Build model
Fixed scope
talk to customers
Trained on your docs and pricing — not a canned FAQ script — it answers at 3 am, in any language, books the call, and captures the lead. Resolving ~65% of routine questions with zero staff time is typical for this type of automation.
Build model
Fixed scope
Timeline
2–8 weeks
Runtime
Monitored
Handover
Owned by your team
Visual summary of what this build includes from quote to handover.
Scope
Written
Deliverables and acceptance checks are fixed before build.
Execution
Instrumented
Retries, logs, and alerts are configured for production.
Controls
Approved
High-risk actions are gated by named human approvers.
Transfer
Owned
Runbook and account ownership move to your team at launch.
There are 168 hours in a week and your team covers about 40 of them. For the other 128 — evenings, weekends, 3 am — every visitor asking about pricing, turnaround, or “do you work with X” hits a contact form and waits. Most don’t wait: they open the next tab in the search results.
The questions themselves aren’t hard. Pull your chat logs and inbox and you’ll find the same 20–25 questions on repeat — roughly 65% of what your team answers by hand is routine (typical for this type of automation). That’s paid staff time spent retyping answers that already exist in your docs.
We index your real material — service pages, price list, policies, past email answers — so the bot replies with your facts, in your tone. Pricing answers come from a locked table, never improvised, and anything off-script gets an honest “let me get a human.”
Claude · Postgres + pgvector · Notion · Google Drive
The bot books calls straight into your calendar, writes qualified leads into the CRM, and opens WhatsApp with context pre-filled. Every conversation ends at one of three exits — lead captured, call booked, or human reached. No dead ends.
Cal.com · HubSpot · WhatsApp · n8n
When the bot hands off, your team gets the whole conversation in Slack — no “can you repeat that?” for the visitor. Every chat is logged and tagged, so you can see what people actually ask and fix the page that keeps confusing them.
Slack · Google Sheets · n8n
ROI math
10 hrs/wk of repeat answers × $25/hr × 52 = $13,000/yr recovered — typical build $2.5–6k
Typical for this type of automation — the free audit runs this math on your real chat and email volume.
Week 1
Top-25 question inventory from your real inbox and chat logs, ranked by volume, with the 5 highest-value automations flagged
Week 2
Conversation flows, tone rules, escalation paths, and the integration map (calendar, CRM, WhatsApp) — signed off before we build
Weeks 3–4
The bot live on your site: trained on your docs, wired to Cal.com and your CRM, tested against 100 real past conversations
Most chatbot builds land at $2.5–6k fixed, quoted in writing after a free audit. Ongoing cost is tool fees only — typically $20–80/mo for hosting and LLM usage with prompt caching. No per-seat fees, no retainer required.
2–4 weeks from kickoff. Week 1 is the audit of your real questions, week 2 the flow design you sign off, weeks 3–4 the build and testing. If a first version isn’t live in 30 days, you don’t pay.
Claude for language, Postgres + pgvector for your document index, n8n for actions like CRM writes and notifications, Cal.com for booking. Every account is in your name — no black box, and firing us doesn’t kill the bot.
Your documents are indexed into a database you control. Conversations are logged with PII scrubbed first, chat history stays in the visitor’s browser session, and the system prompt never leaves the server. We’ll sign your DPA; API traffic isn’t used to train anyone’s models.
Fill this quick form and we’ll send a fixed written quote for this exact build.