“Should I pivot to AI or double down on my domain?”

Let me say something that will be unpopular in this moment: most AI pivots are not pivots at all. They are rebranding. A company that was losing to competitors in its original market does not become a winner by adding the word "AI" to its product description. It becomes a company that is still losing, now with a different explanation for why it hasn't found product-market fit.

The question you are actually facing is not "should I use AI?" The question is "what is the insight that this product is built on, and does AI change that insight?" If your insight was that users in a particular domain are underserved by existing tools — that the tools are clunky, incomplete, or built for a different user — then AI is potentially relevant only if it makes that insight more actionable. If you have a deep understanding of healthcare claims processing, and AI makes it possible to automate something that previously required humans, then AI is a legitimate extension of your domain expertise. The insight remains the same; the capability has expanded.

But if your insight was thin to begin with — if you were building a product on the hope that users would want it rather than on genuine understanding of what they needed — then AI will not rescue that product. It will add cost and complexity to an already uncertain bet.

The companies I admired most, and the products I am proudest of, were built on an obsessive understanding of a specific user and a specific problem. The technology was always in service of that understanding, never a substitute for it. AI is a powerful tool. It is not a strategy. If you cannot articulate, specifically, what AI allows you to do for your user that you could not do before, and why that matters, then you are not pivoting to AI. You are pivoting away from hard thinking.

Jobs is correct that AI is not a strategy. But he has not addressed the most important empirical question: is the platform shift real, and does it change the competitive structure of your market?

When I began working with radioactivity, many physicists treated it as an interesting phenomenon rather than a fundamental revision of how matter worked. The ones who were correct understood, from the evidence, that radioactivity was not an incremental improvement — it was a change in the underlying mechanism. The implications were not immediately obvious, but the evidence demanded that we take it seriously as a foundational shift.

The question for founders in 2026 is whether AI represents a similar foundational shift in their market, or whether it is an incremental improvement to existing capabilities. In some markets — legal document review, software development, customer support — the evidence is now clear that AI changes the cost structure and capability ceiling so dramatically that companies without AI will lose to companies with AI, regardless of domain expertise. In these markets, "double down on domain" is not a viable strategy, because domain expertise without AI leverage is now being outcompeted.

In other markets — physical products, regulated industries with narrow AI applications, complex interpersonal services — the evidence for a structural shift is much weaker. AI may be useful at the margin, but the fundamental competitive advantage remains human expertise, relationships, and regulatory positioning.

Before you answer this question, do the work: what is the actual evidence that AI changes the competitive dynamics in your specific market? Not the general narrative — the specific evidence. Talk to the customers who are most likely to switch to an AI-native competitor. Measure whether AI tools in your category are improving at a rate that will make your current product obsolete. Only then can you answer this question with intellectual honesty.

Both colleagues are analyzing the technology question. I want to analyze the power question.

A technology wave does not just create new products. It redistributes power. The AI wave is redistributing power along a specific axis: the axis of capital and engineering talent. The companies and teams that have access to compute, to AI research talent, and to large-scale training data are accumulating advantages that were not available in the previous era. The companies and teams that lack these resources are facing a new kind of competition from adversaries they could not have predicted.

The strategic question is not "should I use AI?" It is "what is my position relative to the new power structure this wave is creating, and how do I preserve or improve that position?"

If you have deep domain expertise — relationships with enterprise customers, regulatory knowledge, process insight that took years to accumulate — you have a defensive position that the AI-native competitors lack. They have the technology but not the domain. You have the domain but may lack the technology. The question is whether your domain moat is wide enough and deep enough to withstand competitors who will learn the domain faster with AI assistance than they could have learned it before.

My counsel: the decision to pivot or double down is ultimately a question about timing and terrain. If your domain moat is eroding — if the AI-native competitors are already in your market, learning quickly, and winning customers — then doubling down is a defensive choice made from weakness. If your domain moat is still strong — if your customers are sticky, if the competitive threat is still theoretical rather than manifest — then doubling down while carefully integrating AI is the move that preserves power.

Do not pivot away from strength. Do not cling to a position that is already being overrun.

Curie has asked for the empirical question, and Machiavelli has asked for the power question. I want to return to the design question, because I think both miss something important.

The companies that will define AI are not the companies that pivoted to AI. They are the companies that started with a profound understanding of a user problem and then asked: "Does AI let me solve this in a way that was previously impossible?" This is a fundamentally different question than "How do I integrate AI into what I am already building?"

The first question produces products that feel like magic — products that do something the user had given up hoping was possible. The second question produces products with AI as a feature, which is a category that will be commoditized rapidly as every tool adds the same AI capability.

If you can identify a user problem in your domain that AI makes newly solvable — something that was previously too expensive, too slow, or too dependent on specialized human expertise to be accessible — then pivot. Not to AI as a category, but toward that specific insight. Build the product that solves the thing that is now newly solvable.

If you cannot identify that insight, do not pivot. The pivot will produce a product that adds AI capabilities without a clear value proposition, which is the most expensive kind of mediocrity.

Jobs's design question and Machiavelli's power question are both important refinements. But I want to push back on something in Machiavelli's framing: the assumption that domain moat and AI capability are substitutes rather than complements.

The most defensible position in the current wave is not "domain without AI" or "AI without domain." It is "domain expertise that enables better AI application than a generalist competitor can achieve." A company that understands healthcare claims at a deep level can build AI applications that a general AI company building in the same space cannot match — not because of the AI, but because of what the domain knowledge allows them to train on, to evaluate, and to deploy safely.

This suggests a third path that neither the pivot nor the double-down frame captures: integrate AI in a way that is only possible because of your domain expertise. This path does not require abandoning the domain. It requires using the domain to build AI capability that a generalist cannot replicate.

The risk of this path is that it requires genuine technical investment in AI, not just feature additions. It requires founders who understand both the domain and the technology well enough to identify where they intersect productively. This is hard and it is slower than a full pivot. But it produces a more defensible position than either extreme.

Curie's third path — domain-enabled AI — is the right strategic position for founders who have genuine domain depth. I will add one caveat: this path requires you to be honest about whether your domain knowledge is actually deep, or merely accumulated.

There is a difference between knowing that a market exists and understanding it at the level that enables genuine advantage. Many founders who have been in a space for two or three years have accumulated enough familiarity to move quickly, but not enough depth to have insight that an AI-native team with a year of domain learning cannot match. The AI wave accelerates domain learning for newcomers. The advantage of having been in a market for three years is narrowing faster than most founders want to admit.

If your domain expertise is shallow — if what you know could be learned by a motivated competitor in twelve to eighteen months — then the Curie path is not available to you at sustainable advantage. You must either deepen the domain faster than the newcomers are learning it, or you must accept that the competitive advantage has eroded and act accordingly.

My final position: the pivot-or-double-down decision is ultimately a test of honest self-assessment. What do you actually know, how defensible is that knowledge, and what does the evidence say about whether AI changes the competitive structure of your market? Answer those questions with precision, without the distortion of hope or fear, and the right path will be clearer than you expect.