McKinsey Says AI Can Unlock $230 Billion in Upstream Oil and Gas. We Agree. Here's What They Got Right, and What the Industry Still Gets Wrong.

McKinsey just published a detailed sizing of AI's value potential in upstream oil and gas. The headline number: $230 billion at full potential, with $65 billion available today using proven technology.

We have read the full report carefully, and have been building and deploying agentic AI for well delivery and capital operations for years. We read it because when McKinsey publishes a number like $230 billion, the rest of the industry pays attention. And attention is long overdue.

Here is our take on what McKinsey got right, what the industry still gets wrong, and where we see the real path forward.

They Are Right: This Is a Concentration Play

McKinsey's most important finding is not the dollar figure but the shape of the opportunity. Their analysis shows that the top 10 use cases drive nearly half the total value, and the top 60 capture roughly 95 percent. The long tail of AI experiments adds very little.

This matches what we see in the field every day. The companies that are capturing real value from AI are not the ones running 40 pilots across 40 departments. They are the ones that picked two or three high-impact domains and went deep: drilling, production optimization, and risk intelligence.

At SwarmLens, we made that bet early. AssureLens is purpose-built for well delivery and capital operations because that is where the data is densest, the decisions are most consequential, and the cost of getting it wrong is measured in millions per well.

They Are Right: Agentic AI Changes the Game

McKinsey specifically calls out agentic AI as the technology that can orchestrate multistep workflows: diagnosing anomalies, checking equipment history, verifying parts availability, and executing across those steps with human oversight.

This is what AssureLens does today.

Our platform deploys domain-specific, multi-agent AI swarms that ingest thousands of drilling records, surface recurring risks, track which issues were resolved versus which resurfaced, and flag lessons that were never carried forward between wells. In a recent three-well drilling campaign in the Middle East, AssureLens processed 600+ daily and end-of-well reports and delivered 400+ traceable, decision-ready insights in under one week. The same process would have taken 6 to 8 weeks manually.

The key word in McKinsey's framing is "bounded workflows with clear controls and the right human-in-the-loop approvals." We agree completely. Agentic AI without domain expertise is a liability. That is why every insight AssureLens generates is traceable to its source document and validated by subject-matter experts before it reaches a decision-maker.

Where the Industry Still Gets It Wrong: The Knowledge Problem

McKinsey frames the challenge around scaling, commercial models, and workflow redesign. Those are real barriers. But there is a more fundamental problem the report touches on without fully naming: the industry's institutional knowledge is disappearing.

The engineers who built and operated fields for decades are retiring. The reports they wrote sit in filing cabinets and SharePoint folders that nobody reads. The lessons they learned the hard way get relearned the hard way by the next generation, on the next well, at the next rig. This is a serious knowledge preservation crisis. Which happens to be exactly the problem AssureLens was built to solve.

Our partnership with rp², a performance consultancy with over 45 years of drilling expertise and proprietary performance data, exists for this reason. We are not just building AI that processes documents but a system that captures the operational wisdom of an entire generation and makes it available to every team, on every well, at the moment the decision is being made.

Our ambition is straightforward. Turn knowledge into measurable business value for our clients.

— Harish Pillai, CEO, SwarmLens

What McKinsey Gets Right About Commercial Models (and Why It Matters for OFSE)

The report makes a critical point about the OFSE paradox: AI-driven efficiency can reduce the very activity that service contracts reward. Fewer rig days, fewer interventions, fewer billable events. McKinsey estimates that $17 billion of oilfield services revenue could be exposed at current deployment levels. Their solution is outcome-based commercial models: gain-sharing, performance bonuses, SaaS licensing.

We agree, though there's one thing we'd like to add. The companies that will win this transition are the ones that bring both the AI platform and the domain expertise. A technology vendor without field credibility will not convince a drilling superintendent to change how they plan a well. A consultancy without scalable technology won't deliver insights fast enough to matter.

That is why SwarmLens exists at the intersection of both. Our technology is built by people who have worked in the field. Our AI is trained on the data that field teams actually produce. And our commercial approach is designed around the value we create, not the hours we bill.

The Bottom Line

McKinsey's $230 billion figure is useful because it gives the industry permission to invest seriously in AI. But the number means nothing if companies keep spreading their bets too thin, underestimating change management, and treating AI as a technology project rather than an operational transformation.

The companies that will capture value from AI in upstream are the ones that:

1. Concentrate on the domains that matter most. Drilling, well delivery, production, and risk intelligence. Not everything at once.
2. Deploy agentic AI with real domain expertise baked in. Models without operational context produce noise, not intelligence.
3. Preserve and activate institutional knowledge. The biggest risk is that the knowledge AI needs to learn from is lost before anyone captures it.
4. Build commercial models that reward outcomes. If efficiency is punished by reduced revenue, nobody will adopt the tools that create it.

We are already doing this. AssureLens is live, field-validated, and deployed with operators who are turning decades of archived knowledge into a continuous learning and execution loop. If you want to see what $230 billion worth of opportunity looks like when it is actually being captured, talk to us.