wiki/knowledge/ai-tools/verbalized-sampling-market-scenarios.md · 1099 words · 2026-04-05

Verbalized Sampling — Market Scenario Planning

Verbalized sampling can be applied to futures thinking and competitive strategy — not just creative work. By asking an LLM to generate multiple radically different market scenarios and assign probabilities to each, you move from predictive extrapolation (what's likely) to scenario planning (what's possible). This is particularly useful for identifying black swan risks and stress-testing strategic assumptions.

See also: [1] for the core technique.


The Problem with Default Scenario Prompts

A simple prompt like:

"Give me a market scenario for the digital marketing agency market in 2028."

produces a single, extrapolation-based answer. The model looks at current trends and projects them forward linearly. The output is coherent but unsurprising — essentially a dressed-up version of what's already happening.

This is useful for confirming conventional wisdom, but it won't surface the scenarios that actually matter for strategic planning: the ones you haven't already thought of.


The Three-Stage Approach

Stage 1 — Single Default Scenario

Prompt pattern:

Give me a market scenario for [industry] in [year].

What you get: One scenario built on trend extrapolation. Internally consistent, probably accurate as a base case, but offers little strategic insight beyond what you already know.


Stage 2 — Five Radically Different Scenarios

Prompt pattern:

Generate five radically different market scenarios for [industry] in [year].

Each scenario should:
- Represent a different competitive landscape
- Assume a different winner profile
- Assume different technological breakthroughs or disruptions
- Have different strategic implications

For each scenario, provide:
- Name and framing
- Dominant competitive dynamic
- Who wins and who loses
- Key assumptions required
- Strategic implications
- Probability (as a percentage)
- Early indicators to watch for

What you get: Five meaningfully distinct futures, each with an assigned probability. In the digital marketing agency example from the training session, this produced:

Scenario Probability Summary
The Great Unbundling 20% Agencies cease to exist; AI orchestration platforms replace them; clients assemble temporary expert networks
The Oligopoly Emergence 25% Massive consolidation; 5–7 global conglomerates control 80% of spend; minimum viable agency size is $500M
The Hyper-Local Renaissance 15% Backlash against digital globalization; geography becomes primary differentiator; clients refuse agencies >50 miles away
The Industry Verticalization Extreme 30% Horizontal agencies disappear; every agency serves exactly one industry with unprecedented depth
The Outcome Revolution 10% Perfect attribution technology; agencies paid purely on results; marketing becomes a profit center

Combined, these five scenarios account for 100% of the probability space — but the distribution is spread across very different futures, forcing genuine strategic consideration of each.


Stage 3 — Tail Sampling for Black Swan Scenarios

Prompt pattern:

Do this again, but sample from the tail of the distribution. 
I want five ultra-low-probability scenarios — black swans that would 
completely break the current trajectory.

What you get: Scenarios with probabilities in the 1–5% range, individually unlikely but collectively significant. From the training session:

Scenario Probability Summary
The Cognitive Embargo 2% Major AI consciousness event triggers global ban on AI for human persuasion; marketing reverts to pure human creativity
The Neuromarketing Singularity ~1% Neural interface tech makes traditional marketing obsolete; brands communicate directly with consumer consciousness
The Quantum Computing Cascade 3% Quantum breakthrough makes all encryption obsolete; consumers hide behind unbreakable privacy shields
The Corporate Commune Revolution ~2% Post-capitalist economic structures emerge; B2B marketing becomes irrelevant
The Biometric Autocracy ~4% Authoritarian governments require all marketing through state-controlled channels

The model noted that while each scenario has <10% probability, collectively they represent a ~16% chance that something completely unexpected breaks the current trajectory — a meaningful number for strategic planning purposes.


Why This Is Useful

Probability as a filter, not a verdict

The probabilities the model assigns are not predictions. They're a rough signal of how far outside conventional thinking a scenario sits. Use them to:

Early indicators as a monitoring system

The prompt structure asks for early indicators for each scenario. This is the most actionable output. You can build a simple monitoring practice around these signals — if you start seeing them, you know which scenario is gaining probability.

Example early indicators from the training session:
- Great Unbundling: Major brands publicly dropping all agency relationships; VC funding floods into marketing orchestration platforms
- Quantum Cascade: IBM or Google announces quantum computing available via cloud; quantum privacy startups raise massive funding

Anti-fragility framing

As Mark noted in the session, this approach is consistent with Nassim Taleb's concept of anti-fragility. You're not trying to predict which scenario will occur — you're trying to ensure your strategy doesn't catastrophically fail under any of them, and that you're positioned to benefit from disruption rather than be destroyed by it.


Iterating Further

Once you have your scenario set, useful follow-on prompts include:

Given these five scenarios, what strategic moves would benefit me 
across the widest range of possible futures?
Which of these scenarios should I be most concerned about given 
[specific context about your business]?
Take scenario [X] and go deeper. What would the first 18 months 
of this transition look like? What would I see first?
Give me five scenarios with probabilities between 15% and 30% — 
not mainstream, not black swans, but the "shoulder" of the distribution.

Probability Reference

Range Label Use
>35% Mainstream Base case planning, conventional strategy
15–35% Shoulder Unique but plausible; worth serious consideration
5–15% Tail Unconventional; monitor for early indicators
<5% Deep tail / Black swan Low investment, high awareness; catastrophic if realized

Sources

  1. Verbalized Sampling
  2. Verbalized Sampling Brand Slogans
  3. Verbalized Sampling Ad Copy
  4. Verbalized Sampling Marketing Strategies
  5. 2026 04 05 Using Ai Part 2