How To Use AI To Create A Product Strategy That Actually Wins

AI is now woven into every serious product conversation, yet most teams still use it for copy, mockups and feature ideas instead of the hard part: strategy. That gap is why so many AI initiatives feel busy but do not move the company's north star.

Recent research underlines the stakes:

• McKinsey estimates generative AI could add between 2.6 and 4.4 trillion dollars in annual value across industries by 2030
• A later McKinsey report suggests AI software and services alone could reach 23 trillion dollars in yearly economic value by 2040
• Yet a 2024 BCG survey found 74 percent of companies struggle to scale AI value beyond pilots

That gap between potential and realised value is mostly a strategy problem, not a model problem.

AI will not rescue a vague product strategy. It will only help you get lost faster.

Part 1: Put strategy back at the centre

Before you ask what AI tools to use, decide what strategic questions you want help with. Product strategy still needs to answer four basics:

• Who are we building for
• What jobs or pains we are solving
• How the product will create value for the business
• Why customers choose us over alternatives

AI changes how you answer these, not whether you need them.

Metrics: Why strategy matters in the AI era

Leaders that succeed tend to scale about half as many AI opportunities as laggards, but invest roughly twice as much in the ones they pick.

So the game is not "more AI features". The game is "fewer, sharper AI bets anchored in strategy".

Part 2: The AI strategy stack (from foundation to top)

Think of AI enabled product strategy as a stack of layers. Each layer builds on the previous one.

Layer 1: Strategic foundation

• Product vision and narrative
• A small set of measurable outcomes, for example retention, margin or NPS
• Guardrails on ethics, privacy and brand

Layer 2: Data and access

• Customer and usage data you can legally and safely use
• Access to internal documents, tickets and transcripts
• Clear rules on what can be shared with external models

Layer 3: AI research and insight tools

• Desk research assistants like ChatGPT with browsing or Perplexity for synthesising reports and news
• AI market research tools such as Standard Insights, Synthetic Users or Delve for surveys, synthetic respondents and persona hypotheses
• Competitive intelligence tools like Crayon or Klue to track rivals and pricing

Layer 4: Experiment and prototype engines

• Design and build tools such as Figma, Cursor, Copilot, Windsurf or Loveable that turn natural language into working flows or code
• Analytics platforms with AI features, for example Amplitude or Mixpanel assistants that highlight anomalies and segments

Layer 5: Strategy governance and decision rituals

• Cadence for reviewing AI generated insight packs
• Decision forums that include product, data and risk
• A simple log of strategic bets, assumptions and stop rules

The value of AI in strategy comes from the stack you design around it, not from any single tool.

Part 3: Use AI to accelerate strategy creation (but keep humans in charge)

Here is a practical process you can run when you are shaping or refreshing a product strategy.

Part 3.1: AI assisted discovery

Use AI as an insight multiplier, not as a strategy generator.

Inputs:

• Existing strategy artifacts and OKRs
• Export of support tickets, reviews and NPS comments
• Recent win or loss notes from sales
• Key market reports and analyst notes

AI activities:

• Use a research assistant model to summarise each source and highlight patterns in jobs, pains and triggers
• Use specialised market research tools to build "digital twins" of key customer segments, then stress test your early hypotheses
• Ask AI to map competitors, their positioning and recent feature moves

Human activities:

• Decide which problems and segments matter most
• Rewrite AI summaries into your own words
• Explicitly log assumptions you are willing to bet on

Part 3.2: Strategy funnel with AI

Treat AI enhanced strategy creation as a funnel where ideas get filtered and sharpened.

The exact percentages will differ by company, but forcing a funnel mindset keeps AI from generating a fog of unprioritised opportunities.

Part 4: Continuous strategizing with an AI powered feedback loop

Once the strategy is set, the work shifts to steering. This is where AI can shine as an always on sensor for value and risk.

You can design a simple iteration loop.

This cycle is where many AI leaders are already operating. Surveys show that high performers in AI focus on a smaller number of scaled solutions, invest more deeply in them and revisit assumptions regularly instead of treating strategy as an annual event.

Part 5: AI copilot vs AI autopilot in strategy work

A useful distinction in 2025 is between "copilot" and "autopilot" roles for AI in product strategy.

For most organisations, strategy should stay firmly in copilot territory for the foreseeable future.

Part 6: Prioritise AI opportunities with a 2x2

When AI can plug into almost every part of your strategy process, you need a way to choose where to start.

Use a simple 2x2 matrix with horizontal axis showing impact on core product outcomes (low to high) and vertical axis showing risk of incorrect decisions (low to high).

• Quadrant 1: High impact, high risk – "Guarded bets". Examples: automatic pricing changes, AI driven credit decisions, algorithmic curation that affects trust
• Quadrant 2: High impact, low risk – "First wave". Examples: AI enhanced market research, opportunity sizing, pattern spotting in usage data
• Quadrant 3: Low impact, high risk – "Avoid". Examples: black box strategic decisions in regulated products, unmanaged AI driven org changes
• Quadrant 4: Low impact, low risk – "Playground". Examples: AI generated vision drafts, naming exercises, light competitor summaries

Most teams should begin in Quadrant 2 and Quadrant 4. Quadrant 1 deserves attention only after you have strong governance patterns.

Part 7: Respect AI's limits in strategy work

Roman Pichler rightly points out that AI is not a strategist. The 2025 landscape confirms that view. New studies and case stories echo four recurring limits.

Empathy and field reality

AI cannot substitute time spent in customer calls, visits or live product sessions. It can surface quotes and themes but not the emotional weight behind them.

Data fit

Disruptive products or new markets often lack the historical data AI needs. In those cases, AI tends to copy patterns from adjacent spaces and push you toward the average.

Fabricated precision

Models can produce market sizes, personas and competitor strategies with confident tone even when underlying data is thin. Treat AI output as a working hypothesis that must be verified.

Ethics, privacy and energy

External tools may train on data you do not control, expose sensitive information or raise your carbon footprint. Check data residency, retention policies and opt out options before sending proprietary information anywhere.

A short checklist for your next strategy cycle

1. Define the strategic questions AI should help you answer
2. Set up the AI strategy stack from foundation to governance
3. Use AI heavily in research and synthesis, lightly in decision making
4. Run an explicit strategy funnel from AI insights to a small number of bets
5. Operate a continuous Observe → Explain → Decide → Act → Learn loop
6. Start in low risk, high value quadrants before touching critical levers
7. Document where AI is used in strategy work and how you validate it

Treat AI as the smartest junior on your strategy team. Let it read everything, crunch everything and suggest boldly, then make the final calls yourself.

If you do that, AI stops being a novelty feature generator and becomes what it should be in 2025: a force multiplier for clear product thinking.