About Platform Resources Articles Contact
IMA AI  ·  Resources

The Chat Assistant
Framework

How to put an AI assistant on your website that actually helps — answers from what you control, qualifies real intent, and hands a human only the warm leads.
Version v1.0
Date June 2026
Free No form. No email.
The Problem

Most website chatbots make things worse.

They either make things up, or they're a glorified FAQ box.

A chatbot that invents answers erodes trust the moment it's caught. One that only knows a few canned replies sends everyone to "contact us" anyway. And the common "fix" — route every visitor straight to WhatsApp — just moves the problem: now a human answers the same basic questions all day, and the genuinely valuable leads are buried in the noise.

We built assistants for our own sites and brands and refused all three failure modes. This is the framework we use.

The Principle

Answer from what you control. Qualify. Then hand off.

Rule 1Answer from a controlled knowledge base.
The assistant only answers from facts you've approved — products, policies, FAQs. It never invents, and it never quotes a number that isn't in the knowledge base. If it doesn't know, it says so and offers a human.
Rule 2Qualify before you escalate.
Not every visitor is a lead. The assistant reads intent and asks at most a few short questions to tell a buyer from a browser — so a human's time is spent only on the people worth it.
Rule 3Hand a human only the warm leads.
Browsers get a useful answer. Warm leads get captured and passed to a person, with the context already gathered. The human starts the conversation already knowing who they're talking to.
The Funnel

Greet → answer → qualify → hand off.

Every conversation runs the same path.
1GreetA clear opening that sets what the assistant can help with.
2AnswerRespond from the knowledge base, and cite the page or product link. If it's not known, say so.
3Detect intentBuyer, browser, support, or partner? The reply adapts to who's asking.
4QualifyFor commercial intent, ask up to three short questions — what they want, who they are, how to reach them.
5Hand offCapture the lead and context, notify a human, and tell the visitor what happens next.
How It's Built

Two brains: behaviour and facts.

Keep them separate — it's what makes the assistant easy to run.
LayerWhat it holdsHow often it changes
Config (behaviour)Role, tone, what it can and can't do, when to escalate, how to hand off.Set once, rarely.
Knowledge base (facts)Products, prices, policies, FAQs, guides.Whenever the facts change.

Behaviour lives in the config; facts live in the knowledge base. Change a price? Edit the KB — you never touch how the assistant behaves. Change the tone? Edit the config — the facts are untouched.

The Engine

Two models, one auto-switch.

Most questions are simple and should run on a fast, cheap model. The hard ones — a tricky comparison, a long reasoning chain — should escalate to a stronger model automatically, on conditions you define. You get speed and cost control on the 90%, and quality on the 10% that needs it.

You define the switch. Pick the conditions that promote a conversation to the stronger model; everything else stays on the fast one.
Non-Negotiables

The guardrails.

RuleNever invent facts or prices.
If it's not in the knowledge base, the assistant doesn't say it. No guessed prices, no made-up policies, no promises you didn't approve.
RuleDefer what isn't yours to answer.
Medical, legal, or anything high-stakes → point to a professional. The assistant knows the edge of its competence and stops there.
RuleA safety floor under everything.
A content filter sits beneath the whole assistant — the universal lines it will never cross, enforced before any reply leaves.
The Handoff

Warm leads, with the context already gathered.

When a conversation qualifies, the assistant collects the essentials — name, best contact, and a one-line summary — and notifies your team. The human picks up already knowing who they are and what they want. No "how can I help you?" from scratch.

RuleBe honest about the handoff.
The assistant captures and your team follows up — it shouldn't pretend a human is instantly on the line. For anything urgent, give a direct channel, like a phone or WhatsApp number.
Privacy

Keep customer data where it belongs.

Conversations and leads contain personal data. Handle it under your privacy obligations — in Malaysia, the PDPA. Keep only what you genuinely need (leads you're working, and de-identified data to improve the assistant), and never scatter customer data into places it doesn't belong, like a code repository.

Companion Resource

Where the assistant fits in a launch.

The chat assistant is the last step of launching a brand — it only works once the website, shop and knowledge base exist to feed it. The companion playbook covers that whole order.

Pair with this
The Brand Launch Playbook
Collect info → website → shop → knowledge base → chat assistant. Four pieces, five steps, one repeatable order for launching any brand.
Read the playbook →