AI has made me much faster at building things.
I can take an idea and have something working in a few hours. Not a slide showing what it might look like. Something I can actually use, test and change.
For a Product Owner, that feels like a big shift.
It also creates a tempting shortcut: if we can build the idea quickly, why spend so much time exploring the problem?
Because AI hasn’t made discovery less important. We can now build the wrong thing much faster.
A prototype is still a question
I experienced this while creating HitchedFlo, the wedding-planning app I built after becoming frustrated with information being scattered across spreadsheets, emails and notes.
With Claude Code, I created a first version in a few hours and started using it during my actual wedding planning.
AI shortened the path from assumption to evidence. But the prototype was still a question, not an answer. There were still plenty of things to fix, and the product naturally evolved as I used it.
A working product proves that something can be built. It doesn’t prove that the problem matters to other people, the solution is right or anyone will keep using it.
Once an idea has screens, buttons and a name, it starts to feel convincing. I named mine, designed it and started using it.
AI can speed up discovery
I already use AI throughout discovery in my role at Jetstar.
It helps me streamline questionnaires, summarise reporting and work through large datasets that would otherwise take much longer to analyse. It can surface patterns, suggest questions and help turn an early idea into something people can respond to.
That gives me more time to focus on what the information actually means.
But faster analysis isn’t the same as validated customer insight.
AI can identify a possible pattern. We still need to understand why it exists. It can help shape a hypothesis. We still need to test it.
That means speaking with real users, observing how they behave and running experiments such as A/B tests where appropriate. These methods give us evidence that generated summaries and prototypes cannot.
AI should help us get to meaningful testing sooner. It shouldn’t become a reason to skip it.
Prototyping is cheap. Production often isn’t.
Building my own apps gives me plenty of freedom. Working as a Digital Product Owner in a large organisation is naturally different.
A prototype only needs to demonstrate an experience. Turning that experience into a reliable product is another thing entirely.
Take the booking flow at Jetstar. What looks like a straightforward customer journey sits across a complex technology stack, with data moving between multiple systems and integrations. There are technical dependencies, operational impacts, established processes and competing priorities to navigate.
AI can help us explore an idea and make it tangible sooner. It can’t remove the complexity underneath it.
It can also miss edge cases.
A prototype may work perfectly for the obvious journey while overlooking what happens when an integration fails, information is missing or a customer does something unexpected. At scale, those less common scenarios still affect real people.
QA is also very different from clicking through a prototype and seeing the intended journey work once. A production product needs to work reliably across different devices, customer situations and connected systems. It needs monitoring, support and a plan for when something goes wrong.
That booking flow supports millions of dollars in revenue every day, so even a small mistake can be costly. If a customer can’t complete their booking, they may switch to another airline. If they can book but can’t add bags, that is still lost revenue. At this scale, a single decision can affect thousands of customers and have a significant commercial impact.
That places an even higher burden on discovery.
It isn’t enough to show that an idea can work. We need confidence that we understand the customer problem, the commercial impact and the risks before committing to delivery.
The MVP has changed
Traditionally, an MVP was often shaped by delivery constraints. Teams had limited capacity, so they chose the smallest version they could build and release.
AI loosens that constraint during discovery. We can create prototypes, compare approaches and test assumptions much faster.
But it doesn’t necessarily make production delivery cheap.
An MVP is therefore increasingly about restraint and taste, not just what a team has capacity to build.
When adding something to a prototype becomes easy, the harder question is whether it deserves to exist in the product. Does it address the user problem? Is it solving a genuine need, or are we adding it because we can?
Every feature still creates more for users to understand, more edge cases to consider and more technology to support. In a large organisation, it may also introduce dependencies well beyond the feature itself.
Cheap to prototype doesn’t mean cheap to deliver.
Faster learning, not just faster building
The biggest opportunity AI creates isn’t producing more features. It’s shortening the distance between an assumption and evidence.
We can analyse information sooner, explore more ideas and create better prototypes before committing significant development effort.
But learning still needs to be the goal.
Discovery still requires understanding customers, testing hypotheses, recognising weak evidence and navigating technical constraints. Delivery still requires careful development, QA and an honest understanding of how the product will behave in production.
When the customer, commercial and operational stakes are high, moving faster should mean gathering stronger evidence sooner, not lowering the bar.
AI can help us reach answers faster.
Product judgment makes sure we’re answering the right question, and building something that works beyond the prototype.