling software to care and housing operators. Several months earlier, one of her reps had walked into a late-stage meeting with less command of the detail than he realised.

The client’s procurement team had joined for the first time and wanted detail on expansion terms. The rep knew they’d agreed to a pilot but couldn’t share the conditions when asked: which regions were in scope, what the review criteria were, whether a timeline had been discussed.

The meeting ended with a list of points to go back and clarify, and what had felt like a formality started to feel less certain. Fortunately, the deal eventually went through, but it took another two weeks to get procurement back on a call.  

The firm had been using AI-generated prep briefs for several months by that point. Before that, reps built context from notes and whatever they’d retained from the last call. The issue now was that the prompt had been written to give reps something clear and confident before they walked into the room. But it had also flattened any ambiguity that was still under discussion.

For the research behind this, see aibl’s operator research on AI process changes.

The fix was to rewrite the output so it had to show what had been confirmed, what had only been raised, and what was still open, with each point linked back to the call it came from. If a rep couldn’t trace a claim on scope, pricing or rollout to the transcript, it stayed open.