WHATSAPP AI LAYER
A 30-second reply engine on your WhatsApp.
The AI response engine we built for Small Talks is running live — and now it can run for your business. Not a chatbot product — a real response infrastructure: knowledge base, flows, escalation, learning loop, all ready.
LIVE NUMBERS
Running live at Small Talks.
This engine is used by a 22-person team. Thousands of messages handled automatically every month. The numbers aren't estimates — they're live data.
2,342
LINES OF ENGINE CODE
3,041
WP MESSAGES HANDLED
570
ACTIVE CONVERSATIONS
59
KNOWLEDGE BASE ENTRIES
WHAT'S INSIDE
Not a classic chatbot. A six-piece system.
Every piece does its own job, all of them live in one database. Nothing is missing — whatever runs at Small Talks runs for you.
AI First Response
Every incoming message gets a personalized reply within 30 seconds. Active 24/7. OpenAI GPT-4o with Anthropic Claude fallback.
Knowledge Base
50-100 sector-specific Q&A entries. Admin CRUD, categories, tags. Learning loop grows the KB over time.
Conversation Flows
Multi-step scenarios (booking, membership queries, payment tracking). Button-first UI, conditional branching, form capture.
Escalation System
When the bot can't answer, it hands off to a human. Claim system in the message center — two people can't reply to the same thread.
Approved Templates
WhatsApp Business templates already through Meta review (reminders, welcome, winback). City visuals included.
Learning Loop
When the bot can't answer, the question is auto-logged. Admin adds it to the KB with one click. The system gets smarter every week.
INSIDE THE ENGINE
The decision flow.
When a message comes in, the engine runs four steps. Low confidence triggers an auto-handoff. KB hit returns directly. Complex questions get the full context injected into GPT.
01
Intent Classification
Incoming message is classified: info question, booking, complaint, purchase.
02
KB Lookup + Similarity
Knowledge base is searched. If confidence > 0.55, the answer returns directly.
03
Flow Match
If inside an active flow (e.g. booking), the next step is sent automatically.
04
GPT Context Injection
If needed, GPT gets city events, product info, user history as context.
05
Escalate or Answer
If confidence is still low, conversation hands off to a human in the Message Center.
THE COMPARISON
WhatsApp AI module, or the full System?
WhatsApp AI is one piece of jo.works/system. An entry point. If you need the full System, let's scope a bigger installation.
Start with the module, upgrade to Full System as you grow.
WHO IT'S FOR
Who do we build it for?
Service businesses receiving 500+ WhatsApp messages a month where response speed directly drives sales. We tune KB and flows to each sector.
THE PROCESS
Live in three steps.
Not a massive project — a tight rollout. The Small Talks engine is ready; we adapt it to your business.
01
· 3-5 daysBrief + Audit
We map your current WhatsApp flow, the most frequent questions, your current response time, and the leaks.
02
· 1-2 weeksInstall + KB Authoring
Engine is installed, a custom knowledge base is authored, sector-appropriate flows are designed. Meta template review starts.
03
· OngoingGo Live + Optimization
System goes live. First 30 days of close observation, KB additions, flow tuning. Every week gets sharper.
The engine runs every day at Small Talks.
Turkey's largest English-speaking community runs 6 cities with 22 staff on this engine. 2,342 lines of code. 570 active conversations. Not a demo — live, measurable proof.
READ THE SMALL TALKS CASE STUDY→GET STARTED
Is your response layer lagging?
Start with a free short call. We'll look at your current flow and show you what the module can move for you.