I’ve started reading Martin Eriksson’s Decision Stack1, trying to wrap my brain around what people truly mean when they say vision, strategy and goals. To me, these have always been such broad terms that they can feel wishy-washy, loose corporate talk, elusive even.
Recently though, my view has changed. When I look within my team and at the products we’re building, I keep coming back to exactly these questions: how do we make the right bets, why are we here, and how do we get to our goals? I’m trying to figure out how our AI strategy aligns with the broader company vision, how that vision feeds into the decisions we’re making and the products we’re building, and ultimately what metrics and insights we need in order to make the right bets.
Sometimes it feels like there’s a tension between AI strategy and innovation. Innovation calls for experimentation. It is about building tools that may be canned simply because we’re trying to learn things; it is about being ahead of the curve. AI strategy is about making the right calls at the right time, with the right metrics in place. This leads me down even more questions. How can you make a bet on a product when AI is changing so quickly? How do you avoid experimenting your way into a pile of wasted products no one uses? And how do you make bets today that are still relevant six months from now?
A recent Forbes article highlighted that over 550,000 apps were submitted to the Apple App Store last year, the highest number in nearly a decade, and most of them have no users and even less revenue. AI has collapsed the cost of making software. Before AI, the cost of making software lay in the implementation; now it lies in making the right decisions. 2
That’s why I’ve changed my stance on vision and strategy and all those big corporate words. If the products you’re building don’t align with your vision and your strategy, you’re just adding another app to an already oversaturated App Store. Gartner predicts that “over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls”, noting that most agentic AI propositions “lack significant value or return on investment”.3 If the return on investment for AI projects is contested, then it becomes increasingly important for businesses to truly figure out what they wish to achieve and how it aligns with the vision and strategy of the business. Building is cheap; making the right decisions is costly.
One common way people have framed this shift is to say that implementation is cheap, so taste is becoming the more crucial weight. But what does taste mean? I think it oversimplifies a complex idea. Taste implies an aesthetic element, something acquired, but it goes beyond that. It’s being creative and innovative, but also experienced: making sound judgements, having business sense, understanding corporate strategy and setting an ambitious vision. Taste, to me, means you’ve done the experimentation, you know what separates good from bad, and you understand how something fits into the bigger picture. Taste is about curating the right choices.
The things that get built aren’t necessarily the right bets to make; often they’re just an idea that happens to have a working interface. Now that AI can vibe-code a website, write a document, create a PRD, build a whole PowerPoint deck or analyse your spreadsheets, it has removed the friction of that early stage, the one where you haven’t figured it out yet. AI tools are now adding the first brush strokes on an empty canvas. Lenny discusses this in his most recent interview with Andrew Ambrosino and refers back to it as the primal mark, an idea that Justin M. Berg makes in an article over 10 years ago:4
The first brush stroke a painter makes on a blank canvas, known as the primal mark, is especially important because it shapes what the painter subsequently paints. In the same way, the initial raw material employees use in creative tasks can anchor how they build their emerging ideas, in ways that enable or constrain their novelty and usefulness. 5
Whenever we use AI, we have to be careful not to let it define our next brush stroke. If the first brush stroke costs nothing to make, yet shapes every stroke that follows, we should be aware of what we’re giving up. That’s why authors say to “kill your darlings”, and why Pixar implemented the “three pitches rule”, it’s about not getting married to any single idea. With AI, and especially with vibe-coded apps, professional looking documents, and fully summarised analysis, it’s easy to see the shiny output and accept it as the finished product, rather than part of the research, ideation and design phase. That’s why Eriksson’s Decision Stack spoke to me. It pulls you back to the bigger picture, where each brush stroke has to align with the vision, the strategy and the goals of the team.
- Eriksson, Martin. The Decision Stack: How Strategic Clarity Unlocks Organizational Momentum. Decision Stack, 2026. ↩︎
- Majic, Josipa. “The Apple App Store Is Flooded With AI Slop and Legitimate Developers Are Paying for It.” Forbes, March 24, 2026. ↩︎
- Gartner. “Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027.” Gartner Newsroom, June 25, 2025. ↩︎
- Rachitsky, Lenny, host. “OpenAI Codex Lead on the New Shape of Product Work | Andrew Ambrosino.” Lenny’s Podcast: Product | Career | Growth, Spotify, 28 June 2026, https://open.spotify.com/episode/4NV2PIuSxMl7x8pOtuv5j8. ↩︎
- Berg, Justin M. “The Primal Mark: How the Beginning Shapes the End in the Development of Creative Ideas.” Organizational Behavior and Human Decision Processes, vol. 125, no. 1, Sept. 2014, pp. 1–17. ScienceDirect, https://doi.org/10.1016/j.obhdp.2014.06.001. ↩︎