← Blog

Data Visibility

AI Promises $230 Billion in Oil & Gas. The Problem Is the Business Model Blocks It.

The potential exists. The pilots work. And the value still escapes. An analysis of where to focus, how to scale, and why current contracts block the return.

Leo Giménez

Leo Giménez · CEO & Founder, Performy

Expert in sales and human behavior · Nov 22, 2026 · 8 min

The number everyone cites but few know where it lives

The potential value of AI in upstream oil & gas is estimated at $230 billion in capturable value, according to McKinsey analysis. It’s a figure that circulates at every industry conference and appears in almost every technology strategy presentation. The problem is that number rarely comes accompanied by the right question: where exactly is it, who captures it, and under what conditions?

The value isn’t uniformly distributed. Some use cases concentrate the vast majority of the possible return; most active projects in the industry are attacking the lower-impact ones. Not because the team is incompetent, but because the high-impact ones are the hardest to demonstrate, require the most integrated data, and — more often than not — create the most tension with existing contracts.

The real challenge isn’t technological. It’s the combination of three problems that are rarely solved at the same time: where to focus, how to scale from pilot, and how to structure the value agreement when the generated efficiency reduces the activity volume that many contracts remunerate.

The three fronts that concentrate most of the value

Not all AI use cases in upstream compete in the same league. The pattern observed across the industry shows a clear concentration in three fronts, each with very different return profiles, implementation complexity, and payback speed:

Subsurface & reservoir optimization

Reservoir modeling, well planning, and dynamic production management. The highest value potential in absolute terms — and also the greatest data complexity and the strongest dependency on interdisciplinary models.

Predictive maintenance & reliability

Detecting failures before they happen, optimizing intervention cycles, and reducing non-productive time (NPT) in critical facilities. The ROI is more demonstrable in the near term and easier to link to contractual metrics.

Drilling efficiency

Reducing days per well, optimizing parameters in real time, and proactively managing incidents. The fastest payback — and the most contentious for OFSE companies that bill by activity.

The decision of where to start isn’t trivial. Beginning with the front that’s easiest to demonstrate (predictive maintenance) makes sense for building internal credibility. But if the contractual model isn’t ready for the other two, the value capture ceiling is low from the outset.

Why pilots don’t become production

The capital-intensive asset industry has a well-documented pattern: AI projects that demonstrate results on one or two wells, then stall. “Pilot purgatory” isn’t an anomaly in this industry — it’s the default adoption pattern when there’s no active decision to scale.

The most common reasons a successful pilot never reaches production:

  1. Data fragmentation: production, geological, and operational data live in separate systems with inconsistent historical quality and no governance model to integrate them. The pilot worked with manually curated data; at scale, that process isn’t sustainable.
  2. Organizational silos: the person who decides on technology isn’t the one who operates the asset or the one who negotiates the contract. The lack of alignment between the digital team, operations, and commercial creates friction at every expansion decision.
  3. Misaligned incentives: the team that implements the AI doesn’t capture the upside if it works. The savings generated go to the operator’s P&L; the service provider, who put in the effort to implement it, receives the same rate as always — or less, if the efficiency reduced their billable volume.
The challenge isn’t proving that AI works. It’s proving it’s worth more than the organizational cost of adopting it at scale.

The contract knot: when efficiency reduces revenue

Here’s the hardest tension to resolve. A significant portion of OFSE (oilfield services and equipment) contracts remunerate activity: operating days, meters drilled, equipment hours. The model is simple and predictable — until AI enters to optimize that very process.

If AI enables drilling a well in 15 days instead of 20, the operator saves 5 days of day rates. The service provider lost those same revenues. Even if both parties recognize that the operational outcome is better — less risk, less NPT, better well quality — the service provider’s financial model deteriorates with every improvement AI produces.

This creates a structural disincentive that isn’t rational to ignore. OFSE sales teams bringing AI solutions to operators face a paradox: selling something that, if it works, reduces their own revenue base. This is a contract design problem, not a technology problem or a lack of conviction.

Selling something that — if it works — reduces your own revenue base isn’t a technology problem. It’s a business model problem.

How to rethink the shared value model

The teams solving this tension are migrating toward outcome-based models rather than activity-based ones. This isn’t a new concept in the industry — but AI gives it urgency because it makes the gap between value created and value captured more visible and faster to materialize.

Base rate + performance bonus

The service provider maintains a revenue floor and receives upside if AI generates measurable efficiency above an agreed baseline. Risk is shared; incentives align.

Risk/return models

The service provider invests in implementation and recovers a proportion of the savings generated above the operational baseline. Requires defining the baseline precisely — which turns every sales conversation into a technical agreement that must be documented.

Value-linked contractual KPIs

NPT, days per well, cost per equivalent barrel, equipment availability — outcome metrics that replace activity metrics as the basis for remuneration.

The common denominator across all these models is the same: someone has to have agreed, with precision, what success means, how it will be measured, and over what timeframe. That agreement doesn’t emerge from a legal contract — it emerges from sales and implementation conversations.

Data visibility as a starting point, not a destination

Before sharing value, you have to measure it. And to measure the value committed in a complex sale, you need to track what was promised, how success was agreed to be measured, and which conversations defined those parameters. Performance reviews 12 or 18 months after a close depend on agreements that, in most teams, weren’t recorded with any precision.

When the success of a contract depends on demonstrating that AI reduced NPT by 18% above an agreed baseline, and nobody remembers exactly what was defined as the baseline in that meeting from Q3 of the previous year, the value case becomes a dispute instead of an expansion conversation.

This problem compounds in deals with long cycles and multiple stakeholders. The champion who understood the value proposition in depth may have changed roles. The operations manager who agreed on the baseline KPIs may not be in the annual review. Without visibility into what was said, when, and by whom, the team depends on collective memory — which is rarely enough when the number being defended has eight figures.

The value case isn’t defended at the annual review. It’s built in every conversation from the first contact.

Why this shouldn’t be managed without conversation intelligence

All of the above works as a framework. The underlying problem is that applying it requires visibility into every conversation where the value case was defined, negotiated, and agreed. And that visibility, in most commercial teams selling complex solutions, simply doesn’t exist: it depends on what a rep jotted down, what someone remembers, and what someone had time to log in the CRM after a 90-minute call.

That’s exactly what Performy runs automatically across 100% of calls. Every conversation where the value baseline is defined, where the buyer states which KPIs matter to them, where success criteria are agreed — it’s all captured, structured, and available. You don’t depend on the rep’s notes. You don’t lose context when a stakeholder changes. And when the performance review comes around, you have evidence of every commitment rather than a debate about what “was supposedly agreed.”

📋

Map your complex deal value case before it’s too late

45 minutes, your team, the five most important deals this quarter. Identify where the promised value lives, how the current contract measures it, and what conversation is missing to close that gap — before the next performance review.

Leo Giménez

Leo Giménez · CEO & Founder, Performy

Expert in sales and human behavior. Author of “Stop trying to be someone.”

Connect on LinkedIn