How ikeja used AI-powered WhatsApp interviews to capture authentic customer feedback at scale across South African townships - turning a 6% response rate into 45%.
How ikeja used AI-powered WhatsApp interviews to capture authentic customer feedback at scale across South African townships — replacing costly telephonic outreach with always-on, voice-note-rich conversations.
ikeja is a growing Internet Service Provider serving South African townships with its Wave product. As the customer base scaled, the team needed more than satisfaction scores — they needed to understand how customers actually experienced their internet service, what drove upgrades, and where the product was falling short.
The traditional approach — telephonic outreach during business hours — was hitting a wall. Customers in township communities weren't available when agents called, email surveys barely registered, and the conversations that did happen were too rushed to surface anything useful. ikeja needed a way to capture the full voice of their customer base, not just the fraction who happened to pick up the phone.
Running customer feedback at scale in township markets presents a distinct set of operational barriers. The customers are reachable — but not through conventional research channels.
Telephonic outreach was limited to business hours, exactly when township customers were least available. Response windows were narrow and frequently missed entirely.
Rushed phone conversations yielded one-line answers. There was no way to probe deeper into specific pain points — WiFi range issues, reliability during weather, or multi-device needs.
Email survey response rates sat around 6% in communities with limited email access. The standard digital feedback loop simply didn't reach these customers.
Without structured data across hundreds of customer interactions, ikeja couldn't identify systemic patterns — which areas had range issues, which competitors customers were switching from, or what drove word-of-mouth.
Yazi deployed an automated AI interviewer via WhatsApp, triggered 7 days after each Wave installation. The timing was deliberate — long enough for customers to fully experience the service, recent enough for vivid recall.
Each new Wave customer received a WhatsApp message 7 days after installation, opening a conversational AI interview that adapted dynamically to their individual responses.
With 98% WhatsApp penetration in South African townships, the channel removed every barrier. No app to download, no portal to navigate, no phone call to schedule. Customers simply replied when it suited them — and many chose to send voice notes, sharing richer detail than any phone survey could extract.
ikeja's WhatsApp feedback channel didn't just outperform email — it created a depth of engagement that telephonic outreach had never achieved. Customers who would have been unreachable by phone spent an average of 16 minutes in conversation with the AI interviewer.
With 2,000+ structured interviews, ikeja gained visibility into patterns that would have taken years of ad hoc phone calls to uncover. The data delivered clear, actionable intelligence across product development, marketing, and customer success.
ikeja's team accessed three core views: a Media Library with voice notes and text responses, per-customer Interview Transcripts with the AI's adaptive follow-ups, and a Participant Journey stitching the full feedback experience together.
| Telephonic / Email Surveys | Yazi AI WhatsApp Interviewer | |
|---|---|---|
| Availability | Business hours only, customers often unavailable | 24/7 — customers respond at their convenience |
| Response rate | ~6% email, low phone pickup in townships | 45% response rate on WhatsApp |
| Depth | Rushed conversations, surface-level answers | 16-minute avg. conversations with voice notes |
| Probing | Same script for every customer | AI adapts follow-ups to each individual response |
| Voice capture | Manual transcription, if recorded at all | 60% voice notes, auto-transcribed and aligned |
| Scale | Limited by agent headcount and call hours | 2,000+ interviews with zero manual agent costs |
ikeja received a structured, ready-to-analyse data package that turned raw customer conversations into actionable product, marketing, and operational intelligence.
ikeja identified specific WiFi range issues that led to hardware recommendations, discovered weather resilience as a key competitive differentiator, and flagged installation delays in real time for process improvements. Customer service gaps were identified and fed directly into training interventions.
The data revealed that word-of-mouth and sales agents drove nearly equal acquisition — reshaping resource allocation. Speed emerged as the top upgrade driver at 47%, informing messaging strategy. Authentic customer language from voice notes became source material for marketing that sounds like real customers, because it is.
The beauty of it... it just works at the time that the customer is available and at their leisure. Versus, if you're employing someone to do that internally, it's usually during office hours and people aren't really available.
We recently partnered with Yazi on an implementation on their new platform with their AI interviewer. And the results were pretty groundbreaking for us. This AI interviewer was able to steer conversations to places where I don't think we'd get usually on either our telephonic reach outs... The results were quite fascinating and we could see customers' intent around the specific questions that we were asking.
Internet service providers in emerging markets face a fundamental paradox: the customers they most need feedback from are the hardest to reach through traditional channels. Email doesn't land. Phone calls don't connect. And when conversations do happen, they're too brief to surface the nuance that actually drives product decisions.
By meeting customers on WhatsApp — the platform they already use daily — with an AI interviewer that probes as deeply as a good researcher would, ikeja turned a 6% response rate into 45%, a one-minute phone script into a 16-minute conversation, and scattered anecdotal feedback into structured intelligence across 2,000+ customers. The result wasn't just more data — it was data that could actually shape hardware decisions, marketing strategy, and customer success interventions.
Interested in running a similar study?
Book a Demo →
%202.png)