From the
build.

Claire learns from what happens after she sends something. A reply means the approach worked, silence means it did not, and her weights shift a little each week. That loop only holds if a reply is a reply.

It was not. When a recruiter logged a note on a deal, the sync read it as inbound activity, and Claire counted her own handwriting as proof that her draft had landed. Every note she wrote made her more confident in whatever she had done last. We caught it while reviewing the previous week's work rather than from a customer telling us, which is the one part of this I am glad about.

A reply now has to arrive from outside your team before it counts as one. Sends and replies are measured over the same window, which they were not before, so the reply rates you saw last month were flattering. And every value Claire learns sits inside a range a person signed off, so a strange week moves her within bounds instead of past them.

She also stamps her version number on every action and every proposal. If her judgement shifts, you can trace it to the release that shifted it.

CRM Brain ran seven agents for most of a year. Each had a job, a feed, and its own read on your pipeline, which left you holding them together. Firms kept telling us they wanted one competent hire.

There is one now. Her name is Claire, she keeps a single memory across the pipeline, the outreach and the follow-ups, and the app collapsed from eleven pages to four around her.

The change that matters more is smaller than the name. Until this week, disagreeing with the agent meant undoing what it did, then doing that again the next week, and the week after. Now you tell her. Leave that client alone until October. Never open with a question. Route Munich to Sarah. She keeps it, and she applies it without being reminded.

She can also tell a question from an instruction, so asking what she makes of a deal no longer books work you never requested. She runs to a daily spend ceiling. And when the model is unreachable she says so in a sentence, where she used to leave a spinner turning.

CRM Brain started out useful to any agency with a CRM, and that turned out to be the problem. Advice that fits a marketing retainer, a staffing desk and an executive search mandate at once has to stay vague enough to fit all three, and vague advice about a deal helps nobody on a Tuesday morning.

So the product is built for executive search first. Leak detection knows what a retainer looks like, how long a shortlist can go quiet before it is dead, and how far a placement usually sits from its invoice. Those are the numbers a search firm loses money on, and none of them mean much to a marketing agency.

Pricing moved with it. The free trial went away and founding access took its place: one payment of €199 for Growth-tier access, and no seat counting while you are still deciding whether this works.

Marketing and B2B services firms still run on CRM Brain. The scoring is tuned for search, and we would rather say that plainly than claim the product is equally good at everything.

CRM Brain is opening for early access, slower than most software launches and on purpose. Every new team gets a real onboarding first, so nobody logs in to a dashboard full of zeros.

If you're on the waitlist, you'll hear from us directly with your slot. Setup takes about four minutes against your HubSpot account, and the first morning brief lands the next day your pipeline has activity.

Salesforce and Pipedrive support is next. HubSpot is live today.

Most of our early conversations are with Personalvermittlungen in Germany, Austria, and Switzerland, so shipping a real German translation rather than a machine-translated pass moved up the roadmap.

Every core surface is now bilingual and switches per user: Dashboard, Pipeline, Goals, Outreach, Actions, Settings, Onboarding, Analytics, Scoring, Leak detection, and data hygiene. DACH accounts default to German; everyone can switch back.

The agent Situation Room and a handful of auth screens are still English-only while we finish the pass. Those are next.

ITCS is where German staffing and recruiting agencies compare notes on what's actually working. We wanted our first public demo to be the real product rather than a video of it, so we ran the Situation Room live: seven agents reasoning against a real pipeline, on stage.

The question we got most wasn't whether the AI works. It was what happens when it's wrong. That's the right question, and it's why every write to HubSpot still waits for a recruiter's approval by default. More on that below.

It would be easy to build an agent that reads your pipeline and quietly rewrites it overnight. We built the opposite: every agent proposes, and a consultant approves the specific action before anything touches HubSpot.

That trust boundary is slower in the demo and faster in practice. Firms don't have to audit a black box after the fact. They watch the reasoning as it happens, in the Situation Room, and approve what they agree with.

HubSpot stays the source of truth. CRM Brain is the layer that reads it every night and tells you what to do about what it found.

CRM Brain started from a simple observation: a recruiter carrying 200 open deals can hold maybe five of them in working memory at once. The rest gets managed by hope, or it doesn't get managed at all.

A dedicated RevOps hire fixes this, and prices out every agency under about twenty seats. CRM Brain is the alternative: seven agents that read your pipeline nightly and tell each recruiter, in ninety seconds, which deals to push, which are slipping, and why.

It's built and run by one person, bootstrapped, from just outside Hamburg. That shapes the product. It stays direct about what it does and what it doesn't.