AI Is Expanding What Marketers Know. It Is Also Expanding What They Cannot See

Most research on AI in marketing asks how firms should use the tools. Sidney T. Anderson, associate professor of marketing, asks a different question: how does AI change what marketing can know at all? His answer is that AI is not just a faster way to gather information. It reshapes the boundaries of marketing knowledge, expanding what firms know and what they are ignorant of at the same time.

The map

The paper sorts marketing knowledge into four quadrants. Known knowns are the documented base: segmentation models, lifetime value formulas, media response curves, pricing rules. Known unknowns are recognized gaps that research programs and test-and-learn efforts are built to close. Unknown knowns are the tacit assumptions and craft knowledge that guide practice without ever being written down. Unknown unknowns are the disruptions no one can prepare for because no one knows what question to ask.

AI affects each quadrant differently. For known knowns, it automates and scales, now reaching into tasks such as copywriting and customer service that once required human judgment. For known unknowns, it accelerates exploration, but each answered question tends to spawn several new ones, often faster than an organization can validate or act on them. For unknown knowns, large language models can put an organization’s implicit logic into words, which makes hidden assumptions open to challenge, though a fluent explanation is easy to mistake for an accurate one.

The paradox sits in the fourth quadrant. AI surfaces unknown unknowns by detecting patterns people miss, and it creates new ones through emergent behavior. Pricing algorithms have produced collusive outcomes with no firm intending them. Recommendation systems shape preferences instead of matching them. AI agents acting for consumers now transact with AI systems acting for firms, producing market dynamics beyond any participant’s view.

Why the usual playbook falls short

Traditional marketing strategy treats uncertainty as something to reduce. Gather more information, convert unknowns into knowns, and performance improves. The paper argues that this logic holds in stable environments and breaks under AI disruption, where many uncertainties come from feedback loops that cannot be resolved by more data and where structural breaks make past patterns unreliable guides.

Drawing on Nassim Taleb’s idea of antifragility, the paper describes three kinds of marketing organization. Fragile firms optimize around known knowns and become more dependent on current patterns as they automate. Robust firms prepare for known unknowns through scenario planning and contingencies, which works when the range of possible disruptions is bounded. Antifragile firms position to gain from volatility regardless of its form, through optionality, continuous experimentation, and bets with limited downside and open upside.

Reading your own organization

The paper offers diagnostic signals. Knowledge-optimization cultures emphasize long-range planning, detailed forecasts, specialized expertise, standardized processes, and efficiency metrics. Antifragile cultures emphasize rapid iteration, skills that transfer across contexts, structures that reconfigure as conditions change, and metrics that reward responsiveness. Most firms show both, and the paper does not ask managers to pick one. The skill it asks for is recognizing which knowledge situation the firm is in and applying the orientation that fits.

There is an ethical edge to this as well. Firms that cannot recognize the unknown unknowns their own AI systems generate push the resulting risks onto customers, competitors, and regulators. Building adaptive capacity is therefore a governance obligation as much as a competitive one.

The framework is conceptual and its four propositions await empirical testing. The paper also notes that antifragility may favor large firms with slack resources, a concern that AI could accelerate market concentration. For managers, the immediate takeaway is a habit of mind: before investing in AI, ask which quadrant the problem lives in, and whether more knowledge is even the right goal.

What it means for managers

• Match the tool to the type of unknown. Automate what you know, research what you know you do not know, and stop expecting research to resolve what you cannot yet name.
• Over-optimization is a risk. Firms that automate existing processes deepen their dependence on today's patterns and face larger adjustments when those patterns break.
• Build for optionality. Rapid experiments, transferable skills, flexible structures, and metrics that reward responsiveness let a firm gain from volatility rather than absorb it.

Anderson, S. T. (2026). Beyond the knowledge frontier: An epistemological framework for marketing strategy in the age of artificial intelligence. Journal of Macromarketing. 10.1177/02761467261471796