The outcome, the five users and the engagement signal that tells you whether to build or kill the feature. Building AI features has never been faster, but the question that gets skipped: would anyone use the outcome this AI feature produces?
To know if you need an AI feature, we got Hilde Franzsen, Brand and Marketing Director at Stellenbosch UX and brand agency Inkblot, to show us how to get sure
The move: be the AI before you build it
It’s called the Wizard of Oz test. The user thinks the AI is doing the work, but actually, it’s just you (a human) doing it by hand in the background.
“Before you write a single line of code, do what the AI would do manually for five users. Send the recommendation yourself. Write the summary yourself. If a user doesn’t engage when a human delivers it, they’re not necessarily going to engage better when a machine delivers it.”
How to know if you need an AI feature
1. Name the outcome the AI would deliver, not the feature
“AI chatbot” is a feature. “An answer to their tax question in under a minute” is an outcome. Users care about the outcomes, so focus on that: If you can’t describe the outcome in one sentence, the feature isn’t ready to test, let alone build.
2. Find five users who actually need that outcome
Not friends or your team, but real target users with the real problem. Five is enough. You’re looking for a binary signal (do they want the outcome or don’t they), not a statistically significant sample.
3. Deliver the outcome yourself
This is the Wizard of Oz move. Pretend you’re the AI: Answer their questions yourself, write the summary, send the recommendation – whatever they ask for.
The user doesn’t need to know its manual; all you’re testing is whether they engage with and want the outcome.
4. Read engagement, not politeness
Engagement is when they come back asking for more, act on it or tell someone else. That’s real demand for the outcome. Saying “thanks, that’s interesting” means you can discard it.
If four out of five users really engage with the human version, you can build the AI feature. If they don’t, you just saved yourself many hours of work that no one wanted anyway.
The big payoff
Building the AI feature is the expensive part. The Wizard of Oz test takes an afternoon. The cheapest test is always the one you do before you build.
Want the full playbook?
This test is one piece of From Solutioneering to UX: How to Prioritise Designing the Right Things, Hilde’s full masterclass inside the Founder Collab, going deep on how to stop building the wrong things, including:
How to write a value hypothesis that catches solutioneering before it starts
How to find your product’s reason for existence
How to turn feature ideas into testable hypotheses so opinions stop winning over evidence
The full case studies on Snapchat, McDonald’s AI drive-thru, Duolingo and Google Glass — and what each one teaches about getting UX wrong or right
The closing checklist Hilde runs through before briefing any developer, AI tool, or designer
You’ll also get access to 40+ other masterclasses from SA founders and operators on sales, fundraising, paid media, automations and more inside The Founder Collab.
Get more SA tech and business news, tips and business-building workflows and subscribe to The Open Letter.



