The Answers (sort of): Does Product Ops Have a Data Problem
TL;DR: Few answers. But the right questions.
A few weeks ago, I wrote about the questions I was taking to MXP London.
How do we build data foundations for strategic insight?
How do we use AI to amplify judgment without outsourcing it?
How do we ensure customer voice genuinely informs decisions?
Refresh your memory here:
Honestly, I’m not sure I came away with neat answers. But I came back convinced the whole industry is asking the same things.
The speed problem nobody is ready for.
Mixpanel CEO Jen Taylor opened with both palpable enthusiasm and a particular stat that landed:
42% of engineering teams now use AI-assisted coding, up from 6% in 2023, predicted to hit 65% by 2027.
But also,
…faster building means the capacity for faster bad decisions.
This, above all, was the theme I was sensing from my conversations with fellow attendees.
Mixpanel AI, talked up a lot here, is a shift from product analytics to product intelligence - always-on AI that recommends what to do next, not just what happened. Teams moving 10x faster need context 10x faster too. The bottleneck is not dashboards; it is getting the right context to the right people before the decision window closes.
FaceIT’s Maria Laura Scuri’s brilliant session highlighted another important fact: democratising data is essential, but you still need people who know the data well enough to verify what AI recommends. Several organisations shared practical examples; using AI for compliance validation, checking documentation, branding consistency and regulatory wording. Useful, tangible, but still requiring human oversight. I know, because I’m building the same things!
And MCPs are still a hot topic. Vendor after vendor is building MCP integrations so their data feeds directly into AI tools like Claude or ChatGPT. The direction is clear: bring intelligence to where you work, not force people into another interface.
AI is not replacing judgment. It is exposing who has it
The exec roundtable I joined (honoured!) surfaced something I’m hearing too often, too:
PMs and designers are becoming the new bottlenecks, not engineers.
When implementation speed is no longer the constraint, the quality of thinking becomes the constraint.
This connects to a broader theme of role blurring between PMs, designers and engineers, though I wonder how prevalent this really is beyond the ‘top 1%’ of tech companies right now.
AI brings focus and time back to humans for deeper thinking, but only if they use it well. There is a real risk of decision fatigue when AI makes more decisions feel possible. Martin Eriksson’s Decision Stack framework feels relevant here; without clear principles guiding which decisions matter, more capacity to decide just creates more noise.
Re-read my review of the Decision Stack here:
So did I get my answers?
Not exactly. But a room of senior product leaders wrestling with the same questions - and honest about not having answers - tells me the sector is more aligned than it appears.
Few answers. But the right questions. And that is enough to build from.
And what does this mean for Product Ops, dear gentle readers?
(I’ve just won a bet with my wife that I could get this phrase from her favourite TV program into an article - answers on a postcard if you know it too!)
That we are in a prime position to enable these rapidly changing functions that themselves do not exactly know anymore how to define themselves. The tools are there for sure, but this is not about tooling and, once again, for Product Ops, it never was, and overall, my conversations with product leaders at MXP highlighted this like… well… a big yellow highlighter. This is about how people react to changes they are not in full control of. This is how we enable the right decision-making backed by the right data at the right time. AI is making this fast and fluid, but humans are anything but this, so more than ever, it is a reminder to be making the right decisions, not fast decisions.
Graham