Applications to Marketing
This essay builds on my past writings and inputs from AIs (Claude and ChatGPT).
World models are rapidly emerging as the next frontier of artificial intelligence. Unlike large language models, which predict the next word in a sequence, world models predict what will happen next in the world—they build internal simulations of cause and effect. They imagine, forecast, and plan. If LLMs are powerful librarians of human knowledge, world models are flight simulators for decision-making. And when viewed through the lens of marketing, their impact could be nothing short of transformative.
For decades, marketing has been built on pattern recognition. We segment customers into groups, examine historical patterns, and optimise campaigns based on what has worked before. At its core, marketing has been fundamentally backward-looking: the past determines the next best action. World models flip this on its head. They move marketing from remembering to imagining—from projecting history to simulating futures.
- From Customer Segmentation to Individual Simulation
Today, personalisation is rooted in similarity: “customers who bought X also buy Y.” With world models, similarity becomes irrelevant. Instead of predicting behaviour because other customers behaved that way, the AI models the internal dynamics of each specific customer and simulates how that individual will behave in different conditions. This is methodological individualism applied to marketing—understanding each customer as a unique decision-maker, not as a member of a segment. It is no longer “people like her”—it is her.
- From A/B Testing to Counterfactual Experimentation
Instead of running experiments on customers, brands can run experiments inside the twin. A marketing team can test hundreds of interventions—timing, cadence, offer structure, creative style, channel mix—safely in simulation. The winning strategy reaches the customer only after the twin has proven it. This dramatically reduces AdWaste and profoundly raises confidence in execution. You’re not spending to learn; you’re learning to spend.
- From Pattern Matching to Causal Persuasion
Current marketing predicts that a runner will click on trainers because other runners did so before. World models understand why she runs—whether for stress relief, social belonging, or competition—and generate interventions that align with her motivation. Timing becomes as important as messaging; triggers become as critical as offers; subtle nudges outperform overt pitches. The power lies not in showing the right product but in activating the underlying decision driver.
- From Owned Channels to Adaptive Environments
Emails, apps, and websites become dynamic worlds that adjust themselves based on moment-to-moment interaction. A NeoMail is no longer a static message—it becomes a live space that responds to each user’s micro-behaviour, shaping content, rewards, and recommendations automatically. This shift turns the revenue tax paid to ad platforms into investment in owned media that learns and improves. A channel becomes a continuously adjusting environment rather than a delivery mechanism.
- From Designed Journeys to AI-Orchestrated Futures
Today’s customer journeys are flowcharts—courses built by humans and refined through hindsight. Tomorrow’s journeys will be discovered, shaped, and adjusted by the model itself in real-time. Every customer receives the journey most likely to maximise lifetime value—not because it was built manually, but because the model has simulated thousands of futures and chosen the best one. Marketing becomes proactive navigation, not reactive optimisation.
- From Cost Centre to Profit Engine Through Predictive ROI
World models change the role of marketing measurement. Dashboards today describe the past. World models allow marketers to preview the future—predicting churn before it happens, forecasting which Rest customers can be recovered economically, and projecting the impact of expenditure on attention and retention. This makes outcome-based pricing models like Progency feasible: you can guarantee results because you’ve already seen them work in simulation. Marketing becomes a profit engine with measurable, predictable returns, not a reporting function justifying past spend.
The Core Transformation
In short: LLMs answer questions; world models answer consequences. And that single shift moves marketing from reactive communication to proactive orchestration of customer behaviour. When combined with innovations like AI Twins, NeoMails, Co-Marketer agents, and the war on AdWaste, world models may herald the most consequential marketing transformation in decades—the shift from marketing to memories to marketing to imagined futures.
The question is no longer “what worked before?” but “what will work next?” That’s the difference between looking in the rear-view mirror and having a windscreen that shows you the road ahead before you drive it.
**
I asked the AIs to bring this future to life via a story about Ria (customer) and Maya (marketer).
