AI agents for WhatsApp, chat and your website
Agents that answer your customers, act in your tools and know when to hand over.
For businesses that answer the same questions every day, on WhatsApp, in a website chat or by email. I build agents that answer from your own knowledge, look things up in your systems, ask a person before anything new goes out and get better with every reply your team gives. Two of them are ready-made products I set up for you.
Sounds familiar?
- Your team spends hours answering the same questions on WhatsApp and email
- Customers write at night and on weekends and get an answer days later
- An AI prototype works in a demo but nobody trusts it with real customers
- Website visitors leave without finding the page that would have convinced them
How I work on it
- 1
Start with the real questions
We pick one channel and the twenty questions you get most, and build the knowledge base from your team's real answers, not from marketing text.
- 2
Connect your tools safely
The agent can look up a booking or an order, check availability or record a change request through your own systems, with least privilege and read-only wherever possible, for example over an MCP server.
- 3
A person stays in charge
When the agent does not know, it says so, alerts your team with a link to the chat and learns from the reply. New knowledge goes live only after someone approves it.
- 4
Production, not a demo
Staging, tests, monitoring and a budget per model, with a fallback model when the first one is slow or down. The agent is operated like any other critical system.
- 5
Ready-made where it fits
Two products shorten the way: StillAnswer, the AI assistant for WhatsApp, in production for a hospitality business, and Helptempo, the AI chat for websites, whose engine runs the guide you can try on this site. Both work with Gemini, Claude or OpenAI-compatible models.
What you get
- An agent on WhatsApp, in your website chat or both, in your customers' languages
- A knowledge base built from your team's answers, with an approval step
- Connections to your booking, order or CRM data with least privilege
- A staff inbox to read chats, reply and pause the AI per chat
- Monitoring, cost limits per model and a fallback model
Real projects
- Client, surf school & hostel, Portugal, since 2026Booking & operations platform for hospitality64
- Own product, 2026Helptempo: a sales and support chat for any website, built as a reusable engine~5 min
- SaaS & LMS, education sector, since 2025From AI prototype to a production SaaS160
- Hospitality and education SaaS, 2025–2026Custom sales CRM with AI lead scoring2
Questions buyers ask
Will the agent make things up?
It answers only from your knowledge base and your systems, says when it does not know, and hands over to your team. New answers go live only after someone approves them.
Which AI models do you use?
Gemini, Claude or any OpenAI-compatible model, including local models for sensitive data. A fallback model takes over if the first one is slow or unavailable.
How fast can it go live?
The website chat runs on a site in minutes once the knowledge file exists. A WhatsApp agent with your tools connected takes a few weeks, mostly for the knowledge base, testing and the WhatsApp Business approval.
What about data protection?
Keys stay on the server, personal data is left out where it is not needed, and for sensitive cases the model can run on your own hardware. We document which provider processes what.
How do you price it?
After a 30-minute call. The WhatsApp assistant comes either with your own accounts and a guided setup, or all-inclusive with the number and the full setup. Custom agents are a fixed scope or a few days per month.
Read more
- From AI prototype to production: the eight layers I add, in order
- Production AI on Microsoft Foundry: a private reference setup for agents
- Production AI on AWS: Amazon Bedrock and AgentCore in a private setup
- Production AI on Google Cloud: from Vertex AI to Gemini Enterprise Agent Platform
- MCP security in depth: how to let AI assistants into your product without opening the doors
- Building an AI assistant for WhatsApp, end to end: what it takes beyond the demo
- Phone AI agents: the architecture, and the pitfalls nobody shows in the demo
- Streaming AI answers on Azure Static Web Apps: the one header that buffered everything
- FinOps for AI spend: tokens, caching, routing and budgets that hold
- 3,025 pull requests in ten months: an AI-assisted delivery process, from the inside
- The EU AI Act for mid-sized companies: what applies to you, and when
- Free guide (PDF): Connect your company to AI with MCP
Get in touch
Tell me what you're building. I'll reply within 1–2 business days.
- Company
- MADO Consulting
- Follow me on LinkedIn
Available worldwide, remote. On-site on request, depending on the country.
Opens Calendly's booking calendar here. Loading it sends data to Calendly (see Impressum).