How You Need to Think About Product Strategy in the Age of AI

AI has made shipping features cheap, which makes choosing the wrong ones more expensive than ever. Tim Herbig, product strategy coach and author of Real Progress, walks through how to build a product strategy that actually helps you say no. Full write-up: https://www.news.aakashg.com/p/how-the-best-pms-think-about-product-strategy Transcript: https://www.aakashg.com/product-strategy-in-the-age-of-ai-with-tim-herbig/ Timestamps 03:06 - Cheap features make wrong choices expensive 05:33 - The one test every strategy has to pass 07:19 - Ads 09:02 - The core questions every strategy answers 12:39 - Why AI is not a value proposition 18:41 - The Roger Martin test and the stupidity flip 25:11 - Ads 30:59 - Case study - Fair Mile, a used car marketplace 35:43 - Audience trees 42:16 - Differentiation with the 7 Powers 52:48 - How to know if your strategy worked 58:30 - Building the strategy and metrics loop 1:04:33 - What you inherit vs. what you create 🏆 Thanks to our sponsors 1. Fingerprint: Identify good and bad visitors with industry-leading accuracy - https://tinyurl.com/2s3p65p7 2. Ariso: Ship AI agents and features faster, with fewer regressions - https://ariso.ai/aakash 3. Customer.io: Send smarter messages using your product data - http://customer.io/productgrowth 4. Amplitude: The market leader in product analytics - https://tinyurl.com/b7nret7v 5. Product Faculty: Get $150 off their #1 AI PM Certification via code AAKASH150 - https://www.productfaculty.com/?code=AAKASH150 Key Takeaways 1. Cheap features make wrong bets expensive - AI made shipping fast, so picking the wrong feature now costs more, not less. You still have to get people to use everything you ship. The real question is whether any of that output moved customers and the business. 2. A strategy is only real if it helps you say no - Deck, one-pager, or framework, the test is the same. If it doesn't help teams say no to opportunities, it isn't a strategy. A healthy one has most of your team's energy going to work connected to it. 3. Every strategy answers the same core questions - A narrative for why act now, a playing field for who you serve and who else competes, and winning moves for how you deliver value and differentiate. Canvases and templates are just windows into that messy work. Start with the questions, not the template. 4. AI is a resource, not a value proposition - Everyone has access to the same frontier models, so you can't compete on the technology alone. Use AI to amplify the reason customers already choose you. Brand and data are the moats that remain. 5. Run the stupidity flip - Borrowed from Roger Martin. Flip a strategic choice and check whether the opposite is obviously stupid. If it is, you haven't made a real choice. Flip individual components like audience or value prop, not the whole statement. 6. Strategy doesn't have to cover everything - Focusing on paid users doesn't mean turning off the product for free users. The segment choice just has to fit the narrative behind it. Every component has to reinforce the others. 7. Build an audience tree - Land somewhere between too broad and comically narrow. Slice by region, buyer type, and car value until you can say no. Fair Mile lands on repeat buyers of cars worth more than 100K in Germany, and AI speeds up sizing whether that market is big enough. 8. Differentiate on what takes longest to build - Run your winning moves through Hamilton Helmer's 7 Powers. Brand and network effects hold up, while process power and AI features get copied. No one buys a car from you because your marketplace has AI search. 9. Measure behavior change, not tokens - First say in plain English what you will do differently next quarter, then how you'll know it worked. Tokens used is a value exchange for the model provider, not your customers. Ask whose behavior has to change, and by how much. 10. Tie metrics to your specific choices - Generic KPIs like revenue are too noisy and too slow to tell you if a strategy worked. Fair Mile tracks share of cars sold above 100K, margin per transaction, and recommendation score among those buyers. Then map every roadmap item back to those metrics. 👨‍💻 Where to find Tim Herbig LinkedIn: https://www.linkedin.com/in/herbigt/ Website: https://herbig.co/ 👨‍💻 Where to find Aakash Twitter: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com 🧠 About Product Growth The world's largest podcast focused solely on product + growth, with over 200K+ listeners. 🔔 Subscribe and turn on notifications.

Tim HerbigguestAakash Guptahost

Oct 14, 20261h 12mWatch on YouTube ↗

CHAPTERS

0:00 – 3:07
Why cheap AI-built features make bad choices more expensive
Tim frames the AI era’s core product problem: output is dramatically cheaper, but picking the wrong things to build is costlier because you still must drive adoption and business impact. This sets up strategy as the mechanism for focus, trade-offs, and measurable outcomes.
- •AI accelerates shipping, but doesn’t guarantee customer or business impact
- •Cheap features can flood products with low-value ‘output’
- •Adoption becomes the real bottleneck when everything can be built
- •Strategy’s job is to clarify what to achieve and what not to do
3:07 – 9:06
The one test every strategy must pass: does it help you say no?
Tim defines strategy in practical terms: it must enable teams to decline opportunities over a given horizon and align effort to what matters. If your “strategy” can’t be used to reject work, it’s just a list of aspirations.
- •A usable strategy enables intentional yes/no decisions
- •A long list of priorities usually signals lack of strategy
- •If it can’t constrain choices, it isn’t strategy
- •Most team energy should be spent on work connected to strategy
9:06 – 12:43
Strategy’s core questions: narrative, playing field, and winning moves
Tim decomposes strategy into three answerable dimensions: why act now (narrative), where to play (market/alternatives), and how to win (value and differentiation). He emphasizes that frameworks are just containers—real strategy work is messy and nonlinear.
- •Strategic narrative: why us, why now
- •Playing field: who it’s for, what problems, what alternatives exist
- •Winning moves: value proposition and differentiation
- •Canvases/templates are ‘windows’ into messy strategy work, not the work itself
12:43 – 16:20
AI isn’t a value proposition—everyone has access to the same models
Tim argues AI alone rarely differentiates because frontier models are broadly available. Sustainable advantage comes from what’s layered on top—brand, data, and domain expertise embedded into a product experience.
- •AI is a resource to amplify value, not the value itself
- •Embedding a generic chat interface is not differentiation
- •Remaining durable differentiators: brand and data
- •Examples: solopreneur brand (Levels.io) and CPO expertise embedded in ChatPRD
16:20 – 18:51
Where AI helps (and where it can’t): strategy creation vs. strategy content
The conversation distinguishes using AI as a workflow accelerator from making AI the centerpiece of the strategy. Tim lists high-leverage uses (research, synthesis, scenario comparison) while warning that accountability and consequences still sit with the team.
- •AI can accelerate market/competitor research and evidence synthesis
- •AI can generate and pressure-test scenarios faster
- •Teams remain accountable—‘the model told me’ isn’t a defense
- •AI is not a moat; brand and data take longest to build and endure
18:51 – 30:59
The Roger Martin test and ‘stupidity flip’: making real strategic choices
Tim introduces Roger Martin’s heuristic: if the opposite of your strategy sounds stupid, you haven’t made a real choice. He shows how to apply the ‘stupidity flip’ to individual components (audience, value prop, how-to) to force specificity and trade-offs.
- •If the opposite is obviously stupid, the statement isn’t strategic
- •Strategy must answer ‘how’ (specific mechanism), not only ‘what’
- •Apply the flip to components, not only a single one-line strategy
- •Segment focus is valid if it coheres with the narrative (the ‘why’)
30:59 – 35:43
Case study setup: Fair Mile used-car marketplace and the danger of an ‘all-yes’ strategy
Using the fictional ‘Fair Mile’ marketplace, Tim illustrates typical marketplace tensions: trust, inspection, liquidity, and buyer uncertainty. He contrasts an unfocused ‘do everything’ approach with making specific, flippable strategic bets (e.g., first-time vs repeat buyers).
- •Marketplace needs: trust, verification, matching supply/demand
- •Buyers may be uncertain, especially first-time or sight-unseen purchases
- •An ‘all-yes’ strategy can’t say no and leads to chaos
- •Use flips to generate legitimate alternatives (AI vs human-led verification, etc.)
35:43 – 42:25
Audience Trees: narrowing to a focus that’s not ‘comically narrow’
Tim teaches Audience Trees to structure segmentation choices without drifting into either ‘everyone’ or absurdly specific niches. AI becomes powerful here for rapid sizing and validation of segment assumptions, while still preserving strategic trade-offs.
- •Start with very broad vs comically narrow endpoints to calibrate focus
- •Slice audiences by meaningful dimensions (region, buyer type, car value, etc.)
- •Rearranging slices still leads to a clear chosen segment
- •AI helps with market sizing and segment research to sanity-check focus
42:25 – 46:31
Differentiation using Helmer’s 7 Powers—what still works in an AI world
Tim applies Hamilton Helmer’s 7 Powers to clarify which moats are plausible and durable. He highlights that some powers (brand, network effects) remain hard to copy, while AI features (e.g., AI search) often won’t drive true marketplace differentiation.
- •7 Powers as a checklist for feasible differentiation
- •Brand and network effects are durable; process advantages are catch-up-able
- •Switching costs often don’t fit certain products (e.g., car marketplaces)
- •‘Captured resource’ via AI models is weak when models are widely accessible
46:31 – 52:48
No fixed sequence: strategy is a web of linked choices
Tim explains you can start anywhere—audience, differentiation, channel—but must ensure consistency across the narrative, playing field, and winning moves. He also cautions against ‘winning on price’ unless you have structural advantages to sustain it.
- •Strategy components are interdependent; coherence matters more than order
- •Differentiation must match audience realities and competitive alternatives
- •Be explicit about why customers choose competitors today
- •Competing on price is usually a losing game without subsidies or structural edge
52:48 – 58:30
From strategy to measurement: proving strategy worked (and avoiding ‘token metrics’)
Tim focuses on how to know whether strategy is working: describe what you’ll do differently and how you’ll measure it, in plain English, before choosing a framework. He warns against machine-centric AI metrics (tokens, model efficiency) that don’t represent customer value.
- •Start with ‘What will we do differently?’ (the essence of strategy)
- •Then define ‘How will we know it worked?’ before picking OKRs/dashboards
- •Avoid AI-internal metrics that aren’t customer value (e.g., tokens used)
- •Pick leading vs lagging indicators thoughtfully; balance speed with meaning
58:30 – 1:04:35
Building the strategy–metrics loop: tie metrics to specific strategic choices
Using Fair Mile, Tim demonstrates linking narrative + playing field + winning move to metrics that directly reflect those choices. The loop closes when metrics inform course-correction, and roadmap items map back to those strategy-derived metrics.
- •Strategy example: focus on repeat high-end German buyers + white-glove inspection/delivery
- •Choose metrics tightly coupled to choices (share of >100K cars sold, margin, NPS in segment)
- •Avoid overly generic KPIs (e.g., ‘increase revenue’) as execution metrics
- •Use metrics to validate strategy and trigger iteration; map roadmap items back to metrics
1:04:35 – 1:12:12
What you inherit vs. what you create: strategy levels and ‘Evergreen benefits’
Tim addresses multi-level strategy: teams inherit constraints from higher levels but must narrow further until they can say no in their scope. He introduces ‘Evergreen benefits’—leveraging what the company already excels at (distribution, integrations, brand) rather than reinventing from scratch.
- •Use the strategy stack: higher-level choices funnel into more specific ones
- •Inherit decisions that already enable ‘no’; refine those that don’t
- •Apply the stupidity flip within your scope to sharpen choices
- •Leverage ‘Evergreen benefits’—existing company strengths—as the foundation for team strategy