What We Were Solving
How We Built It
Custom AI chatbot built on OpenAI with RAG, integrated with WhatsApp Business API and HubSpot CRM, trained on six months of real enquiry data.What They Needed
The client is a Johannesburg-based B2B professional services firm with around 400 active clients. Their three-person customer service team handled everything: service enquiries, onboarding questions, document requests, scheduling, billing queries, and general support. They were receiving hundreds of inbound enquiries per month across email, WhatsApp, and their website contact form. Response times were slow during business hours and non-existent after hours and on weekends.
Client satisfaction was slipping, dragged down by slow response times and the need to repeat information across channels. They couldn't justify hiring another support person. They needed a way to handle routine enquiries immediately while freeing the human team for complex work.
What We Built
We built a custom AI chatbot integrated with their website, WhatsApp Business API, and their existing HubSpot CRM. This wasn't a template chatbot with scripted responses. It was a context-aware AI assistant trained on the client's specific knowledge base.
We started by auditing six months of customer enquiries and categorising them by type and frequency. The vast majority were routine, low-to-medium complexity questions, which made them perfect candidates for AI handling. We built a knowledge base of 240 question-answer pairs covering every common enquiry, product detail, process explanation, and policy.
The chatbot was built using OpenAI's API with retrieval-augmented generation. It understands context within a conversation, connects to HubSpot for client data, triggers automated workflows for document requests, links to Calendly for scheduling, and hands off to human agents with full conversation context when needed.
Getting It Right Over Time
Since over 60% of inbound enquiries came via WhatsApp, we integrated the chatbot with WhatsApp Business API so clients got the same experience in the app they already use. We then monitored every conversation for three months, expanding the knowledge base and adding new automated workflows as we spotted gaps.
How It Turned Out
The chatbot now handles the majority of monthly enquiries without any human involvement. The support team spends significantly less time on routine questions and redirects that capacity to revenue-generating activities. Response times dropped dramatically and client satisfaction scores improved. The investment paid for itself within the first few months.