Building fast is no longer the challenge – building the right thing, fast, is.
In our webinar on April 23, Levi9 Delivery Managers –Nevena Majstorovic Atanasovski, Tamara Djenadic, and Boban Poznanovic explored what product engineering looks like in the age of AI: where to slow down, how to measure what matters, and why human judgment remains at the center of it all. The key takeaway: AI is not the strategy. AI is the accelerator – and humans decide the direction.
The Acceleration Trap
AI has dramatically lowered the cost of building. You can prototype in hours, run discovery in days, and ship features at a pace that would have been unimaginable two years ago. That’s genuinely powerful – but it comes with a risk that’s easy to miss when everyone’s excited about the demo.
Speed can create pressure to skip the questions that still matter the most: Why are we building this? For whom? What problem are we actually solving? Those questions haven’t changed. What’s changed is how fast we can answer them.
And here lies the acceleration trap – building more before validating sufficiently. Prototypes are a great foundation for conversation with stakeholders – not a substitute for it. The human in the loop must not be considered as bottleneck. Removing conversation out of the picture is what represents the risk and the trap.
What Good Product Work Looks Like Now
The fastest phase of any product lifecycle is delivery. Where we shift our focus now is discovery. The real opportunity for us is to take time that we saved in delivery and reinvest it into learning and discovery.
Tamara shared a concrete example from a recent project. A customer came in with a working prototype – vibe-coded, but functional. Using AI tools, the team quickly extracted the embedded business logic and rules. What they found out by analyzing the identified problem and business logic is that it didn’t seem like the most critical one.
That signaled the team that right path is to get back to basics: organize user interviews, use AI to syntheses and analyze the inputs to come up to the real problem statement. Using AI tools and the right data, team was able to create comprehensive PRD and basic user flows in two weeks time. What would require six to eight weeks, couple of years ago, now could be done in the fraction of the time, having the same, if not more quality outputs. AI handled tasks were human used to invest a lot of time, freeing the team to focus on confirming, challenging and fine tuning the result.
The outcome: a cleaner handover to design, fewer rounds of rework, and changes that used to require significant effort now applied in minutes. Discovery can be done right and fast – but only if the right questions are asked first.
Delivery, Governance, and the Right Kind of Constraints
Faster discovery means more frequent, more complete specs arriving at the delivery team’s door. Two-week sprints still exist – but the expectation of what ships in two weeks has fundamentally changed.
Boban’s point on governance cuts through a common misconception: it doesn’t need to be reinvented for the AI era, and it shouldn’t become more rigid in response to it. The foundations are already there – how teams are structured, how stakeholders stay engaged, what quality means. What AI changes is how efficiently teams can work within those foundations.
Governance isn’t more bureaucracy. It’s the guardrail that makes going fast sustainable. Regulation, data privacy requirements, quality standards – these existed before AI and they remain relevant. What Levi9 brings to client collaborations is a setup that holds both sides of the equation: the business thinking on what needs to be achieved, and the process discipline on how to get there.
What's Changing for People and Teams
Most evident change is that existing roles are changing and merging. The delivery managers of two years ago spent most of their time tracking progress, removing blockers, keeping things aligned. Much of that is now automated. What that role looks like today – what Tamara called an “Outcome Lead” or “Value Delivery Lead” – is someone who asks more questions, spots what doesn’t add up, and brings delivery experience and knowledge into conversation with customers.
The same shift is happening on the technical side. Engineers are reinvesting their time into better understanding the business problems but also bringing their perspective into conversations about problems we are solving. The lines that once clearly divided the responsibilities of the roles are no longer there. We must all evolve, bringing our unique knowledge into business discussions.
What’s not changing is where human judgment lives. AI can challenge your logic, synthesize information, and surface patterns. It can’t bring the experience of working with a specific customer, navigating a specific domain, or reading the room in a difficult stakeholder conversation. That stays with the people.
What Winning on Repeat Looks Like
The webinar closed with a simple framework for what it takes to keep delivering value in this environment:
- Fast learning, not just faster delivery. Build, collect feedback, adapt, repeat. The speed advantage is wasted if it doesn't feed back into better decisions.
- Clear metrics, not just more activity. Output metrics still matter – timelines, velocity, cycle time. But they need to sit alongside outcome metrics: revenue impact, customer satisfaction, cost reduction, quality in production. Both tell part of the story.
- Smarter governance, not heavier processes. Frameworks and standards are a foundation, not a constraint. Use them to stay on course while moving quickly.
- Evolving leadership, not just new tools. The more important question isn't how fast the tools let you go. It's what you want to achieve with them – and that's a question for people, not AI.
Teams that win won’t be the ones that shipped the most. They’ll be the ones that learned, adapted, and delivered well – again and again.
In this article:
