By Horst Heinze, Head of CX and Product Strategy, Inovo
About six months ago, I was working at a hot scale-up in California — the kind of place where ideas move faster than slide decks and every problem feels urgent. One afternoon, our CEO knocked on my door and said, “We need you to reimagine our entire ordering journey — but make it AI-powered.”
This is the story of what unfolded after that knock — and what I learned about turning big AI promises into a real, working customer experience. No hype — just the hard truths.
The first thing I did was pull together what I still call my “mean team.” Strategists, designers, developers — people who were top of their game in their craft. But here’s the kicker: nobody was a true AI expert. And that’s more normal than most people admit.
AI is still new. The playbooks are thin. So we needed people who weren’t afraid to admit what they didn’t know — and who were hungry to figure it out. The best thing we did was build a team that could think across strategy, design, and tech — and who were comfortable exploring the unknown together.
Our CEO had strong ties into the Y-Combinator network and lined up deep dives with some of the hottest AI startups on the planet. Those sessions were humbling. Spoiler: even the people building frontier AI tools are figuring it out in real time.
Unlike designing a standard checkout flow — where you follow well-worn UX patterns — creating new AI-powered experiences means you’re pioneering. There’s no manual. That’s why our starting point wasn’t a polished strategy doc — it was messy, hands-on experimentation.
If you want to understand what AI can actually do for your customers, you have to get your hands dirty early. Build, test, break it, and learn fast.
Normally, I approach any new experience through three lenses: desirability (does it solve a real customer need?), viability (does it make commercial sense?), and feasibility (can we build it?).
With AI, feasibility flips everything. If you don’t know what the tech is capable of — or where it will fail — how can you design something delightful?
We ran dozens of small experiments to map where AI was genuinely useful — and where it wasn’t. Only then did we focus on desirability. We leaned into human-centred design, mapping real customer pain points, mocking up rough front-end prototypes, and putting those concepts in front of actual users.
The truth is, customers don’t care that it’s AI under the hood. They care that it works, that it solves a problem, and that it feels intuitive. AI for AI’s sake is just expensive noise.
This part hit us hard. Once we started using our AI prototype ourselves, the cracks showed. Often, the AI didn’t respond the way we wanted. I’d sit there playing the role of the agent in my head — If I were the AI, I’d answer like this. But the model would answer differently.
That’s when it clicked: AI has its own agency. You don’t ‘hard-code’ it like a simple script — you guide it, but it still has freedom within constraints. And that freedom is both powerful and risky.
A self-service or autopilot AI will always produce outputs that can vary. So you need to govern it, just like you’d manage an employee. How do you make sure it doesn’t go rogue when you update something? What happens when the underlying LLM changes? It’s a genie in a bottle — but that genie needs constant oversight.
Managing an AI experience isn’t one-and-done. It’s continuous. We started running vibe tests — playing with different prompts and scenarios to see if the AI “felt right” in its responses. Did the tone match our brand? Did it understand context? Any odd logic or cringe-worthy hallucinations?
We layered on evals — automated tests that give the AI input data and measure outputs against clear criteria. Every time we updated the model’s knowledge base — its ‘brain’ — we’d run hundreds of these tests to make sure we didn’t break something else.
And we learned quickly: accuracy is only half the battle. Latency can kill trust too. Customers won’t wait forever for a response — especially when they’re expecting instant, human-like help. We had to get ruthless about tightening prompts, choosing the right LLMs, and optimising for speed without sacrificing quality.
When we finally rolled out our new AI journey — alpha, then beta, then scale — it felt like crossing the finish line. But really, it was just the starting line.
AI isn’t static. It learns, drifts, and sometimes surprises you. Your business changes. Your data evolves. New models emerge. Each of these can knock your AI’s performance off track — or worse, introduce risks you don’t see coming.
If you want AI to work for your customers in the real world, you have to treat it like a living system. You can’t set and forget it. You have to keep testing, governing, optimising — and always asking: Is this still right?
If there’s one thing I’d say to anyone about to start an AI project, it’s this: nobody really knows everything. The hype is huge — but the real work happens in the trenches. It’s about experimenting, failing fast, staying close to your customers, and chasing accuracy day after day.
At Inovo, we live this reality alongside our clients. We bring platforms that work, people who’ve learned the hard lessons, and the discipline to turn bold AI ideas into practical, orchestrated experiences that deliver real outcomes — not just slides and promises.
Because AI is hype — until you make it real.
Horst Heinze, Head of CX and Product Strategy, Inovo
Horst Heinze is the Head of CX and Product Strategy at Inovo. He’s spent his career helping organisations bridge the gap between bold ideas and real-world impact — from building AI-powered customer journeys to leading teams that make experience design, strategy, and technology work together seamlessly. He’s passionate about turning hype into outcomes and helping businesses across Africa deliver customer experiences that truly build trust.
Feel free to connect with Horst on LinkedIn or email hheinze@inovo.co.za if you’d like to chat about how to develop AI-powered experiences that actually work in the real world.
Contact Us
VUKA is the trusted media partner to key professionals, policy makers, suppliers and
manufacturers. We provide unparalleled opportunities for industry-wide connection.
