What Every CPO Needs to Understand About AI Right Now
- Jul 16
- 4 min read
Updated: Jul 27
Written by Marina Shulga
When people talk about AI inside product teams, the focus almost always lands on engineers and designers. The more consequential change is happening one level above them. The pace at which a product organisation moves, the questions it can answer without specialists in the loop, and the kind of judgement it now demands from leadership, all of that is being rewritten right now. A CPO whose tooling literacy stopped in 2023 has already missed a generation of capability. Multilingual content, synthetic avatars, and instant translation all looked remarkable two years ago and have become the baseline. The work that earns a product leader their seat now happens in the decisions about where AI enters the workflow, which handoffs collapse, and what the team is freed up to attempt instead.
I've spent the last several years across EdTech and fintech product organisations, and what I'm watching is the role being redrawn through three operational shifts and one strategic consequence that's starting to separate teams that move from teams that watch.
The role itself has changed shape
The differences between teams now show up in how they integrate AI: where in the funnel it intervenes, what it replaces, and what it forces a human to do with sharper craft. A CPO who can't read those choices critically becomes a passenger in their own organisation. A product leader is now judged on how clearly they understand the three-month consequences of a tooling decision, well beyond whether they could write the code themselves.
Three operational shifts to watch
The first runs between design and development. Product design is becoming less about approving a finished visual and more about being part of the build itself. As Figma generates production-ready code from the visual layer, the transition into frontend development becomes smoother, as engineers work from code reviews rather than mockups. Teams that hold onto the older visual-approval process tend to carry its overhead well before they see anything return from the newer way of working.
The second is the integration of discovery and ticketing. With Cursor used as a development environment, AI models such as Claude accessed through approved APIs or enterprise services, and Jira connected to the engineering workflow, the distance between identifying a product need and turning it into executable work becomes much shorter.
Where a CPO once spent effort moving work between research and execution and keeping the two aligned, that effort now goes toward setting the quality bar for what enters the flow in the first place.
The third is analytics. With events placed across a product and connected to a platform like PostHog, natural-language prompts can run the kind of analysis that used to require a Discovery brief and a product analyst. A reasonable question can be asked and answered without having to queue for it. The output still needs a sharp reader at the other end; an AI-assisted read is only as good as the question the CPO asks of it, and the framing they bring to what comes back. The analyst doesn't disappear from this picture either; their day reshapes around the questions that warrant deeper investigation.
Another question worth pointing this at: where do users drop off in parts of the product the team doesn't directly control? Most teams can build a clean, friendly interface, but a funnel still has blind spots where users quietly drop off, often when the journey stops telling them what's happening or what to do next. Checkout is the clearest example. Once a user is handed to a payment provider, they move through a redirect, a 3DS check, and a chain of steps that the product team doesn't run. When something breaks in that chain, the team is left tracing what failed, where the routing went wrong, and which part of the experience confused the user.
None of these rollouts lands without friction inside the team, and each has to happen anyway. And these three are only a small part of what's shifting in how tech companies run products right now. The larger the organisation, and the wider the markets it distributes into, the longer the list of challenges a product team has to work through to deliver to users with quality, on time, and above all effectively.
The strategic consequence
The compound effect of these three changes shows up at the strategy level, not at the tooling level. When a landing page can be localized in around 2 hours using pre-built templates, the arithmetic for entering a new market changes. A test that used to require a quarter of preparation now costs a meeting. When product analytics are reachable without a specialist in the loop, the standard for evidence-based decisions rises, because the friction excuse has been removed. Data has gone from one input among several to the biggest one in the room, pushing aside the intuition and market-reads strategy that used to rely on it.
Teams that have been letting processes stand in for direction will find this uncomfortable. Those who already move with clarity get more room to work. AI shortens the distance between an idea and its output, and that's what shows which kind you are. Where working processes aren't in place, adding AI doesn't fix the backlog or the actual work of building the product. It speeds up whatever's already there, so a group running on chaos just speeds up the chaos. A CPO's job right now is to know which pattern is theirs and to close the gap before it shows from the outside.
The shape of product strategy hasn't changed. What's changed is cost, speed, and complexity, to the point that execution no longer holds a company back. Now the limit is what a leader decides to do with the room those shifts have opened, which is what the three points above all add up to.
