Sponsored post: When I get stuck on something these days, my first move is usually to ask my AI. An answer shows up in seconds, it sounds reasonable, and more often than not, I use it.
I doubt I'm alone in this. In just a few years, AI shifted from an experimental side quest to a key part of how work gets done—and product teams have been at the forefront of this. In airfocus by Lucid's recent survey of 500 product leaders, 96% said they use AI in their daily work. 71% said their results would suffer if it were removed from their workflows tomorrow.
However, if you multiply "ask my AI" reflex by every product manager, every day, a bigger problem appears. Product managers now spend their days optimizing all sorts of tasks with AI, but it's mostly happening at the individual level, which can yield inefficiencies. Assembling the necessary context ad hoc for each request wastes time and tokens, and often produces generic, non-applicable output. Few companies have managed to bring that to the institutional level, often resulting in alignment, communication, and clarity issues.
Where does this gap stem from, and why is it so hard to overcome?
The fuller survey results made me look at my own reflex differently, not because it's wrong to use AI that way, but because the data draws a distinction that is often overlooked: using AI a lot, versus using AI well.
The adoption question is settled, the maturity question isn't
80% of respondents said their organization has a clearly defined AI strategy. More than half (57%) also said that strategy is only informal. The two aren't necessarily contradictory. You can have executive sponsorship, a shortlist of approved tools, and genuine enthusiasm, and still miss a real answer to the harder questions: Where should AI create the most value? What data should it work from? Who's accountable for its output? How do you know it's improving decisions rather than just producing more of them, faster?
That’s the AI maturity gap: activity that looks like strategy but isn't operationalized as one. Adoption means AI is present in the workflow. Maturity means it is scaled to the institutional level and reasons over a shared context layer.
Speed was never the hard part
The book Prediction Machines breaks every decision down into three parts: prediction, judgment, and action. AI has made prediction dramatically faster and cheaper. Judgment and action remain untouched, and judgment has always been the expensive, slow, human part of the equation.
That's the trap in my own "ask my AI" habit, and likely in the individual-level pattern the survey is picking up more broadly. The answer arrives fast and sounds reasonable, which makes it easy to mistake speed for correctness. Building fast, or answering fast, was never the hard part of product work. Deciding well was, and still is. AI has changed how quickly you arrive at the decision point, not how much judgment the decision itself still requires.
The real constraint: trust
When we asked leaders what's holding back further scaling, the top answer wasn't a tooling gap. It was:
- Trust in AI outputs (40%)
- Security and privacy concerns (39%)
- Lack of training (35%)
That echoes with another pattern revealed in the report. Teams use AI for the parts of the job where clarity matters most:
- 57% use it for analytics and experimentation
- 52% for user research
- 47% for customer feedback analysis
Yet 48% say they still struggle to separate signal from noise. AI can summarize faster than any of us. But if what it's drawing on is fragmented, or rebuilt from scratch by every person who asks it a question, the result is more noise, produced faster and apparently more reasonable.
As Spencer Cowley, Product Manager at airfocus by Lucid, put it recently: “AI agents are only as useful as the context they can access.”
Bridging the gap: How to build trust in AI
The research provides a clearer line to draw between two things that are often treated as the same: how often a team uses AI, and how well it's set up to use AI's output responsibly, as a team, not as 500 individuals each swimming in their own lane.
Nearly every product org has answered AI adoption by now. Maturity is where most of the real work still is. And in most cases, what stands between the two is trust.
So here are a few tips I’ve found used in my own maturity journey, to build trust in AI:
- Focus on value: Deploy AI intentionally where it yields the highest impact, whether it’s speed, synthesis, surfacing trade-offs, and/or communication.
- Ground AI in real product context: Embed tools directly within roadmaps, OKRs, and customer feedback loops.
- Design for trust and transparency: Make the data foundation behind AI outputs fully traceable.
- Reduce noise, increase relevance: Eliminate data clutter caused by redundant dashboards, drafts, and summaries.
- Integrate AI into your operating model: Treat AI as a core component of decision-making rather than an isolated experiment.
The full report goes deeper into the strategy paradox, the blocker rankings, and where the signal-to-noise struggle is most evident. Worth a look if any of this sounds familiar.
