How a threat-detection startup stopped losing a week to every exec review

Before every board and exec review, this Atlanta security company's three PMs spent days reconstructing what had been decided, shipped, and learned. Now a decision log records all of it as it happens, and a bot trained on that log answers the questions execs actually ask. QBR prep fell from two weeks to two days.

2 days
QBR prep, down from 2 weeks
94%
of exec follow-ups answered same day
100%
of experiments logged with outcomes
3
PMs freed from deck archaeology

The situation

Every quarter, the same ritual: for two weeks before the exec review, the product team stopped doing product work and started doing archaeology. Which detections shipped, what the false-positive experiment showed, why the endpoint agent got deprioritized. The answers existed, scattered across Slack threads, Linear comments, and three people's memories, and every quarter someone rebuilt them into a deck by hand.

The CTO's frustration was sharper: the questions execs ask are predictable. What changed, what did we learn, what does it cost, when does the thing I care about land. The team was rebuilding answers to the same twelve questions four times a year.

The turn

The fix had two parts. First, a decision log that fills itself: every experiment, scoping call, and trade-off made inside the workspace is recorded with its rationale at the moment it happens, the way large orgs log every experiment internally, except nobody had to be assigned to maintain it.

Second, a bot on top of the log that knows what product and engineering executives ask. Before the next review, the CTO queried it directly: what shipped in detections this quarter, what moved the false-positive rate, where is the endpoint agent. Answers came back with links to the decisions behind them. The two-week deck build became a two-day review of answers that already existed, and between reviews, execs stopped waiting for the quarterly meeting to ask.

The questions were never the hard part. Every exec asks the same twelve things. The hard part was that answering them meant excavating three months of Slack. Now the log answers, and my PMs do product work in the two weeks before a board meeting.
CTO, threat-detection platform

How Second Axis fits

01

A decision log that fills itself

Experiments, trade-offs, and calls are captured with rationale as they happen inside the workspace, not reconstructed later.

02

A bot that knows what execs ask

Trained on the org's own decision log, it answers the recurring exec questions with citations, on demand, any week of the quarter.

03

QBR prep from living material

Review decks assemble from logged decisions and measured outcomes instead of two weeks of archaeology.

04

Short documents, linked evidence

ICs stopped writing long defensive documents. A one-page summary with links to the log carries every review.

Results

AreaBeforeAfter
QBR prep time~2 weeks, 3 PMs2 days
Exec follow-up questions answeredDays, sometimes next QBR94% same day
Experiments logged with outcome and rationale~30%100%
PM time on status reconstruction per quarter~60 hrs~8 hrs
Length of median review document14 pages1 page plus links

What came next

Week 0Workspace connected to Linear, Slack, and the experiment tracker. Log backfilled one quarter.
Week 3Decision log coverage hits 100% of active experiments.
Week 6CTO runs first exec-question session against the bot; answers cite the log.
Quarter 1QBR prepped in two days. Board follow-ups answered during the meeting.
Quarter 2Execs query the bot between reviews. Deck archaeology retired for good.

The two weeks per quarter came back as roadmap work, and the CTO got something he did not expect: better questions. When execs can self-serve the status answers, the quarterly meeting spends its time on the decisions that actually need the room.

Company details anonymized at the customer's request.