AI assistants. Answers your customers can rely on, with a person one step away.
We build assistants for your website, WhatsApp or internal tools that answer from approved knowledge, capture requests and hand over to a person when needed.
For organizations handling repetitive questions at volume from patients, policyholders, guests, customers or staff.
AI assistants & chatbots
We build the knowledge base from approved sources, design the conversation and the handover rules, and connect the assistant to the channels your customers use. Sensitive topics and anything uncertain go to a person, and every conversation is logged for review.
- (01)Answers grounded in approved sources
- (02)WhatsApp, web and internal channels
- (03)Clean handover to people
- (04)Multilingual where you need it
Tools we use for this.
- WhatsApp Business Platform
- OpenAI
- Anthropic Claude
- Retrieval (RAG)
- Next.js
- Supabase
We choose the stack that fits your requirements. Tool names describe our skills, not partnerships.
Sprint.
4 weeks.
Four weeks on one goal, with working software at the end.
Included
- (01)Knowledge base from approved sources
- (02)Conversation and handover design
- (03)Channel integration (web, WhatsApp)
- (04)Guardrails and sensitive-topic rules
- (05)Testing with real questions
- (06)Analytics and review dashboard
Not included
- (01)Channel fees, such as WhatsApp conversation charges
- (02)Writing new policy content
- (03)Fully automated decisions on refunds, claims or medical questions
How it runs
- (01) · Week 1
Scope
Agree topics, sources, languages and when to hand over.
- (02) · Weeks 2–3
Build
Knowledge base, conversation flows and channel setup.
- (03) · Week 4
Pilot
Soft launch with review of every conversation, then widen.
Also available as: Build & Support (Staged, ongoing). Compare the formats.
You might also
look at these.
Can it work on WhatsApp?
Yes, through the WhatsApp Business Platform, alongside web chat or internal tools.
What if a customer asks something it should not answer?
Sensitive or uncertain questions go to a person with the conversation attached. You decide which topics the assistant must never answer itself.
Which languages does it support?
Major languages work well with current models. For less widely supported languages we test quality first and keep a person in the loop.