The demo is the product: Ergofon, our AI phone answering agents, from 90 demo configurations to a real customer pilot

The challenge

Small and medium businesses lose revenue to missed phone calls. Calls after hours, during lunch, or while staff are with customers go unanswered. A receptionist costs EUR 1,500-2,000 or more per month and still leaves nights, weekends, and callers who speak another language uncovered. We built Ergofon as our own answer: an always-on AI voice agent that picks up every call, holds a natural conversation in the caller's language, and takes bookings.

A voice product has a second problem: trust. Nobody buys a voice agent from a slide. You have to hear it. So we made the demo the product.

What we built

ergofon.com is built around talking to the product itself. Visitors speak with ERGO, a sales agent that sells Ergofon by being Ergofon - "You're talking to a voice agent right now" is a deliberate proof point in its script. ERGO's whole sales conversation is mapped out like a flowchart: it sorts callers into four types and follows scripts, decision points, and polite exits for each, so we can review and improve the conversation visually.

Visitors can also try a working demo agent for their own industry, in their own language: ten industries, nine languages, 90 configurations in total. The demo agents do real work, checking available time slots and booking appointments during the call.

The third piece proves it works beyond demos. Eva, a German-speaking agent that guides students and parents through apprenticeship opportunities, was built for a real pilot customer and runs from the same system.

How we built it

One system serves every agent. Instead of writing 90 agents by hand, each one is assembled the moment a call starts from shared building blocks: a common foundation covering who the agent is, its safety rules, how it handles off-topic questions, and how it speaks naturally on the phone, combined with one of ten industry profiles and one of nine language profiles. Adding an industry or a language means adding one profile, not rebuilding anything, and every profile is checked automatically at startup, so a mistake surfaces before the first call instead of in the middle of one.

Behavior is tested before every change goes live: automated tests cover all ten industries, and a second AI listens to scripted test conversations and checks that the agent stayed on script - the medical clinic agent must refuse to give a diagnosis, agents set to German must answer in German.

A phone call falls apart if the timing is off, so we engineered the conversation itself: the agent starts preparing its reply while the caller is still finishing, answers without awkward pauses, and does not talk over the caller. The AI's internal deliberation is filtered out before it reaches the voice, so the agent never reads its own thinking aloud.

What this build demonstrates

Ergofon is past the prototype stage: two agents are live, and the Eva pilot was built for a real customer. The next step - giving every customer their own phone number and their own agent setup - is already fully planned. Building Ergofon taught us where voice AI actually gets hard: not in getting answers from an AI, but in timing, natural turn-taking, keeping the AI's thinking out of its mouth, and proving how an agent behaves before each release. That is the experience we bring to any voice or conversational AI project.