How AI is reshaping product development, and what teams need to thrive

We share our working map of how AI is reshaping product work — from delivery to discovery to strategy — and what it means for teams that want to stay ahead. No hype, just a practical view of what’s emerging. One thing is clear: empowered teams are key to AI adoption.

Christophe Achouiantz · 2025-04-10 · 12 minutes · Product Model

The year AI changes product development?

2025 might be a turning point. Not because we’ve figured out how to “do AI,” but because organizations are now using it, at scale.

Meta’s Mark Zuckerberg talks about automating mid-level engineering. Salesforce’s Marc Benioff says their AI tools have boosted engineering productivity by 30%, enough to pause hiring. These statements are worth paying attention to. But whether they signal deep change or short-term optimization is still unclear.

Given that, let’s take a closer look at what AI may bring to product work based on current trends.

Generative AI is exploding across industries, yet we're still in the early adoption phase. In this article we attempt to connect emerging patterns and paint a possible development. It’s our guess on where product development is going. We’re still experimenting and learning, like everyone else, and tomorrow will bring new insights.

A simple lens: three types of work

To orient ourselves in this expanding field, we’ve found it helpful to break down AI’s role in product development using three categories:

  1. AI-Driven Work: repetitive, scalable tasks where AI can increasingly work on its own.
  2. AI-Augmented Work: tasks where humans stay in the lead, but AI boosts speed, breadth, or insight.
  3. Human-Led Work: work that requires ethics, judgment, creativity, or deep contextual understanding and where AI is used in a marginal role.

Three categories of AI work

We think that, for the near future, these categories fit well to the three main types of product work:

  • Product Delivery will become increasingly AI-Driven
  • Product Discovery will become increasingly AI-Augmented
  • Product Strategy will stay primarily Human-Led but with AI-Augmented research

We’ll walk through each of these, based on the trends we see and where we think it might go.

Delivery: becoming AI-driven

Of all the areas in product development, we believe that delivery is changing the fastest.

Tools like GitHub Copilot, Cursor, and custom LLMs are now a real part of the workflow for many engineers. As our former colleague Henrik Kniberg put it, “Tasks that used to take days now take hours, and hour-long tasks take minutes.”

This has a number of implications:

  • Faster release cycles. AI is helping automate not just coding, but testing, documentation, and deployment. Continuous delivery becomes easier, without the need of heroic effort.
  • Smaller teams. AI enables teams to get more development work done faster. Teams may become smaller, allowing them to move faster due to simpler coordination.
  • More cross-functional teams. AI enables teams to reduce the number of engineers so that other roles can be part of product teams, yet keeping the headcount low. This makes teams more capable and creative, able to take on larger problems and opportunities.
  • Lean cost models. Some companies, like Klarna, have replaced large delivery orgs with smaller AI-supported teams, keeping output steady while reducing cost. For companies in cost-sensitive markets, we expect to see the same amount of work done by smaller teams.

Product delivery will become increasingly AI-driven

This shift is still in motion, and uneven across industries. But we think it's likely that much of product delivery will become increasingly AI-driven, with human oversight focused on tough problems, edge cases, exception handling, and innovation.

Discovery: supercharged by AI-augmentation

Discovery, the ongoing work of understanding how best to solve problems and why a solution is worth pursuing, has always been time-consuming and hard to scale. We believe this is the area where we’ll see some of the most exciting (and practical) AI applications.

AI will not replace a team’s judgment during Discovery, but it will amplify their ability to see patterns, test ideas, and synthesize data across different inputs.

Here’s what we see is becoming possible:

  • Real-time insight generation. Teams will be able to process large amounts of qualitative data, like customer interviews, support logs, or reviews, and find emerging patterns in hours, not weeks. Note that current LLMs are not fully capable yet.
  • Synthesizing across sources. AI will help connect the dots between customer feedback, usage behavior, business performance, and even competitor activity.
  • Faster iteration and testing. Teams will use AI to generate prototypes, simulate flows, or predict behavioral reactions based on past data, helping them explore more ideas before building.
  • Shorter learning loops. When discovery becomes faster, teams will be able to validate assumptions earlier and more often.

AI will amplify product discovery

This will not make Discovery easy. But it will make it faster and broader, if teams are equipped to use these tools well. AI amplifies discovery capabilities, but only when teams have the discipline to guide it with clear intentions and apply rigorous quality control to its outputs.

Strategy: still a human-led game

Strategy remains, and will likely remain, a deeply human capability. Or at least the core activity in strategy: to evaluate and balance multiple ambiguous and contradictory needs within known and unknown constraints.

Sure, AI can help with market research and competitor analysis. Research that took months now takes but a few prompts. It’s like using satellites to create a map instead of sailing around in a frigate and drawing by hand. It speeds up and democratizes research, leveling the playing field by giving everyone the same advantage.

While AI excels at generating ideas, summarizing research, and simulating outcomes, it lacks the contextual awareness, ethical grounding, and intuition needed to make the kinds of decisions that determine long-term success.

Critical product strategy work will stay mainly human-led

That’s why we think strategy will stay mainly human-led, even in AI-native product organizations. AI can support, but it can't feel what matters. It lacks product sense.

The key human super power: Product Sense

At the core of all product work - regardless if it’s during strategy, discovery, or delivery - there is a key human super power needed to deliver a great product: product sense.

Product sense

Product sense is the ability to recognize what makes a product good, not just in theory, but in practice. It’s built through experience, exposure, and reflection. It’s what allows teams to:

  • Spot real user needs behind noisy data
  • Understand what will resonate in a given market
  • Recognize when something “feels off” before users complain
  • Navigate complex trade-offs between user value, business value, and feasibility
  • Understand how to implement a feature, present some data, making it look good

Product sense can’t be copied or queried, it has to be cultivated. It's what allows teams to make strategy not just rational, but relevant. It allows teams to discover the solution that fits real life, It allows teams to implement features that feel right.

As Steve Jobs put it in ‘The Lost Interview’:

“Designing a product is keeping 5,000 things in your brain and fitting them all together constantly in different ways to get what you want. Everyday you discover something new, a new problem or an opportunity to fit these things together a little differently. It's that process that is the magic! That's a team effort.”

Empowered teams matter more than ever

Product sense - the magic according to Steve Jobs - is a team effort.

But it must be teams that own outcomes, not just outputs. Teams that don’t wait for instructions. Teams that shape direction. We believe these empowered teams are best positioned to use AI wisely to create something great. But this only works if they have:

  • A strong product sense. Individually and collectively, the team needs to know what “good” looks like. This takes time, coaching, and exposure to real customers.
  • AI literacy. Teams need to know what the tools can do, where they help, and where they mislead. This isn’t about turning everyone into prompt engineers, it’s about integrating AI into how the team works.
  • Ethical awareness. AI increases leverage, but also risk. Teams need to think clearly about privacy, fairness, transparency, and accountability, especially when AI is acting on their behalf.

In short:

AI changes how we work, but not what good product teams are built on. Judgment, curiosity, empathy, and ownership are more important than ever.

What should organizations do?

There’s no blueprint. We’re all still learning. But here are a few early lessons:

✅ Give teams real mandates. Empowered teams only succeed when they own outcomes, not just delivery.

✅ Invest in product sense. Help teams see what “good” looks like, and give them exposure to the full product experience.

✅ Build AI capabilities into team workflows. Let teams experiment with different tools to find the right one for different tasks. Then, give these tools to product operations to facilitate adoption, creating playbooks, examples, etc.

✅ Keep your eyes on value, not just velocity. AI will make it easier than ever to move fast, but the real question is whether you’re solving the right problems.? In the end, what matters isn’t just how much you build, but whether you're building something worth building.