How we build AI telesales agents.
From conversation design to a live pilot — our process step by step, plus answers to the questions we hear most.
How we work
1. Discovery
We look at your sales process, target industry and the objections your team hears most — this becomes the basis for the conversation design.
2. Conversation design
Opening, discovery, pitch, objection-handling, close — as a conversation guide with prepared objection handling, not a script read aloud. Your voice, cloned with consent, becomes the agent's voice.
3. Build & configure
The agent goes live in our system: prompts, objection library and voice are wired together and connected to the telephony.
4. Test calls
Structured test calls before anything goes live. We measure latency, turn-taking and outcome quality — and only tune based on what we actually observe.
5. Pilot & launch
A small, monitored pilot first, real feedback, then a controlled rollout. No feature-building during the pilot — only measuring.
Frequently asked questions
Does the AI disclose that it's an AI?
Yes. The agent is instructed to say at the start of the call that it is an AI, and never to deny it when asked directly.
Which languages are supported?
There are agents in German, English and French; the language is resolved per call, so the agent speaks the right language for each lead.
Do we need to change our phone number or CRM?
No. We can integrate with your CRM instead of replacing it. Calls run over our telephony; connecting your own SIP trunk is in preparation.
Where is the call data stored, and who can access it?
Transcripts, outcomes and recordings are stored under your account.
Can we use our own voice?
Yes — with a short reference recording and your consent, we clone the voice your team already uses on calls.
How long until we can go live?
It depends on conversation design and test-call rounds, not on infrastructure — we'd rather run one extra round of test calls than launch a rough pilot on a deadline.