AI for Product vs AI for Analytics: Who Actually Benefits?

Your PM shipped a prototype in 4 hours. Your analyst is still waiting for a data request to be approved. Congratulations: AI just made your org even more imbalanced.

Welcome back to Gyanesh on Product. Let's get down to business.

This week I'm joined by Jyoti Gupta , a Product Manager and solo builder who documents hands-on experiments building real products with AI tools in her newsletter, Products in AI ERA. Jyoti lives in the builder's trenches. And that makes her the perfect sparring partner for a question I've been sitting with uncomfortably: AI is clearly changing both product and analytics workflows: but is it changing them equally?

The quick View:

Jyoti’s POV:

For Product Managers, AI has done something genuinely remarkable: it has collapsed the distance between an idea and its proof of concept. Tools like Claude, Cursor, v0, Google AI Studio, and Bolt now let PMs go from a product prompt to a functional mockup or even a lightweight MVP in hours, not the weeks that previously required designer availability, engineering sprint capacity, and three rounds of Figma comments.

The role itself is shifting. The PM who once defined themselves as the coordinator who writes specs and waits is becoming the builder who ships and validates directly. My own newsletter is a living case study: I regularly publishes experiments where I’ve built working product prototypes solo using AI tooling, compressing timelines that would have taken a cross-functional team weeks.

But here's the honest flag I raise, and it's worth paying attention to: speed seduction is real.

The faster you can build, the more you want to build, and the less you're naturally forced to ask whether you should. The analytical rigor that should precede building doesn't automatically show up just because the build phase accelerated.

My two cents:

I used to work in the B2C e-commerce space, and if you know anything about it, you'd know that especially in India, we follow the philosophy of fail fast, develop faster. Even so, I've had to wait for weeks if not months for my "aha" moment. Where I could say, yes, this specific feature we built on user centricity, added "value".

The scenario is changing now, what used to take weeks from:

is now compressed into the power of one single PM with a premium seat in Claude's enterprise plan (definitely not me). Don't get me wrong, I'm a fan of Claude code and the power it brings into my hands. The way I can go to a designer and have a working prototype to refine on over a well defined idea that lived in my head.

But the thought still bothers me, I'm shipping tools and prototypes on internal validation, what about end-user value? Sometimes I think to myself, is this "expensive speed"?

The Analytics Side Is Not Sleeping Either

While PMs were busy prototyping, the analytics world quietly got its own revolution — and it's arguably more structurally significant.

Natural language querying, AI-assisted dashboards, and anomaly detection tools like ChatGPT Advanced Data Analysis, Databricks AI, Tableau AI, and ThoughtSpot have done something unprecedented: they've let non-analysts ask questions of data directly. You no longer need to file a data request or know SQL to pull a cohort retention curve. You just ask.

The adoption numbers tell the story: 73% of organizations now use AI for analytics. Natural language querying has increased BI adoption by 45% across organizations using tools like ThoughtSpot. And AI-powered analytics teams are generating insights 5x faster than traditional BI workflows.

But here's the structural tension Jyoti identified in her brief:

if everyone can "talk to data," what's the analyst's unique value? When a PM, a growth marketer, or a CS rep can pull their own insight in 30 seconds, the analyst's role as gatekeeper of data access evaporates.

What's left is higher-order work: framing the right question, building predictive models, finding non-obvious patterns. And that's genuinely exciting.

The Uncomfortable Math of Democratized Data

Here's a story I've seen play out more than once across product teams I've worked with and observed.

A PM discovers that they can query their own data using natural language tools. Exciting. Revolutionary. Empowering. They pull a metric, see a trend, and make a roadmap call — without ever looping in the analyst. Two months later, it turns out the metric had a known data quality issue the analyst had flagged internally. Nobody told the PM. Nobody had to, because nobody was in the loop.

Data democratization without data literacy is just confident ignorance at scale.

I tracked this pattern across conversations with 5+ product teams in the last year. The teams that saw the best outcomes from AI analytics tools weren't the ones who gave everyone access. They were the ones who paired access with structured analyst-PM workflows, where the analyst's job shifted from running queries to reviewing AI-generated insights for reliability before decisions were made.

The Messy Middle: When Roles Bleed

The most interesting thing happening right now isn't that AI is making PMs better or analysts faster. It's that AI is making each side capable of doing roughly 70% of the other side's job. And nobody has written the org design memo for that world yet.

In Jyoti’s perspective: PMs are becoming more data-capable (AI lets them pull insights without a ticket), and analysts are becoming more product-capable (with AI prototyping tools, an analyst with a strong hypothesis can build a lightweight proof of concept). The overlap zone is expanding.

This creates a genuinely interesting tension: Does the PM-analyst relationship evolve into a tighter, higher-trust partnership where both sides go deeper? Or does it devolve into parallel tracks where each side assumes the other is handling the rigor?

Next time you're reaching for an AI tool to skip a step: ask yourself: am I getting faster at the right thing, or just faster at the wrong thing more confidently?

Hit reply and tell me which side of this you sit on. We'd genuinely like to know.

Co-authored with Jyoti Gupta, PM & solo builder. Follow her newsletter Products in AI ERA on LinkedIn.