AI that works your playbook, not around it.
Ask a sales tool how your pipeline looks and it will usually flatter you: a big number, a colourful funnel, everything trending up and to the right. In our live demo for 20+ European medtech distributors, we asked the same question and got something rarer — a verdict.
Ten open deals, €621,600 — and the system called out that not one of them had a next step scheduled. No task, no meeting, nothing. Every deal quietly drifting, flagged before a manager had to notice. The rule it applied is one any good sales leader already believes: no next meeting means not active.
Everything you'll see here ran on a fictional demo distributor, built for the session — we say so on every screen. What matters is the mechanics, and the mechanics are what runs in production.
The question that separates a chatbot from an operating system
Later in the demo we asked something no generic AI can answer honestly: "This defibrillator is overdue, out of warranty and has no service contract. Per our service policy and sales playbook, what should we do?"
The answer didn't come from the internet. The system read the company's own service policy and sales playbook, applied them to one specific serial number, and named its sources: the never-serviced rule triggering an immediate maintenance proposal, the escalation threshold long since crossed, the contract-first rule for uncovered devices — and the playbook's own seven-year clause tipping the recommendation from "repair" to "replace". Four of the company's rules, firing on one device, cited like a careful colleague would cite them.
Generic advice is easy and worthless: any chatbot will tell you to "consider contacting the customer". An operating system tells you what your company does in this situation, because it knows what your company does. That's the whole difference.
From answer to action — with a human at the end
Then the natural next step: "Draft the email." The system read the device record, kept the policy reasoning, wrote the client email — factual, calm, with the real serial number and dates — and filed it as a draft on the account. And then it stopped.
Nothing was sent. That's deliberate, and it's the division of labour we believe in: the AI drafts from real data and your own rules; your people decide. Not because the technology couldn't press send — because a medtech business shouldn't want it to. The value isn't replacing your best account manager. It's giving every account manager the preparation your best one does on a good day, in seconds, on every account.
What makes it possible: the AI has to know your company first
None of this works as a bolt-on feature. It works because the company's operating knowledge — service policy, sales playbook, delivery rules — is written into the system's knowledge base, where the AI reads it before it answers. That setup is smaller than it sounds: a handful of documents, an afternoon of structured work. We've described the method, and a real delivery of it at a Czech distributor, in "Your AI shouldn't start from zero every day" →
Once it's there, every question gets answered the way your company would answer it — consistently, at any hour, by an AI that cites your rules back to you. Your playbook stops being a PDF nobody opens and starts being how the system behaves.
The takeaway
Here's the test worth running on any AI you're offered: ask it a question whose correct answer depends on your rules, not general knowledge. If it can't cite your policy, it's a chatbot. If it can — and then drafts the action and waits for your person to approve it — it's an operating layer.
Getting your playbook into the system is an afternoon. Seeing what it does on your data is 30 minutes. Book the working session →
Owner: Jan Safka · Healtek field notes, August 2026. Clips are from the live IntMeDA demo (4 Aug 2026), running on a fictional demo distributor; the knowledge-base delivery referenced is a real customer project.