Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

How human AI simulation unlocks decision intelligence

Learn from a new opportunity in experiential and strategic decision-making.

[Photo: Getty Images]

Profile photo of Chad Reynolds

If the word “simulation” feels too sci-fi for you, that’s okay. But you’ve likely been carrying out simulations all along. You’re playing out your ideas in your head or through focus groups, or spreadsheets, or learning through iterative doing.

But you’re discovering what worked only after you launch the campaign and deploy your capital. 

New AI environments allow teams to simulate scenarios and observe how populations of AI-powered consumer personas react, make decisions, and influence one another. Intelligence from how possible decisions unfold.

How we deployed a simulation

In our work with a global ice cream manufacturer, we used primary consumer research that they had collected. We developed a population of AI consumers that reflected their key segments across flavor profiles, occasion states, and brand affinity.

Instead of debating new product concepts and scenarios, the team simulated a world where thousands of consumer agents reviewed new product ideas, shared feedback, and synthesized recommendations. Each agent was a unique persona trained on real consumer inputs, representing a distinct global market with cultural nuances and wrestling with rising costs while trying to feed their family.

The result: An end-to-end simulation of product innovation across three global markets in under two hours. The deliverable population could explain why they scored ideas the way they did, which themes stood out to each of them, and the specific reasons why.

Team members can test how different consumers respond to pricing, messaging, distribution channels, and cultural framing. One direction might accelerate adoption in urban markets but stall in secondary cities. Another might resonate emotionally with younger demographics but create operational pressure. A third might reveal an unexpected use case that reshapes positioning.

Thousands of micro-decisions compound into patterns and reflect how word-of-mouth might spread differently across geographies, channels, and audiences. By the time leaders decide what to ship to the real world, they’ve already experienced hundreds of plausible futures. Their decisions reflect experience (seeing it happen) rather than speculation (imagining what might happen).

5 architectures of experience

Five practical capabilities allow teams to play out future scenarios. In these environments, each capability informs the next.

1. People layer: The foundation of human input relies on real-life communities, fans, experts and stakeholders providing behavioral signals, cultural energy, and lived experience grounded in reality. Authentic human insight reflects how real populations think and adapt over time. People steer the world and offer ways to explore scenarios, observe interactions, and impact the simulation.

2. Data layer: This secondary layer draws on customer research, database records, sales data, search behavior, call-center themes, product reviews, competitor activity, and social sentiment. It seeks the patterns in what people say and feel publicly about a brand, category, or cultural moment.

3. Agent layer: Agents operate as active simulation participants. They collect research, scan for patterns, generate hypotheses, compare possible decisions, and stress-test ideas at a scale no human team could manage manually. The agents take on the identities of your audience, your retail stores and your channels. They transform data into AI forms that can use tools and models, and interact with one another.

4. Model layer: The model layer gives agents the behavioral and cultural intelligence to interpret what is happening and why. This goes beyond LLMs like GPT, Gemini, and Claude that reason to generate the next word. The new People Models simulate human behavior, providing agents with unique identities that understand how a specific human being thinks, identifies, evolves and responds to change.

5. Workflow layer: The final layer connects all of the parts, moving exploration into action. A goal enters the system, such as testing whether a new product should lead with taste, price, sustainability, or convenience. One agent refines the strategy, another generates creative options, another simulates a population of likely market responses. The output could be a sharper brief, a prioritized concept, or a launch recommendation.

Simulating an experience through these five layers generates understanding, and decisions develop from the capacity to explore, generate hypotheses, and refine their direction through simulated experience rather than estimated instruction.

Close the reality gap

Simulation builds continuously updating shared-world models based on layers of real human behavior, cultural context, and market signals. These systems understand what happened and why, and can adapt faster than any plan we could write.

For example, a furniture retailer might set a goal of deciding whether a new campaign should emphasize low prices, durability, room-by-room inspiration, or faster delivery.

One agent could refine the strategy by comparing those directions against customer pain points, purchase triggers, and competitive positioning. Another could turn the strongest directions into creative territories or sample headlines. A third could simulate market response by showing how first-time apartment renters, growing families, and value-conscious homeowners might react differently to each idea.

The finished prioritized campaign brief could lead with price for immediate sales, supported with family-friendly marketing, and reserve room-inspiration creative for social and upper-funnel sales channels.

Applied in real-time

The pattern is already beginning across different categories. In a CNBC interview, Bob’s Discount Furniture’s CEO credited “synthetic focus groups” as key to the company’s growth and market outperformance, speaking from the stock exchange floor on the occasion of the company’s IPO. SpaceX tests Starship trajectories through thousands of simulated scenarios, each iteration revealing failure modes that inform the next physical prototype.

The question is moving from “What do we think will happen?” to “What did people experience when we simulated it?”

Before the next campaign brief is approved, marketers should run one simulated decision sprint. Choose the audience that matters most, feed the system the creative territory, offer, channel plan, and budget assumptions, then use the agent responses to identify which idea deserves investment, which needs refinement, and which should be cut before media dollars are spent.

This isn’t SimCity.

The question now is:  “What world do we want to build?”

Chad Reynolds is the founder and CEO of Vurvey Labs.