Principles
AI should know its job — and its limits.
Seven principles, and the last three matter most. An agent gets relevant context, uses approved knowledge, acts through explicit tools, follows communication policy, respects commercial authority, escalates the decisions that matter, and leaves everything it did visible.
Communication policy
Spain · enforced before a call is placed
These are enforced by the platform before a call is placed, not described to the model and hoped for.
It uses approved knowledge, not its own fluency
Claims about your service, your pricing and your capabilities come from records somebody approved. An agent that can compose a plausible claim will eventually compose a false one, in writing, to a customer.
Authority is enforced outside the model
“Never offer more than 5%” in a prompt is a preference. The refusal happens at the point the offer would be made, which is the only place it counts as a limit.
Communication policy is deterministic
Calling hours, attempt caps, opt-outs, disclosure and consent are checked by the platform before anything is sent or dialled. None of it depends on the model choosing to comply.
Everything it did is visible
Every message, every call, every action, with which agent and which version did it. An automation layer you cannot inspect is one you will eventually switch off in a hurry.
Questions this answers
The things you currently have to ask somebody.
Each of these is a question an owner asks regularly and currently answers from memory, a spreadsheet, or by interrupting whoever is nearest.
What can the AI say about our company?
What can it offer, and what needs me?
Could it contact somebody it should not have?
What exactly did it say to that customer?
Where this stops
What it does not do.
None of this makes an agent good at selling. A correctly bounded agent pointed at a badly chosen audience will do the wrong thing consistently and within its limits, and choosing the audience is your judgement rather than the platform's. Nor can these controls judge tone: an agent permitted to email customers can send something clumsy, and no boundary catches that — which is why draft-first exists and why most teams use it for a fortnight. And we hold no position that this makes you compliant with any marketing or telecoms regime; it enforces the rules you configure, and configuring them correctly for your market is your decision with your own advice.
Next
Where this connects.
Make sure every opportunity gets a next action.
Let AI handle the repetitive engagement, keep follow-up moving, and bring your team in when a customer needs a real conversation.
Agents identify themselves on every call · Nothing commits without you · [email protected]