Each deal you close, won or lost, tells you something about why a buyer picked you or picked a rival. Trouble is, that story usually gets squeezed into a single HubSpot field, typed by a rep rushing to their next call. One of those notes on its own is close to useless. Line up a hundred of them from a quarter and they show you plainly where you're winning, which competitor is taking deals off you, and the point where your pipeline keeps quietly leaking.

So we have an agent on Machine read our closed HubSpot deals every week and post what it finds to the sales channel in Slack. It reads deal data and does nothing else; a Slack post is all it ever produces. It's how we keep an eye on our own win rate.

The problem

Reps record close reasons and rival names in HubSpot because the form makes them, and then nobody looks at those fields again. A lost deal gets a short reason, perhaps a competitor's name, and slides into the archive. Because no one adds them up, the same losing pattern can run for six months. Maybe one competitor keeps undercutting you on price in a single segment, or one type of deal always stalls at the same step. Eventually a sales leader senses it during a QBR, on instinct alone.

The usual answers don't solve it. A win-rate figure on a dashboard shows that the number moved, not the reason. A quarterly deal review is thorough, but by the time it's presented it's describing three months ago, far too late to change how you sell this week. And no one is keen to read two hundred close reasons by hand every Monday morning.

What we built

Once a week, a schedule on Machine starts an agent. It opens a fresh session (a cloud sandbox) that can read HubSpot and nothing more, then fetches every deal marked closed-won or closed-lost since its previous run. It splits the results by segment, competitor and price band, and pins down the stage at which the lost deals actually fell away. From that it draws out the themes, attaches a concrete suggestion to each, and posts the summary in Slack. HubSpot is left exactly as it was.

How it works

The pattern

Each week a cron opens a fresh session with a connector that can only read HubSpot. The rules for breaking deals down are kept as a skill. The agent goes through every closed deal, and the Slack post is the only thing it writes.

Guardrails

The agent can see every closed deal in HubSpot, so its access is narrow and runs in one direction only:

• Isolation. Each run happens in a sandbox of its own. The session can reach HubSpot and nothing else, and the Slack post is the only thing that comes out.
• Scoped secrets. The HubSpot key is stored encrypted in the Secrets Manager and only handed to the agents you grant it to, when the run begins.
• Read-only. The HubSpot connection can't write. The agent has no way to change a deal's stage, close reason, competitor or amount. All it can do is report what's already recorded.
• Report only. The agent never reaches out to a rep, a prospect or the buyer on a lost deal. It points out the pattern, and a person decides what to do about it.
• Everything is code. How it breaks deals down, its skill and its HubSpot permissions are files in the repo. They're versioned, and any change goes through a reviewed change request, not a toggle on a settings page.

The outcome

The reasons behind each win and loss used to sit untouched in a HubSpot field. Now they turn into a weekly roundup in the channel sales already lives in, with the themes spelled out and a suggestion next to each one. The agent does the reading. The team decides how the playbook changes.