By Proud Dzambukira, Group Head of Product, Watu
A few years ago at Meta, I watched what happens when a platform becomes part of people’s daily lives. The rules change. People stop comparing your service to your competitors and start comparing it to the best experience they had that day, on any app, in any category. Immediacy becomes the baseline. Context becomes an expectation. Friction becomes unforgivable.
That lesson has stayed with me as I now lead product at Watu across several markets in Africa and LATAM, because it explains the gap every customer experience operator lives with. Expectations are rising faster than our operating models can change. Customers want instant answers, in the channels they already use, from a company that knows who they are and does not make them repeat themselves. They do not care that we are managing legacy systems, fragmented data, regulatory constraints and cost pressure. They compare us to WhatsApp and M-Pesa, not to the lender down the road.
That is where many AI conversations go wrong: we treat AI as the transformation, when it really only amplifies the system around it. AI is not the strategy. System design is the strategy.
AI is powerful. It can classify intent, translate, summarise, route, score quality and resolve simple needs in seconds. But it does not create outcomes on its own. It creates outcomes only when it is built into the entire service system: the channel, the identity layer, the workflow, the escalation path, the knowledge base, the quality checks, and the product itself. Without that, AI becomes theatre: a chatbot here, a pilot there, a polished demo that leaves the real economics of service untouched.
At Watu, this is not abstract. We finance opportunity-seekers across several African markets, and for most of them, the phone is the bank branch, the support desk, the repayment reminder, and the main access point to economic activity. That changes what customer experience even is. It is not a department at the edge of the business, or a call centre that cleans up after the product fails. It is the operating system through which the customer experiences the company. And at scale the question becomes unforgiving: how do you serve millions of people with speed and dignity without letting the cost of serving them rise in lockstep with their number?
That is where AI earns its place, but only when we stop asking “what can we automate?” and start asking “what outcome are we trying to improve?” A customer checking a balance does not need a person. A customer asking how to pay may not. A customer whose account status has changed needs automation plus a clear explanation. A distressed customer with a complex problem needs a human being. A suspected fraud case needs a different workflow entirely. AI is useful when it helps us make those distinctions faster and more consistently, not when it is pointed at everything indiscriminately.
The lesson I carried from Meta is that scale amplifies design. In a small system, good people can paper over weak foundations. At scale, they cannot. Weak knowledge bases become wrong answers at volume. Poor integration becomes repeated customer frustration. Unclear ownership becomes backlog. AI does not hide these flaws. It exposes them. That exposure is uncomfortable, but it is also where the real transformation begins.
That is the shift I would put above all others: every customer interaction is a product signal.. If thousands of people call to check a balance, the balance is not visible enough. If they call to confirm a payment, the confirmation is not trusted enough. The point of reading those signals is not to automate the complaint. It is to remove the reason for it. Treated this way, customer experience stops being a cost centre and becomes one of the richest sources of product intelligence in the business.
In African markets, this is not just a philosophy. It is a necessity. We cannot assume unlimited data, five apps per customer, a single shared language, or that voice-only support will stay affordable. Those are not weaknesses. They are design inputs. They force us to build systems that are mobile-native, conversational, and product-led from the start, rather than slowly digitising service models built for elsewhere.
So the crossroads is not AI versus humans. The real question is whether we use AI to build customer systems worthy of trust: fast when the need is simple, careful when the stakes are high, transparent when the answer matters, accountable when the system fails, and intelligent enough to keep improving.
At Meta, I learned that scaled systems shape behaviour. At Watu, I have learned that in African markets, trust decides whether that behaviour endures. Our job is not to build AI that sounds human. It is to build systems that act humanely.
Read Also: MultiChoice Launches Customer-Centric Everything Can Wait Campaign Ahead Of FIFA World Cup 2026