Shell AI Strategy: AI in the Cost Plan, Savings Not Yet Banked

Shell treats AI as a lever in its cost plan. CEO Wael Sawan names AI, and agentic AI in particular, as one of the ways Shell will reach the top of its target to cut $5 billion to $7 billion of structural costs by 2028. He also says Shell is not yet counting on those savings. Its AI record in operations goes back to 2018, and it is now adding agents to it.

The CEO ties AI to cost. On the fourth-quarter 2025 call in February 2026, Sawan said he expects Shell to hit the higher end of the cost target, adding: “And AI is one of those key elements. Agentic AI is one of those elements.” He was also candid about the limits: “We are not yet banking all sorts of cost reductions coming out of agentic AI, because, to be honest, we’re still learning.”

AI is not written into the plan itself. The cost target dates from the March 2025 Capital Markets Day, which set cumulative structural cost reductions of $5 billion to $7 billion by the end of 2028, compared with 2022. The announcement did not mention AI. In the 2025 annual report, published in March 2026, AI appears as context: Sawan’s review cites “the rapid rise of artificial intelligence.”

The focus is operations and functions.

• Maintenance: AI monitors more than 13,000 pieces of equipment across Shell’s assets
• Subsurface: AI-assisted seismic interpretation and agentic AI for wells and production, with SLB
• LNG plants: real-time sensor data and AI models that adjust equipment settings for efficiency
• Corporate functions: Sawan said agentic AI is “playing up very much in our functional journey”

Outside, AI is a demand story. The 2025 annual report names demand from data centers, AI compute clusters and electrification as “a major demand driver” for US power in 2025. Shell Energy supplies about 140 gigawatt hours a year of renewable-matched electricity to Kao Data, a UK operator of data centers built for AI (July 2026).

Data is the precondition. Sawan credits “the investment we have been making in data cleanup over the past few years” and “the data-centric architecture that allows us to scale up AI’s benefit across the organization” (February 2026).

Partners build much of the industrial AI.

• C3 AI: has run Shell’s predictive maintenance program since 2018. In June 2026, the two agreed a multi-year expansion that adds AI agents for root cause analysis and remediation across asset operations, on Microsoft Azure.
• SLB: in December 2025, agreed to work with Shell on agentic AI across subsurface, well construction and production, using SLB’s Lumi data and AI platform.

Operations leaders describe a shift to scale. Mansi Madan Tripathy, chair of Shell’s companies in India, said in April 2026 that “AI and advanced digital technologies are becoming powerful drivers of productivity and competitiveness,” and that Shell is moving from pilots to company-wide use.

AI leadership sits in a specialist team. Amy Challen is Global Head of AI (AI Magazine, 2026). Dan Jeavons, who led Shell’s AI program for years, left in 2025.

Training is described only for specialists and managers. In January 2024, Challen said Shell had trained data scientists and technologists in AI ethics and compliance, and general managers and IT managers in using and assessing generative AI. She added that “our change management program isn’t there yet.” No company-wide training standard has been published.

The workforce is shrinking. Headcount fell from 103,000 at the end of 2023 to about 85,000 at the end of 2025, during the cost program. Sawan has named AI as one lever for cost, but Shell has not attributed job cuts to AI.

Governance is not described. The sources reviewed contain no published AI principles or oversight structure.

AI fluency, coded against the Lead with AI Fluency Matrix (September 2026 edition):

• Fluency floor: Not published. Training has been described for specialists and managers, not for the whole workforce.
• Highest level evidenced: Level 3 (Integrating), company-claimed. AI is built into maintenance, seismic interpretation and LNG plant operations. Agents for root cause analysis and remediation are being added, but the sources reviewed do not show AI running a process end to end, so Level 4 is not evidenced.
• Portfolio design: Partner-built industrial AI. C3 AI, SLB and Microsoft supply the platforms, a central AI team sets direction, and AI for the wider workforce is not described.

Shell does not report AI value on its own. AI sits inside the wider structural cost program, and Shell has not said how much of the savings come from AI:

• End of 2025: $5.1 billion of structural cost reductions since 2022 (February 2026)
• 2026 to July: a further $700 million, after which Sawan said Shell is “now halfway through that band” (July 2026)

The only AI value figure comes from a vendor. C3 AI’s president said the predictive maintenance program has produced “hundreds of millions of dollars in economic value” (June 2026). Shell has not confirmed a figure.

Candor is a strength here. Saying that agentic AI savings are not yet banked is more useful to investors than an unverified number. The next step is to report them, which is the move from AI adoption to AI value.

Next markers: whether Shell attributes part of its cost savings to AI, results from the C3 AI agents and the SLB work, and how the 2026 annual report describes AI.

Change log

• 2018: Shell begins its AI predictive maintenance program with C3 AI.
• March 2025: The Capital Markets Day sets a $5 billion to $7 billion structural cost target for 2028.
• July 2025: Dan Jeavons, who led Shell’s AI program, leaves.
• December 2025: Shell and SLB agree to collaborate on agentic AI.
• February 2026: Sawan names AI and agentic AI as elements of the cost program, with savings not yet banked; $5.1 billion of cost reductions achieved.
• March 2026: The 2025 annual report cites the rise of AI and AI-driven power demand.
• June 2026: C3 AI and Shell expand predictive maintenance with AI agents.
• July 2026: Shell Energy renews its supply deal with AI data center operator Kao Data; a further $700 million of cost reductions is reported.

• Disclosure: 2 of 4. Shell answers where it is going, a leaner company with AI in the cost plan, and where it will focus. What it will take is answered only for data and partners, with no training or governance disclosure. How it will know is not answered: AI savings are not separated from the wider program.
• AI is a cost lever, named by the CEO as part of reaching a $5 billion to $7 billion target.
• Shell is candid that it is not yet counting agentic AI savings.
• Industrial AI is long-running and partner-built, with predictive maintenance in place since 2018.
• Headcount is falling, and Shell has not linked the cuts to AI.
• The say-do gap is disclosure. A company that ties AI to its cost plan does not yet report what AI contributes, or how it governs and trains for it.
• Watch the cost reports. The test is whether AI gets its own line in the savings.

What the numbers show: Revenue fell 30% from 2022 to 2025 as oil and gas prices came down from their 2022 peak. Headcount rose to 103,000 in 2023, then fell to about 85,000 by the end of 2025 as Shell cut costs and reshaped its portfolio, so revenue per employee fell 23% over the period. In 2025 alone, headcount fell 11% and revenue 6%, so revenue per employee rose 6%. Shell names AI as one lever in its cost program but has not attributed job cuts to it. These figures do not show that AI caused any change.

Revenue, headcount and revenue per employee, FY2022 to FY2025

• FY2022: $381.3bn revenue, 93,000 employees, $4.10M per employee
• FY2023: $316.6bn revenue, 103,000 employees, $3.07M per employee
• FY2024: $284.3bn revenue, 96,000 employees, $2.96M per employee
• FY2025: $266.9bn revenue, 85,000 employees, $3.14M per employee

Change, FY2022 to FY2025: revenue down 30%, headcount down 9%, revenue per employee down 23%.

Notes: Revenue is as reported. Headcount is employees at year-end, rounded to the nearest thousand; 2025 includes portfolio companies. Revenue moves mainly with oil, gas and refined product prices.

These figures give context on scale and productivity. They do not show that AI caused any change. Sources: Shell 20-F filings and annual reports.

What is Shell’s AI strategy?

Shell uses AI mainly to lower costs and improve operations. CEO Wael Sawan names AI and agentic AI as elements of its plan to cut $5 billion to $7 billion of structural costs by 2028, and Shell uses AI for predictive maintenance, seismic interpretation and LNG plant optimization.

‍

Which companies does Shell work with on AI?

C3 AI has run Shell’s predictive maintenance program since 2018 and in June 2026 expanded it with AI agents, on Microsoft Azure. SLB is working with Shell on agentic AI for subsurface, well construction and production.

‍

How much has Shell saved with AI?

Shell has not said. It reported $5.1 billion of structural cost reductions by the end of 2025, and Sawan said in February 2026 that it is not yet counting savings from agentic AI.

‍

Is Shell cutting jobs because of AI?

Shell has not attributed job cuts to AI. Headcount fell from 103,000 at the end of 2023 to about 85,000 at the end of 2025 during its cost program.

‍

How is this profile put together?

Lead with AI profiles use public sources only, such as annual reports, earnings materials, filings and on-record interviews, dated at the time of writing. AI fluency is coded against the Lead with AI Fluency Matrix. Company figures are marked as company-claimed unless they have been independently verified.