Cutting WhatsApp support cost with tiered AI triage
Client anonymised. Named references are available under NDA during due diligence.
The client operates in loyalty and customer engagement, running a consumer-facing programme that fields a steady stream of member enquiries over WhatsApp. As the customer base grew, so did the support workload. They engaged Canaan to absorb rising contact volume without letting support cost climb in lockstep with it.
The challenge
Every WhatsApp enquiry was handled manually by an agent. Roughly a fifth were repetitive FAQs — the same questions about receipts, points and devices, answered over and over. Because each contact needed a person, headcount scaled one-for-one with volume: more customers meant proportionally more support cost, with no way to break the link between the two.
The solution
Canaan deployed a tiered AI chatbot through Zendesk as a gatekeeper on the WhatsApp channel. It resolved the repetitive tier outright, without an agent. For everything it could not close, it pre-filled the ticket with the relevant receipt and device data before handing to a human. Agents stopped opening cold tickets and started each case with the context already in front of them.
The outcome
The repetitive tier was deflected before it ever reached an agent, resolved in seconds. That broke the one-for-one link between customer growth and support headcount — monthly salary cost stopped rising in step with volume. Agents were freed to concentrate on the enquiries that genuinely needed judgement, each ticket arriving already populated with the context they used to gather by hand.
Each figure is published with its calculation basis on request.
- 18%
- Reduction in monthly salary OPEX
- 18%
- Contact volume deflected
- ~30s
- Time to resolution on deflected contacts
Key takeaways
What carries across.
- Triage the repetitive tier first — a fifth of contacts rarely need a human at all.
- Use the bot as a gatekeeper, not a replacement: deflect what it can, enrich the rest.
- Pre-filling tickets with context removes the cold-start cost of every handover.
- Breaking the link between volume and headcount is what actually contains support cost.