WhatsApp support and sales
Answers, qualifies, books, and passes to your team only what needs a person.
Example of what we measure: time to first reply, % resolved without a human, bookings per week.
Custom AI agents that answer customers, sell and book appointments on WhatsApp — and take the repetitive work off your back office. Built by a team with 15+ years of putting software into production. Tested before launch. Measured after.
Mechanism, not magic
Every decision the agent makes leaves a record: what it understood, which tool it called, what it answered. That's how you audit it, measure it and make it better.
Simulation of a BYIA conversation · fictional data
09:41:02 whatsapp ← "hi, can I see dr. paula on thursday?" 09:41:02 classify → intent=booking confidence=0.97 09:41:03 tool.call calendar.find_slots(provider="paula", day="thu") 09:41:03 → 3 open slots: 9:30am · 2:00pm · 4:30pm 09:41:03 tool.call patients.lookup(phone="+55 62 ····") 09:41:04 → returning patient · insurance active 09:41:04 whatsapp → "Hi Marina! On Thursday I have 9:30am, 2pm or 4:30pm. Which works best?" 09:41:31 whatsapp ← "2pm" 09:41:32 tool.call calendar.book(slot="thu 2:00pm") 09:41:32 tool.call reminders.schedule(-24h, -2h) 09:41:33 ✓ appointment confirmed · no human needed
The cost of doing nothing
They asked for a quote after hours. Nobody answered. By morning, they'd signed with whoever replied first.
Everyone has built an AI demo. Almost no one has put one into production, with integrations, metrics and an owner. Demos don't pay the bills.
Expensive people copying data from one system to another and answering the same question for the hundredth time. Work a machine does better.
The offer
Your first AI agent answering on WhatsApp — in production in 45 days.
Investment: set after the diagnostic
A map of your current flow, volume and cost per conversation, plus a numeric goal, agreed in writing.
Built for your process and plugged into what you already use: calendar, CRM, spreadsheets, ERP.
A test suite built from real, anonymized conversations. The agent goes live only if it passes.
Every conversation, cost and error, visible to you. No black box.
Change the tone and the rules yourself — no waiting on us for every small tweak.
When a conversation needs a person, the agent hands it over with the full context.
We review the agreed metric with you every week and keep tuning the agent.
Before and after, with your numbers.
Ranked by return and effort.
If the agent doesn't pass the acceptance criteria defined in the AI X-Ray, we keep working at no cost until it does.
How it works
Week 1
Interviews, data and a goal in writing.
You: two 1-hour calls.
Weeks 2–5
Agent, integrations and a test suite.
You: review sample conversations.
Day 45
Live, with a dashboard and a human handoff.
You: sign off on acceptance.
Day 60
Before and after, and the roadmap for what comes next.
You: choose the next process.
Product
In beta with its first client clinics
BYIA is a WhatsApp front-desk agent for medical clinics, wrapped in a complete practice-management system. It comes from the same team behind Kadash — and it's already in beta with its first client clinics, which help shape every release before general availability. It's the same rigor we bring to your project.




What we build
Answers, qualifies, books, and passes to your team only what needs a person.
Example of what we measure: time to first reply, % resolved without a human, bookings per week.
Reads invoices, contracts and receipts; checks, classifies and enters them into your system.
Example of what we measure: documents processed per hour, review rate.
Your team asks, the agent answers with the source — manuals, policies, history.
Example of what we measure: questions resolved, search time saved.
Questions in plain language become queries and reports, without waiting in the data team's queue.
Example of what we measure: reports generated, time to an answer.
Agents that run workflows across systems, with rules and approvals.
Example of what we measure: hours saved per month, errors avoided.
Engineering
Our team comes from platforms where mistakes are expensive. We bring the same rigor to AI.
OpenAI and Anthropic, chosen per task, cost and latency. No lock-in to a single vendor.
Business rules, action limits and forbidden answers, defined before the first test.
Suites of real, anonymized conversations the agent has to pass before every release.
Every call, cost and decision logged and traceable (LangFuse).
Minimal data, anonymized test sets and role-based access, under Brazil's data-protection law (LGPD).
Applications on Cloudflare Workers: close to your users, scalable, and no servers to manage.

Who's on your project
Founder · Software engineer and architect
15+ years putting software into production — backend, platforms, distributed systems, cloud/edge and applied AI. At Kadash since 2020, he runs projects end to end: diagnostic, architecture, build and operation.
Our team has built production software for
Questions
It will, sometimes — people do too. That's why the agent has guardrails, goes live only after passing its evals, and hands the conversation to a human when it isn't sure. Every mistake is logged and becomes a test.
We use only the data we need, anonymize it for testing and control access by role. We handle everything under Brazil's data-protection law (LGPD) and put in writing who can access what.
No. The agent plugs into what you already use — calendar, CRM, spreadsheets, ERP. If something has no integration, we build it.
It depends on the process, the volume and the integrations — so the AI X-Ray sets the investment, in writing, before you commit to anything. The first 5 clients get founder pricing.
The agent stays live, with 30 days of hands-on support from us. On day 60 you get the ROI report and decide: we keep running it, we move on to the next process, or you take it over.
Your prompts, your rules and your data are yours. Code ownership terms are spelled out in the contract, before we start.
AI X-Ray · free
Fill in the calculator. If the number makes sense, request the full AI X-Ray: a 30-minute call and a one-page report within 72 business hours, covering the recommended process, feasibility, risks, a goal and an investment estimate.
72 hours a month you could get back; R$3,240 in potential value a month; R$38,880 a year.
Estimate based on your numbers: volume × minutes × % repetitive × cost per hour. Not a promise of results — it's the starting point of the X-Ray.