At a glance
WHAT IT’S REALLY ABOUT
AI makes output cheap; strategy must force real product choices.
- Tim Herbig argues that in an AI era where feature production is dramatically cheaper, the critical constraint shifts to making the right strategic choices that drive adoption and business impact.
- He defines product strategy as a set of opinionated, testable choices that enable teams to say “no,” and he recommends Roger Martin’s “opposite isn’t stupid” test (“stupidity flip”) to validate whether real choices were made.
- Tim explains that AI alone is rarely a defensible value proposition because the underlying models are widely accessible, making brand and proprietary data the most durable differentiators for many products.
- He introduces practical tools—audience trees for narrowing target segments and Hamilton Helmer’s Seven Powers for evaluating differentiation—to help teams clarify playing field and winning moves.
- For measurement, he advises starting in plain English (“what will we do differently” and “how will we know it worked”), avoiding machine-centric AI metrics like tokens used, and selecting metrics that tightly map to strategic bets.
IDEAS WORTH REMEMBERING
5 ideasCheap feature output makes strategy more—not less—important.
As AI reduces the cost of shipping features, the penalty for shipping the wrong features increases—because you still must drive adoption and business impact. Strategy’s job is to force choices and provide a basis to evaluate whether those choices changed customer behavior and business outcomes.
If it doesn’t help you say “no,” it isn’t strategy.
Tim’s litmus test is whether a strategy helps teams and stakeholders intentionally say yes/no over a time horizon (e.g., 6–18 months). If it can’t meaningfully exclude options, it’s likely a set of aspirations, not a strategy.
Use the “stupidity flip” to pressure-test whether you’ve made real choices.
Use Roger Martin’s test: flip a strategy statement to its opposite; if the opposite is “stupid on its face,” you haven’t made a real choice (e.g., “make salespeople more efficient”). Decompose the strategy into components (audience, winning move, etc.) and flip each component until alternatives are plausible.
AI is a resource, not a moat; differentiation shifts to brand and data.
AI rarely functions as a standalone value proposition because competitors can access the same frontier/open-source models. Sustainable differentiation tends to come from slower-to-build moats like brand and proprietary data, with AI serving as an amplifier of an existing value proposition.
Nail the audience first—then everything downstream gets clearer.
Audience definition is a high-leverage strategic choice; Tim recommends making the target “comically narrow” enough to create focus, then sanity-checking market size. Audience trees help you systematically narrow by dimensions (region, buyer experience, car value, etc.), and AI can accelerate the research and sizing behind those cuts.
WORDS WORTH SAVING
5 quotesIf it doesn't, it's not a strategy.
— Tim
AI by itself in your strategy is not a value proposition. Everyone has access to the same frontier models. You can't compete on that technology alone.
— Tim
The point is, tokens used is not a value exchange for your customers. It's a value exchange for Anthropic.
— Tim
Well, the question is, is the opposite of your choice, of your strategy choice, stupid on its face? If the opposite is stupid, it's you haven't made a real choice.
— Tim
You will still be held accountable for the consequences of your strategy. You can't say, "No, ChatGPT told me this is a great strategic bet. It's not my f- problem that we lost ten million in the market."
— Tim
High quality AI-generated summary created from speaker-labeled transcript.
