The Debrief That Writes Itself

The Company

Halversen & Krug is a boutique recruitment firm in Hamburg, about 40 people. They work mainly in professional services and financial advisory, placing senior candidates across the DACH region. They're not a big-volume shop. They run maybe a dozen serious searches at any one time, and they take the process seriously — structured scorecards, trained interviewers, panel interviews at the final stage. It's a well-run firm.

The Pain

Miriam Krug, the managing partner, co-interviews every final-stage candidate. So do two or three of her senior consultants. After the call ends, everyone goes back to their desk. One person's on another call ten minutes later. One types a few notes into the ATS. Someone else writes a longer thought in the panel Slack thread. Miriam gets an email from the third interviewer that evening with a completely different take. By the time they actually sit down to decide, usually the next morning, nobody's quite sure what anybody said. They piece it together from three different places, they go round the table again to reconstruct things they already covered, and a lot of the sharper observations from the call just don't make it into the final conversation. Miriam's been doing this for years and it still happens almost every time.

The Move

When an interview is marked complete in their ATS, a pipeline kicks off automatically. It pulls the scorecard fields the interviewers submitted, searches the relevant Slack thread from the panel channel, and picks up any emailed feedback that's come in. An LLM reads all of it together and produces a short structured summary — where interviewers agreed, where they diverged, any specific concerns that were raised, and an overall read on the candidate. That gets posted into their hiring Slack channel within a few minutes of the interview ending.

This isn't about replacing the debrief. It's just that when Miriam opens the channel, the raw material is already organised. The interviewers aren't reconstructing from memory. The email that came in at 8pm is already in there. They can argue about the actual decision instead of spending twenty minutes working out what everyone thought.

Tools that do this today: Make or n8n to build the pipeline, the ATS webhook to trigger it, Slack's API to pull thread content, and any decent LLM via API to run the synthesis. It's a few connected steps, not a product.

The blind spot

The debrief call feels like it's working. People talk, they reach a decision, they move on. The problem is quiet — it's the stuff that got lost before the call started. That's harder to notice because you don't see what's missing.

The pattern

The same setup works in a few other situations:

• A professional services firm that runs client project reviews — feedback comes in from multiple people across email and Teams, and nobody consolidates it before the partner meeting
• A VC or family office doing investment committee prep, where deal notes, call summaries, and analyst comments are all in different places when the IC starts
• Any team doing structured sales debriefs after enterprise demos, where the account team each has a different read on what the prospect actually said