Lead qualification
Greets an enquiry in seconds, asks the three questions you'd ask, captures name, area, budget and urgency, and hands a complete lead to a person. On WhatsApp, on the website, or both.
AI that does real work in your business, not a chatbot that says "great question"
Assistants built on large language models (Claude, OpenAI), grounded in your own documents and prices, connected to WhatsApp, your website, email and your CRM. They qualify leads, answer the questions you answer every day, draft quotes and summarise calls. A human stays in the loop.
What we build
Five jobs an LLM can take off your plate this month
FAQ and after-hours answers
Grounded in your price list, policies and service pages, so it answers accurately and says "let me get a person" when it isn't sure. No invented prices.
Quoting assistant
Turns a rough enquiry into a structured quote request or a draft quote from your rate card, ready for you to approve and send.
Call and message summaries
Transcribes calls and long WhatsApp threads, pulls out what the customer wants, and logs a clean summary against the lead in your CRM.
Document extraction
Reads IDs, payslips, invoices, application forms or plans, extracts the fields you need and files them, with a person checking anything uncertain.
Internal copilots
A private assistant over your own SOPs, product sheets and past quotes, so staff get answers from your knowledge, not the internet's.
How it's built
Production-grade, not a demo
The same architecture we run for large lead-generation operations, sized for a small business.
Models
Claude and OpenAI models via their APIs, chosen per task for cost and accuracy. Swappable later without rebuilding the rest.
Grounded in your data
Retrieval over your documents, prices and pages (RAG), so answers cite your material. Guardrails on what it may promise.
Channels
WhatsApp Business API, website chat, email and voice. One assistant, one memory of the customer, whichever channel they use.
Integrations
Leads, summaries and documents pushed to Bitrix24, HubSpot, Zoho, Google Sheets or any system with an API, via webhooks or direct calls.
Hosting
Deployed on AWS (Lambda, API Gateway, S3, DynamoDB) or Cloudflare Workers, in your own account where possible. You own the keys.
Monitoring
Every conversation logged, every hand-off tracked, alerts when the assistant is unsure or a lead waits too long. A dashboard shows what it's handling.
Claude APIOpenAI APIRAGWhatsApp Business APIAWS LambdaAPI GatewayS3DynamoDBCloudflare WorkersPythonNode.jsWebhooks
Pricing
All pricingStart with a plan, not a bill
AI projects go wrong when they start with the tool. Ours start with a Discovery Session and a written plan.
AI & Automation Setup
From R12 500
Automate lead handling, enquiries and everyday admin.
- Automation plan and priorities
- Website chat or WhatsApp assistant flow
- CRM or spreadsheet connection
- AI response structure and guardrails
- Testing and handover
Monthly Digital Support
From R20 000 / month
A retainer across website, campaigns, design and reporting.
- Agreed monthly hours
- Website updates and campaign management
- SEO improvements
- Reporting and dashboard upkeep
- Design and automation improvements
Model usage (API tokens) and WhatsApp conversation fees are paid to the provider at cost, in your own account. We tell you the expected monthly run cost before you commit; for most small businesses it is a few hundred rand a month.
Questions
AI FAQs
Will it make things up?
Not about your prices or promises. Answers are grounded in your documents, the assistant is told what it may and may not commit to, and anything outside that gets handed to a person. We test it against real enquiries before it goes live.
Where does my data go?
To the model provider's API for processing (Claude or OpenAI, under their business terms, no training on your data) and to your own systems. Hosting is in your AWS or Cloudflare account where possible, so you can revoke our access any time.
Do I need a CRM first?
No. Many clients start with WhatsApp and a Google Sheet. The assistant can write there, and move to a CRM when the volume justifies it.
How long does it take?
A single-channel assistant with a defined job (say, WhatsApp lead qualification) is typically live in two to four weeks after the Discovery Session. Multi-channel setups with CRM integration take longer and are scoped in the plan.
What does it cost to run?
Model usage is metered per conversation and is cheap at small-business volumes. WhatsApp charges per conversation window. We show you both estimates in the plan and set alerts so there are no surprises.
Tell us the job you'd hand to a good assistant
Qualifying leads, answering the same ten questions, reading forms. We'll say whether an LLM should do it, and what it would cost.