The AI tools I actually use as a Product Manager (and the moments they save my week)

From a messy signal to a confident decision using a small set of tools that earned their place.

A few times a month, I get the same question. “What AI tools do you use as a product manager?”

The funny part is that people usually expect a punchy list. A stack. A magic combo. Something that makes product work feel clean and linear. But product work is rarely clean. It is a mix of half-signals, conflicting opinions, and that one tiny UX issue that quietly destroys activation while everyone debates the roadmap.

So instead, I want to share the way these tools show up in my real routine. The moments they save my week.

Here are the AI tools for product managers I use weekly.

Who this is for: Product Managers in SaaS, growth, and B2B products who want to move faster from signal to decision without losing quality.

Perplexity Pro for user research

I use it when I need to quickly understand how people talk about a problem, what alternatives exist, and which patterns repeat across competitors.

It is especially useful for two situations:

when I am entering a new category and need to get oriented quickly
when I suspect our users are feeling a pain point we have not named well yet

A prompt that consistently gives me useful output:

“Summarize the top pain points for this persona in this category. Group by job to be done. For each pain point, list triggers, workarounds, and how strong products address it.”

So basically, I treat it as a shortlist of hypotheses worth validating.

This is where I go from signal to hypothesis.

Fireflies for call summaries

User calls are emotionally persuasive. After five calls, it is easy to remember the most charismatic person or the one complaint that matches your existing beliefs. But for the bigger picture, I use Fireflies. It helps me keep user feedback honest.

What I typically extract after a batch:

recurring pain points
the exact words users use when they struggle
where they hesitate, ask questions, or change direction
action items and decisions that should not get lost

This is where I turn qualitative noise into structured evidence.

Hotjar for behavioral friction

Users do not open a ticket to say: “Your UI makes me slightly uncertain, so I procrastinated and churned quietly.” They just silently disappear. Hotjar helps me catch that quiet friction. I use it heavily for onboarding and activation, because early flow quality is where retention begins.

This is where I validate friction with real behavior before I invest in solutions.

UX Pilot for rapid prototyping

There are numerous tools for vibe coding and design prototyping. But UX Pilot is the best I’ve found for my workflow so far. I find it consistently produces usable variants at this time.

UX Pilot helps me move the conversation from opinions to options. I have been using it for almost half a year, and what I like is that it often suggests small but high-leverage details like better layout hierarchy, clearer CTAs, or visualizations that make the product “click” faster.

The prompt I use most is centered on the first value:

“Design a flow for this persona to achieve this job. Optimize for first value in 60 seconds. Include empty states and error states. Give me three variants with different information architecture.”

This is where I stop debating and start shaping solutions I can test.

GPT-5.2 Pro for my PRD quality gate

I often use GPT-5.2 extended thinking to make the spec sharper, simpler, and harder to misinterpret. So my approach consists of multiple steps and roles: gather information, structure, align with the goals and current metrics by one agent and then criticize by another agent.

A prompt I come back to:

“Review this PRD like a skeptical Staff PM. Identify unclear assumptions and missing edge cases. Rewrite it shorter but more concrete. Add acceptance criteria, instrumentation, and rollout guardrails.”

It turns a PRD from a document into an execution tool.

This is where I turn a solution idea into an executable plan with fewer surprises.

shadcn/ui and Cursor: the fastest way to ship UI without losing control

shadcn/ui helps because it keeps UI consistent and maintainable. You own the code. You are not trapped by a library’s constraints. Basically, it gives you a more controllable approach than generated UI.

I use Cursor because it speeds up building and refactoring in a real codebase, especially when I need to iterate quickly while keeping the UI maintainable.

So Cursor plus shadcn makes the build phase feel lighter, customizable, and fast. I like to use it to build small real apps and easily test them on my localhost.

As an alternative, I like Rocket.new. While Cursor is good for developers (or technical PMs), it could be difficult for non-technical PMs who do not understand how a typical IDE works. Lovable is good for non-coders, but there is less control and it is harder to evolve into a real product. So Rocket.new helps with strong Figma-to-web conversion and broader use cases (dashboards, mobile).

This is where prototypes become real features.

PostHog for experiments

The painful truth is that many experiments die in analysis. In my experience, people always tend to argue about metrics, segments, tracking quality, and whether the lift is real. PostHog AI agents help me compress that loop.

I use them to:

turn questions into analysis quickly
surface segments that matter
detect tracking gaps before we trust conclusions
summarize outcomes in decision language

This is where I close the loop and turn results into decisions.

TL;DR: the stack

Rapid UX prototyping: UX Pilot
Shipping UI fast (with control): shadcn/ui + Cursor
Polishing PRDs and specs: GPT-5.2 Pro
Alternative fast Figma-to-app flows: Rocket.new
Customer call summaries and insights: Fireflies
User research: Perplexity Pro
Behavior analytics and friction mining: Hotjar (AI highlights)
Rapid experiments and analysis: PostHog AI agents

One last question

Which AI tool has genuinely earned a permanent place in your workflow, and which one did you expect to love but quietly stopped using?

If you are working on onboarding, activation, retention, or building AI copilots, and you want a second brain to pressure test hypotheses and experiments, feel free to reach out.

More on PLG, AI adoption, and growth experiments:

Enjoyed this article?

Get new PLG + AI notes in your inbox.

No spam. Practical playbooks and experiments. 1–2 emails/month.

Email in