Automation / AI / Business system
AI Enquiry Agent
An automated lead-handling system that reads a customer enquiry, collects what is missing, creates a structured lead, ranks it, alerts the owner and schedules the follow-up.
Problem
The enquiry was not the problem. Everything after it was.
Small businesses receive messages from a website form, Instagram, WhatsApp and walk-in referrals. Getting the enquiry is rarely the bottleneck — most owners get more than they can process.
The cost sits in what happens next. Someone reads the message, works out what the customer actually wants, asks the same three or four questions they asked yesterday, waits for a reply, copies the details into a spreadsheet, decides whether the lead is worth chasing, tells whoever needs to know, and then tries to remember to follow up in two days.
Every one of those steps is repetitive, and every one of them is a place where a lead quietly disappears.
Repeated for every enquiry
- Read
- Understand
- Ask
- Collect
- Record
- Notify
- Follow up
Approach
Structure the message first. Then the rest becomes rules.
- Unstructured message
- Intent + information detection
- Qualification
- Structured lead
- Automation actions
The design decision was to stop treating an enquiry as a conversation to be managed, and start treating it as a record to be completed. Once a message becomes a structured lead, every step after it is a rule rather than a judgement call.
The processor reads the message for intent, business type, requested features, budget and timeline — in Malay or English, since that is how people actually write. It then compares what it found against what a usable lead needs, and asks only for the gaps.
Qualification uses a transparent points rule rather than a hidden model. The owner can always see why a lead scored what it did, which matters more than the score being clever.
System
Six components, one direction of flow.
Customer
Website form, Instagram, WhatsApp
Enquiry interface
Embedded assistant · quick choices + free text
Enquiry processor
Intent, business type, features, budget, timeline
Qualification engine
Missing-field check · Lead Priority Rule
Lead store
Structured lead record · status · history
Owner notification
Hot leads surface immediately
Follow-up scheduler
Booked on the customer's channel
Each stage is a service behind an interface. The portfolio build ships the demo implementations, which run entirely in the browser and need no API keys. Production adapters implement the same methods.
- Interface
- Next.js App Router · TypeScript · Tailwind CSS · Framer Motion
- State
- Zustand with localStorage persistence
- Processing
- Deterministic keyword and pattern rules — no API key required
- Architecture
- Service interfaces with swappable demo and production adapters
Swapping DemoEnquiryProcessor for an LLM-backed processor, or DemoLeadService for a database-backed one, is a change in one registry file.
Result
What changes for the owner.
- Enquiries become structured leads
- Missing information is identified earlier
- Important leads become visible
- Owner actions become easier to prioritise
- Follow-ups are less dependent on memory
This portfolio implementation runs in Demo Mode. Real messaging, CRM and AI integrations can be connected through the same workflow architecture. No performance claims, ROI figures or client numbers are presented here — the demo is the evidence.