Journal / Product
How AI is changing product management
AI is not just a feature teams ship. It is reshaping how product teams are structured, how decisions get made, and who gets to make them. At Toggl, we restructured the product team to work in exactly this way, and the shift changed how we think about hiring, planning, and ownership.
Smaller teams, broader ownership
For years, product organisations scaled by adding headcount. More engineers, more designers, more PMs. AI changes that equation. A small, well-equipped team can now cover ground that previously required three times the people. Tooling handles the repetitive work: boilerplate code, first-draft copy, data queries. That means tribes and squads can shrink without losing output.
The line between designer, PM, and developer is blurring fast. Designers use tools like Cursor, Claude Coe or Codex to generate and iterate on real components. PMs write and run scripts to pull data or prototype logic. Developers spend less time on mechanical tasks and more on architecture and judgement calls. Role boundaries matter less than outcomes. With AI handling more of the execution layer, truly autonomous teams are closer than ever. A small team with the right tools and clear goals can now own a problem end to end, from insight to shipped solution, without depending on centralised functions at every step.
Problem briefs and PRDs are increasingly generated automatically, with team members reviewing and steering rather than writing from scratch. They still serve as communication tools for people, but their primary audience is shifting. More and more, these documents exist so that AI can use the description as a foundation for writing code. A well-structured PRD fed into a coding agent produces far better output than a vague prompt.
Customer insight belongs to everyone
Historically, user research required dedicated researchers, tooling access, and time. CRM data sat in sales. Support tickets lived in Zendesk. Session replays were watched by a handful of people. AI changes the access model. Engineers, designers, and PMs can now query customer feedback directly, surface themes from hundreds of support conversations, and watch replays without routing everything through a research team. Insight becomes ambient rather than gated. The whole team has context, not just the PM who attended the last round of interviews.
Goals over sprints
Sprints made sense when work was predictable and planning horizons were short. Two-week cycles imposed a useful rhythm but also a lot of overhead: planning, retros, backlog grooming. That overhead consumes time without always driving better outcomes. With AI in the workflow, iteration speed increases. The bottleneck shifts from execution to direction. What matters more is having a clear goal, for a two-week period, a month, or a quarter, than adhering to a sprint ceremony. Teams that orient around outcomes and move fluidly will outpace those optimising the sprint ritual.
The automated product loop
A/B testing used to be the growth team's territory. Getting a test designed, instrumented, and analysed required tooling access, statistical knowledge, and enough traffic to reach significance quickly. AI lowers all three barriers. If traffic supports it, every product team should be running their own experiments rather than queuing behind a central growth team.
When traffic is too low for A/B tests to reach significance in a reasonable timeframe, the answer is not to wait. It is to talk to customers. Qualitative feedback from ten to twenty users can surface the same signal a quantitative test would have confirmed in three months. AI helps here too: summarising interview transcripts, tagging themes across calls, and identifying patterns across sparse data.
The logical endpoint of all of this is a closed loop: customer issues surface automatically, possible solutions get generated and ranked, experiments run and report results, and the best option gets implemented. Team members set direction and make final calls rather than doing the mechanical work at each step. That loop is not fully realised yet, but the components exist. Teams that start building towards it now will have a significant structural advantage over those still running the process manually.
These are still early days and the patterns are shifting quickly. If you are seeing this play out differently in your team, or have thoughts on where product management is heading, feel free to share. Feedback and different perspectives are always welcome.
