Beyond P10/P50/P90: AI, Vectored Uncertainty and the Future of Investment Decisions

The Problem: Knowing the Range Is Not the Same as Knowing the Direction

Investment decisions have traditionally relied on deterministic economics: forecast production, prices, costs and cash flow, then calculate NPV, IRR, VIR, payout and breakeven value.

We know these forecasts are uncertain, so we add sensitivities and Monte Carlo simulation. Instead of one NPV, we obtain P10/P50/P90 outcomes and probabilities of meeting investment hurdles.

But knowing how widely NPV may vary does not tell us which way the world is moving—or why.

I think of this distinction as scalar versus vectored uncertainty.

• Scalar uncertainty asks: How much might the outcome vary?

  
  

• Vectored uncertainty asks: What plausible future are we moving toward, what forces are driving us there, and what would tell us that direction is becoming more likely?

The Need: Model Coherent Futures, Not Just Variables

Consider two scenarios with $100/bbl oil. One results from strong global economic expansion; another from geopolitical supply disruption.

The oil price may be identical, but inflation, interest rates, drilling costs, demand, political risk and economic growth could be dramatically different. Consequently, acquisition value may also be different.

This is where Generative AI offers an intriguing capability.

AI can combine economic conditions, interest rates, inflation, employment, VIX, energy security, geopolitics, climate policy, commodity markets and industry fundamentals to generate several internally consistent, credible scenarios.

Importantly, it can explain the causal logic behind each scenario rather than simply producing another number.

From Scenarios to Signposts

Each scenario should contain observable signposts.

A recession scenario, for example, might be preceded by tightening credit, declining PMIs, increasing VIX, weakening employment, rising inventories and downward revisions to energy demand.

An AI-enabled dashboard could continuously track these indicators and estimate something such as:

Expansion: 48% | Recession: 37% | Supply Shock: 15%

These are not certainties. They are changing assessments as new evidence arrives.

Each scenario can then feed the investment model, recalculating NPV, IRR, VIR, payout, breakeven value and bid strategy. Monte Carlo can still operate within each scenario, quantifying uncertainty around that particular future.

The Opportunity: From Valuation to Decision Intelligence

The objective is not to replace deterministic economics or Monte Carlo. It is to add the missing dimension:

Deterministic valuation → Probabilistic uncertainty → AI-generated scenarios → Signposts → Dynamic investment decisions.

Such a platform could extend far beyond oil and gas acquisitions. Private equity, investment banks, infrastructure investors and corporate M&A teams all face the same fundamental question.

Not simply: “What is this acquisition worth today?”

But:

“Which future are we moving toward, how will we know, and what will this investment be worth if we get there?”

That may be where Generative AI becomes far more valuable than simply generating another forecast.