Scaffold · Guides

Best AI tools for product managers in 2026, by the job you need

Choose by the job, one tool each. For drafting and research, a general assistant: ChatGPT, Claude, Gemini or Perplexity. For interviews and feedback, Dovetail or Productboard Spark. For meetings, Granola. For analytics, Amplitude. For prototypes, Figma Make or Lovable. For the strategy all of them serve, Scaffold. Keep the judgement yours.

By Cyril Le Roux · Updated 19 November 2026 · 14 min read

Disclosure: Produxity builds Scaffold, one of the tools below. Where the others beat Scaffold, we say so. Prices were read from each vendor's own page on 19 November 2026; where a page could not be read or publishes no price, we say that instead. Produxity is not affiliated with any other vendor named here.

Monday morning, and the browser has eleven tabs of AI tools in it. One writes PRDs, one summarises calls, one builds prototypes, one promises to tell you what to build next. Each has a free trial, and each trial wants your customer data. By Friday you have drafted more than ever, and the roadmap review still turns into a negotiation.

Most lists of AI tools for product managers stop at the tools. This one sorts them by the job they take off you, says what each costs, and spends some time on the one job that should stay yours.

Why this guide reads the way it does

Cyril Le Roux teaches Product Focus's flagship product management course and co-authored its AI course, and has been a senior product leader at BT, Photobox, Zoopla and TransferGo. The question he asks of any task, and the thread through this guide: is this a job for me, or something an AI could take care of?

His answer for most product managers: far more of the job than they hand over today, and never the choice of direction.

Three questions to ask of any AI tool

  1. Does it take a whole job off you? Sales enablement material, customer service training, first drafts of specs and release notes: none of these deserves much of a product manager's week any more. A tool that saves ten minutes inside a job you still do by hand is a nice-to-have.
  2. Can you check what it tells you? Never trust an AI completely. Good domain knowledge lets you spot what does not add up, and a tool that shows its sources lets you check the rest. Data analysis tends to be reliable; claims about markets and competitors need the most checking.
  3. Could you defend the result? You are the one who explains the roadmap to leadership. If you cannot say why an AI-drafted answer is right, you have not saved time; you have borrowed it.

Hand AI the work that feels like progress. Take back the time for the work that is.

Tactical work drowns strategy because it is easy and it feels like progress. AI is the first chance in years to claim that time back. The point of the tools below is what you do with the hours: understanding your market, your customers and your competitors well enough to choose a direction. Use AI there too, to move faster.

Job 1: a general assistant for drafting, research and thinking out loud

ChatGPT (OpenAI)

Free; Go $8 and Plus $20 a month; Pro from $100 a month

Best for: Drafting, summarising and analysing data, with the widest choice of plans and integrations

Watch for: Starts every conversation from what you paste in, unless you set up projects and memory with care

Claude (Anthropic)

Free; Pro $17 a month billed yearly or $20 monthly; Team from $20 per seat a month billed yearly

Best for: Long documents, careful writing and reasoning through a messy brief

Watch for: Same limit as any chat: it knows your product only as well as the context you give it

Gemini (Google)

Free; Google AI Pro £18.99 a month (the page showed UK prices only); for business, Gemini with Google Workspace from $14 per seat a month

Best for: Teams that already work in Google Docs, Sheets and Gmail

Watch for: Strongest inside Google Workspace; less of an advantage if your documents live elsewhere

Perplexity

Free; Pro $17 a month billed yearly

Best for: Fast market and competitor research with the sources listed beside each answer

Watch for: Cited sources make checking easier, not optional: open the ones your argument rests on

Job 2: customer research and feedback

Dovetail

No free plan listed; Bronze $4,800 a year plus credits, marked coming soon; Silver and Gold through sales

Best for: Turning interview recordings and transcripts into tagged, searchable evidence

Watch for: The synthesis is only as good as the interviews: it cannot fix leading questions

Productboard Spark

On request; Business includes 500 Spark credits per maker a month, Enterprise more

Best for: Teams already running feedback and roadmaps in Productboard who want an agent working over that data

Watch for: Its value depends on Productboard being where your feedback already lives

Enterpret

No published price; demo on request

Best for: Larger teams pulling support, sales and survey feedback into one place and ranking it

Watch for: Built for volume; a team with a few dozen requests a month may not need it

Job 3: meetings and notes

Granola

Basic free; Business $14 and Enterprise $35 per user a month

Best for: Notes from back-to-back calls without a bot joining the meeting

Watch for: Ask the other side before you record, and check the summary against what was agreed

Job 4: product analytics

Amplitude

Free up to 2 million events a month; Growth and Enterprise on request

Best for: Asking questions of usage data in plain language, with AI agents on every plan

Watch for: The answers are about what users did, not why; pair them with research

Job 5: mock-ups and prototypes

Figma Make

Free Starter plan with daily AI credits; Professional full seat from $16 a month, with 3,000 AI credits

Best for: Prototypes that stay next to your designers and their design system

Watch for: Agree with your designers where a PM prototype stops and design work starts

Lovable

Free plan with daily build credits; paid plans priced by credits, not per person (prices could not be read on the date of this guide)

Best for: A working web prototype from a conversation, to put in front of customers

Watch for: A clickable prototype can make an untested idea look finished

Job 6: discovery, documents and the roadmap

Jira Product Discovery

Free for up to 3 creators; $10.50 or $26.25 per creator a month

Best for: Ideas, insights and roadmaps next to the delivery work in Jira, with AI to organise and connect feedback

Watch for: Its home is the backlog and the roadmap, not the choice of direction

Notion AI

Included on Business, £16.50 per member a month (the page showed UK prices only); a trial on Free and Plus

Best for: Searching and drafting across the documents your team already keeps in Notion

Watch for: It answers from what is in the workspace, so out-of-date pages give out-of-date answers

Two techniques make these tools far more useful, and neither needs software. Before recording a customer interview, learn the approach in Rob Fitzpatrick's book The Mom Test: ask about what people did in the past, not what they would like in future, and no AI summary will be better than the questions behind it. And for competitors, have an assistant draft a profile of each rival that includes the strategy it appears to be following and the play you could make against it. A feature comparison dates within months. A rival's strategy tells you where its features are going.

Job 7: the strategy the other tools serve

Every tool above makes you faster at something. None of them decides where the product is going, and without that, faster means more features pulling in more directions: death by a thousand cuts, delivered sooner. There are three ways to do this job.

A general assistant and a disciplined process. Cheap, fast and enough for an experienced product manager who knows which techniques to use and in what order. The risk is a plausible strategy built in one conversation, with nothing to show what it rests on six weeks later.

A roadmap or discovery tool with AI. Jira Product Discovery, Productboard and Aha! all give a strategy somewhere to live next to the roadmap. They record the answers well. Arriving at them is still your work.

Scaffold. Scaffold is an AI product strategy tool for product managers and product engineers. Paste a link and drop in your documents: a free research run drafts your starting point, Scaffold suggests the approaches that suit your product's stage, and Strategic Options compares every direction you kept with changing nothing. You choose and validate. Here is what sits on every product's map:

  1. 01Context: where you areCompany Foundations, Persona, PEST Analysis, Competitor Profiles, Porter's Five Forces, Value Proposition Canvas
  2. 02Route: how you get thereStrategic Options, Strategy Narrative ("The Trip"), Now, Next, Later
  3. 03Destination: where you're goingProduct Vision, 3 Horizons

Scaffold owns the judgement, never the backlog. It does not track delivery: no sprints, no statuses, no code generation. That stays in Linear or Jira, and the research, meeting and prototype tools above stay useful alongside it.

For the AI tools that help with the strategy work itself, compared job by job, see the best AI tools to help a product manager build a product strategy.

Where the others beat Scaffold

And it is the wrong tool for one kind of product manager: the one in a very early startup whose CEO runs the strategy hands-on. There, the job is to execute that strategy faithfully, not to write a separate one. Scaffold could still be useful to that CEO.

A note on lists like this one

Including ours. Feature lists date within months, and every vendor here ships AI features constantly. A tool's features today tell you less than where its maker is heading. Read this guide as a snapshot from 19 November 2026, and check each vendor's own page before you buy.

Choosing, in five lines

  1. Start with one general assistant, and use it for a week on real work before adding anything.
  2. Add a specialist tool for your biggest recurring time sink: interviews, meetings or analytics.
  3. Prototype only what you will put in front of customers.
  4. Check the sources behind any claim your argument rests on.
  5. Spend the hours you win on the strategy, and keep the choice of direction yours.

This week, list the five tasks that took most of your time and ask of each: a job for me, or for an AI? Then decide what you will do with the hours.

Related guides: how to build a product strategy, and the alternatives to Productboard and Aha! Roadmaps.

Questions product managers ask

What are the best AI tools for product managers right now?

Pick one per job. A general assistant (ChatGPT, Claude, Gemini or Perplexity) for drafting and research. Dovetail or Productboard Spark for interviews and feedback. Granola for meetings. Amplitude for analytics. Figma Make or Lovable for prototypes. Jira Product Discovery or Notion AI next to the roadmap. And for deciding the strategy all of them serve, Scaffold.

Do I need a dedicated AI tool for product management, or are ChatGPT and Claude enough?

For drafting a PRD, a release note or an interview guide, a general assistant is enough, which makes single-purpose writing tools a hard sell. A dedicated tool earns its place when it holds context the chat does not: your feedback history, your usage data, your design system, or the earlier analysis your strategy rests on.

ChatGPT or Claude: which should a product manager pick?

Try both on the same real task for a week, because the difference depends on the work. Claude is strong on long documents and careful writing; ChatGPT has the wider set of plans and integrations. Claude Pro is $17 a month billed yearly or $20 monthly; ChatGPT Plus is $20 a month.

Are there free AI tools for product managers?

Yes, and enough to start. Claude, Gemini, Perplexity and ChatGPT all have free plans. Granola's Basic plan is free, Amplitude is free up to 2 million events a month, Figma's Starter plan includes daily AI credits, and Jira Product Discovery is free for up to three creators. All checked on 19 November 2026.

How much should a product team spend on AI tools?

Start from the hours, not the budget. A general assistant costs roughly $17 to $25 per person a month at the published prices in this guide, and covers most of the drafting. Add a specialist tool only when it takes a recurring job off you that the assistant cannot, and drop any tool nobody opened last month.

Which AI tool is best for analysing customer feedback?

Dovetail for interviews and research recordings; Productboard Spark if your feedback already lives in Productboard; Enterpret for large volumes across support, sales and surveys. Whichever you use, treat each request as evidence about a problem, group requests by problem, and then ask how strongly solving it serves the strategy. Tools do the grouping well.

Is there software that turns customer interviews into a PRD without the copy and paste?

Partly. Research tools such as Dovetail tag and summarise interviews well, and a general assistant can draft a spec from the quotes you give it. The step between them, deciding which problem the PRD should solve and why now, stays with you. Keep each quote linked to its interview so reviewers can check the claim.

What AI tool should I use to create mock-ups and prototypes?

Figma Make if your designers work in Figma and you want the prototype next to their design system; a Professional full seat starts at $16 a month. Lovable if you want a working web prototype to put in front of customers. Either way, a polished prototype can make an untested idea look finished, so show it to customers before stakeholders.

How can AI help with product analytics?

It lets you ask usage data questions in plain language: where people drop off, what retained users do differently, which features are barely used. Amplitude includes AI agents on every plan, and a general assistant can analyse an exported CSV. Data analysis is one of the more reliable uses of AI. The data says what users did; research says why.

Can a product manager use AI to understand the codebase?

Yes, for reading, with care. A coding assistant such as Cursor or GitHub Copilot, pointed at the repository, can explain how a feature works and where its logic lives. Ask it to cite the files it read, check the answer with an engineer before you repeat it, and leave changes to the code to the people who own it.

How should I decide which AI model to use for which task?

Keep a short list of your recurring tasks and test two models on each with real material. You will usually find one model better for long writing, another for data, another for research with sources. For decisions that matter, compare answers from more than one model and look hardest at where they disagree.

Can AI write my product strategy?

It can draft one, and that is the risk: a plausible strategy nobody can defend. AI is good at research, competitor profiles and options. Choosing between options needs domain knowledge of your market, customers and competitors, and you are the one who will explain the choice to leadership. Use AI to widen the options and to test them, then choose yourself.

Should I use AI for decisions with real risk?

Use it for the inputs and keep the decision. AI can gather evidence, summarise research and argue against your preferred option. It should not make the call. Never trust it completely: check the sources it cites, and rely on your domain knowledge to spot what does not add up. If you cannot explain a recommendation, do not adopt it.

How do I check AI output before I rely on it?

Open the sources behind any claim your argument depends on, and check them yourself. Ask the model what would make its answer wrong. For market claims, compare answers from two models and treat agreement as stronger evidence than either alone. Be most careful where the answer is fluent and you know least about the subject.

Is my product data safe with AI tools?

It depends on the plan and the vendor, so read each one's data and training terms before you paste in anything confidential. Business and enterprise plans usually give more control than personal ones. Ask where data is stored, which AI providers process it, and whether your content is used to train models. Scaffold's answers are on its Data and security page.

Will AI replace product managers?

It is replacing parts of the job, and that is an opportunity. Sales enablement material and customer service training no longer deserve much PM time. Keep asking whether a task is a job for you or something an AI could take care of, and reinvest the time in understanding your market, customers and competitors. Judgement about direction stays with you.

Everyone on my team asks their own AI. How do we keep a shared view of the product?

Write the shared view down where everyone, and every assistant, can read it: the vision, the strategy as a few principles, and the evidence behind them. Private chats drift apart because each starts from different context. Present the strategy before every roadmap discussion, so people argue from the same starting point, whatever their assistant told them.

How are product managers using AI day to day?

For first drafts of specs, release notes and stakeholder updates; summaries of interviews and meetings; quick analysis of exports; prototypes to test an idea; and research on competitors and the market. The ones who gain most use the time saved for the work AI cannot do for them: talking to customers and agreeing the strategy.

How many AI tools should a product manager use?

Fewer than you think. Start with one general assistant and add a specialist tool for your biggest recurring time sink, often interviews or meetings. Each extra tool is another place context lives, so a third or fourth tool needs to save more time than it costs to keep current. Most solo PMs need two or three.

Should I choose AI tools by comparing features?

Only last. Feature tables date within months, and every vendor in this guide ships AI features constantly. Choose by the job you need done, then by where each vendor appears to be heading, and only then by features. The tables here are a snapshot from 19 November 2026, with prices read from each vendor's own pages.

How is Scaffold different from asking ChatGPT for a strategy?

A chat starts from what you paste in. Scaffold starts from your product's link and your documents: a free research run drafts your starting point, Scaffold suggests the approaches that suit your product's stage, and Strategic Options compares every direction you kept with changing nothing. You choose and validate. Either way, you have to explain the strategy, so convince yourself first.

Who is Scaffold wrong for?

Anyone looking for a backlog, roadmap or feedback tool: the other tools in this guide do that and Scaffold does not. It is also wrong for a product manager in a very early startup whose CEO runs the strategy hands-on, where the job is to execute that strategy faithfully. Scaffold could still be useful to that CEO.

About the author

Cyril Le Roux teaches Product Focus's flagship product management course and co-authored its AI course. He has been a senior product leader at BT, Photobox, Zoopla and TransferGo, founded Produxity, which builds Scaffold, and is a co-founder of Frntir.ai.

Sources, read 19 November 2026