For the smallest real scope — one workflow, a core knowledge base, one channel — building a voice AI agent takes about a week. The process runs from a scoping afternoon through build and internal testing to a demo you try yourself, then go-live and a month of tuning on real conversations. Your side of it is two afternoons, roughly one to two hours each. More workflows, integrations, or channels add build days — the same scope that moves the price.
What does the implementation process look like?
Getting a voice agent live is a one-time onboarding, not an open-ended technology project. The path: confirm the business case, build the agent, test it, put it live, tune it on real conversations. Then it runs, and we keep it current.
The phases in order:
- Discovery — before the first call, we research your business and arrive with a hypothesis.
- Scoping — one afternoon with you: confirm the case, lock one workflow, set the rules.
- Build — internal, three to four days: knowledge base and agent built in parallel.
- Internal testing — simulated conversations and human review, before you see anything.
- Demo and adjustments — a second afternoon: you talk to the working agent.
- Go-live — a soft launch if you want one, then full 24/7 with monitoring.
- Tuning — the first month after launch, included in the setup. Ongoing maintenance after that.
Work normally starts about two weeks after the contract is signed. That window covers the infrastructure check that comes before the agent's brain gets built.
Discovery and scoping: one afternoon of your time
Discovery is our time, not yours. Before the first call we analyze your business — website, services, pricing, reviews — and form a use-case hypothesis: the single most valuable job an agent could do for you. The first session opens with our findings for you to confirm or correct, not with a questionnaire.
The scoping afternoon takes one to two hours and settles four things: whether the agent makes business sense, which single workflow it handles first, what knowledge exists and in what shape, and the business rules — what the agent may promise and what needs a human.
Bring three things:
- A decision-maker. Someone who can lock the workflow and set the boundaries in the room.
- Your knowledge as it is. Website, price lists, FAQs, internal documents. Messy is fine — structuring it is our work.
- Your rules. What the agent may say, what it must never promise, and when a human takes over.
The build: about a week for the smallest scope
The build is internal work; it needs roughly none of your time. Two tracks run in parallel for three to four days. The knowledge track turns your materials into a structured knowledge base the agent answers from, with open questions flagged instead of guessed at. The agent track builds the agent itself: persona and tone, conversation flow, connections to your tools, and voice calibration — speed, tone, pronunciation of your business terms.
Then we test before you see anything. Over fifty simulated conversations — routine questions, edge cases, deliberately difficult ones — plus a human review of the transcripts and the voice. Our own bar for moving on: would we trust this agent with your customers right now?
For the smallest real scope — one workflow, a core knowledge base of up to about twenty documents, one channel — the build takes about a week. That week is the build clock, counted from when the build starts, not from the day the contract is signed.
Demo, adjustments, go-live
Your second afternoon is the demo. You talk to the working agent — a real conversation, not a walkthrough of slides. The question on the table is simple: would you be comfortable if a customer talked to this today? Your feedback is captured in writing and drives a day or two of adjustments.
Go-live can be gradual. A soft launch puts the agent on limited pages or limited hours first; full 24/7 follows when you are comfortable. From day one the agent runs with monitoring — alerts fire on low-confidence answers and unexpected topics, so nothing drifts quietly.
What happens after launch?
Real customers ask different things than anyone predicts. The first month after launch exists for exactly that: two to three feedback cycles on real conversations — filling knowledge gaps, tuning tone and behavior, watching how often the agent resolves, hands off, and books. This tuning month is included in the setup.
After the first month, maintenance takes over: updating information when your business changes, monitoring quality, making small adjustments. It runs on what the conversation data shows, not on a calendar of meetings.
Why does more scope mean more build time?
Every addition to the scope is real build work, so it adds build days. Ballpark figures, never a quote — the real plan is confirmed when we scope your case:
- The base setup — integration, the agent's knowledge and conversation flow, one channel, a first round of testing — is about a week.
- Real integrations — three or four tools, actions beyond the website such as booking a meeting in your system — add about a week.
- Visual presentation — the agent scrolling your site, changing pages, highlighting content — adds about a week.
- Each additional channel adds a couple of days. A channel is a deployment, not a rebuild.
- A complex codebase adds a day or two; extensive testing adds about half a week.
We size the narrow case first: the smallest agent that does the job well. Extensions come later, if they earn their place.
Build time is also the pricing logic. Setup starts from €6,000, and the same things that add build days — workflows, integrations, channels, knowledge depth — are what move the price.
Common questions
How long does it take to deploy a voice AI agent?
For the smallest real scope — one workflow, a core knowledge base, one channel — the build takes about a week. Larger scope adds build days: real integrations or on-site visual presentation each add about a week. Work normally starts about two weeks after the contract is signed; that window covers the infrastructure check before the build.
How much of my time does the implementation take?
Two afternoons, roughly one to two hours each. The first is the scoping session where we confirm the case and set the rules; the second is the demo where you talk to the working agent and give feedback. Everything between and after is our work.
Can a voice agent really be production-ready in a week?
Yes — but the week is the build, not the whole relationship. In about a week there is a working agent to try; then comes a month of tuning on real conversations, included in the setup, and ongoing maintenance after that. Fast and production-ready — not a weekend demo.
What do I need to prepare before the first call?
Very little. We research your business before the call and open with our own hypothesis for you to correct. Bring a decision-maker, your existing materials — website, price lists, FAQs — and a view on what the agent may and may not promise. Messy documents are fine; structuring them is part of the build.
Do we have to go live everywhere at once?
No. A soft launch puts the agent on limited pages or limited hours first, and full 24/7 operation follows when you are comfortable. Either way the agent runs with monitoring and alerts from day one.