14 OCTOBER 2026 / ABOUT FIVE MINUTES TO READ

What changes when AI becomes part of the work?

Oil and gas has moved beyond AI announcements. Some systems adjust equipment. Others change how people plan, investigate and visit the field. The useful question is which part of the work improves—and whether the benefit survives outside the pilot.

Start with the operating decision.

Devon changes lift settings. Oxy changes field attention. ADNOC changes rig oversight. These are different operating choices, even when all three are described as AI.

Devon Energy Occidental ADNOC

Give each tool a clear job.

QB’s planning tool separates a numerical solver from the assistant that explains its output. Vickery uses document models to work through land records. A language model’s useful role depends on the task.

QB Energy / Quantum Vickery Energy / Quantum

Copy the measurement plan, too.

A rollout count, time saved and a production gain answer different questions. Before repeating a project, agree the eligible assets, comparison, review work and complete cost.

Devon Energy BPX Energy Burk Royalty Expand Energy

EXHIBIT 01 / THE WORK BEHIND THE AI

Each tool has a different job.

Choose a workflow. Read what the sources establish, then consider the next test. The proposed extension is separate from what the company reports.

Wells / Devon

scaled · source 2026-08-05

The information

Live well measurements

The operating tool

Specialist lift optimization and control

Language-model role disclosed

No language-model role established in the selected disclosure.

People’s part

Operating and override responsibilities need more detail.

A PROPOSED NEXT TEST

Help engineers investigate exceptions and follow unresolved problems. Keep equipment control in the specialist system.

Our proposal; not an announced company plan.

Read the complete project

Sources & limits

Devon Energy — deployment evidence
2026-08-05 · Operator
Devon’s operator-hosted Q2 2026 transcript (2026-08-05) says its closed-loop AI system autonomously optimized 1,000 wells in real time, 24/7; management said 200 wells in March and a path to broad deployment across >2,000 additional Permian wells, including first legacy Coterra wells. This supersedes the Q1 count of >850.

Devon Q1 2026 earnings call — Smart Gas Lift
2026-05-06 · Issuer-hosted earnings transcript
Prepared remarks p. 4; Q&A pp. 12–14; 21 October additional locator: PDF and printed p. 4; Clay Gaspar prepared remarks, three waves of AI

See the recorded source reviews

The eligible-well denominator, later lift-type mix, fleet-wide benefit and detailed control limits are not disclosed.

Field work / Oxy

scaled · source 2025-08-07

The information

Well data and exception signals

The operating tool

Sensors, automation and AI support routeless work

Language-model role disclosed

No language-model role established in this operating disclosure.

People’s part

Field crews respond to the wells needing attention.

A PROPOSED NEXT TEST

Assemble the history behind an exception, coordinate a response and check what happened.

Our proposal; not an announced company plan.

Read the complete project

Sources & limits

Occidental Q2 2025 Earnings Call Transcript
2025-08-07 · Issuer-hosted earnings-call transcript
Page 3, Vicki Hollub prepared remarks

See the recorded source reviews

The contribution of AI is not separated from sensors and automation. Dispatch logic and overrides are not described.

Drilling / ADNOC

scaled · source 2026-08-04

The information

Live information across the rig fleet

The operating tool

A shared remote drilling operations platform

Language-model role disclosed

The selected release does not isolate a language-model contribution.

People’s part

Engineers oversee operations and make decisions.

A PROPOSED NEXT TEST

Prepare shift handovers and assemble evidence across rigs. Check every recommendation against the operating systems.

Our proposal; not an announced company plan.

Read the complete project

Sources & limits

ADNOC and SLB Deploy AI Platform Across Over 120 Drilling Rigs to Strengthen Upstream Performance
2026-08-04 · Operator press release
Release body: paragraphs on RTOC deployment, workflow, operating scope and reported metrics

See the recorded source reviews

This establishes fleet oversight, not autonomous rig control. Downtime claims are not a recurring annual benefit.

Planning / QB Energy

operating · source 2026-04-30

The information

Terrain, candidate locations and engineering constraints

The operating tool

Geospatial processing and a numerical planning solver

Language-model role disclosed

An embedded assistant explains optimization output.

People’s part

Engineers review plans and trade-offs.

A PROPOSED NEXT TEST

Track changed assumptions and prepare comparable scenarios for review. Test decision quality as well as speed.

Our proposal; not an announced company plan.

Read the complete project

Sources & limits

Geospatial field planning for an unnamed Quantum portfolio operator
2026-04-30 · Microsoft customer story
Field development planning solution; Improved speed, scale, and precision; updated algorithm paragraph

Quantum 2026 Stakeholder Report
2026 · Sponsor stakeholder report
Pages 11–15; page 68 disclaimer

See the recorded source reviews

The solver and assistant have different roles. Faster planning does not establish better capital returns.

Land / Vickery

operating · source 2026-04-30

The information

Scanned deeds, leases and conveyances

The operating tool

Document recognition and structured extraction

Language-model role disclosed

Azure OpenAI reads records for ownership, heirship and lease availability.

People’s part

Difficult records go to human review.

A PROPOSED NEXT TEST

Trace each finding to the underlying record and route uncertain ownership chains for review.

Our proposal; not an announced company plan.

Read the complete project

Sources & limits

Land-title research at an unnamed Quantum portfolio company
2026-04-30 · Microsoft customer story
From the field to the back office and beyond — land-title research paragraphs

Quantum 2026 Stakeholder Report
2026 · Sponsor stakeholder report
Pages 11–15; page 68 disclaimer

See the recorded source reviews

Processing speed does not establish legal correctness. Accuracy, full costs and replacement of title opinions are not established.

Plants / Aramco

operating · source Undated; accessed 2026-09-27

The information

Gas-plant operating information

The operating tool

Reinforcement learning within process control

Language-model role disclosed

No language-model role established in this plant-control account.

People’s part

Detailed approval and override boundaries are not disclosed.

A PROPOSED NEXT TEST

Explain process deviations and prepare maintenance or engineering follow-up, within existing plant procedures.

Our proposal; not an announced company plan.

Read the complete project

Sources & limits

Aramco AI and Big Data
Undated; accessed 2026-09-27 · company announcement
Fadhili Gas Plant section

See the recorded source reviews

One plant. Operating envelope, interlocks, full costs and repeatability at other plants are not described.

These layers explain the selected workflow. They are not a complete technical architecture. An unspecified language-model role does not establish that the company has no other LLM projects.

EXHIBIT 02 / WHAT THE NUMBERS MEAN

An impressive number still needs context.

These four disclosures measure different things. The charts keep the unit, population and comparison beside the claim.

Devon Energy / Operator earnings-call disclosures

Separate rollout size from results.

March 2026200wells

May 2026850+wells

August 20261,000wells

Devon’s reported rollout grew from March to August 2026. March is retrospective; May is a minimum. Later scope includes legacy Coterra wells.

Before you use this number

The earlier 2–3% gas-lift pilot gain does not establish the result across the later 1,000 wells.

Ask for a fleet result, an eligible-well denominator and total intervention cost.

Sources & limits

See the recorded source reviews

The eligible-well denominator, later lift-type mix, fleet-wide benefit and detailed control limits are not disclosed.

Burk Royalty / Operator/vendor-coauthored paper

Keep the units beside the improvement.

Electricity · US$ / well-month · zero-based scale

Before660

After376

Field time · hours / month · separate zero-based scale

Before126

After18

Burk’s eight-month account reports lower electricity spending and field time falling from 126 to 18 hours per month.

Before you use this number

An integrated sensing, analytics and control system produced the reported changes. AI’s separate contribution and full costs are unresolved. Producer counts conflict between the paper and abstract.

Ask for the baseline tariff, comparable production, full system costs and continued results.

Sources & limits

Field Optimization Reimagined: full paper
2026 · Operator/vendor-coauthored technical paper
PDF pp. 3, 7–9; figure on p. 8

Pump-by-exception monitoring and control at a West Texas lease
2026 · Operator/vendor-coauthored technical paper
SWPSC 2026052 abstract; linked full paper PDF pp. 3, 7–9 (site configuration, eight-month results, economics and projections)

See the recorded source reviews

The sources conflict on well count. The paper’s profit and payback estimates do not establish full-cost net return.

BPX Energy / Operator/vendor-coauthored paper

The people who adopt may be different.

6% average oil increase on adopters

441wells in the pilot

?wells adopting recommendations

BPX’s abstract describes a 441-well pilot, but the oil result concerns wells that adopted recommendations. The adopter count is not disclosed.

Before you use this number

Applying 6% to all 441 wells would change the population. The measurement period and matched comparison are not established in the abstract.

Ask what happened on all eligible wells, including declined recommendations.

Sources & limits

Daily lift recommendations across a 441-well Permian pilot
2026 · Operator/vendor-coauthored technical paper
SWPSC 2026051 abstract: method, 441-well asset and reported results

See the recorded source reviews

Do not apply the 6% to all 441 wells. Adoption selection and the oil-result denominator remain unresolved.

Expand Energy / Operator/vendor paper and supplier-authored JPT account

A short result is still a short result.

70–139% cumulative gas increase

63–83 daysReported comparison period

0 days365 days

The shaded interval marks days 63–83 on a one-year scale. Annual estimates are reported separately.

The Expand account covers nine Haynesville wells across eight pads and compares cumulative gas over 63–83 days.

Before you use this number

More than 80 MMscf per well per year is an estimate. The short-period percentage is not an annual or fleet-wide result.

Ask for longer follow-up and well-level comparisons under similar liquid-loading conditions.

Sources & limits

Autonomous liquid unloading on nine Haynesville gas wells
2026 · Operator/vendor-coauthored technical paper
SWPSC 2026001 abstract; JPT 1 July 2026, Haynesville Field Results

Edge Computing and IIoT Enable Autonomous Artificial Lift and Well Optimization
2026-07-01 · Supplier-authored technical account
Field Results: Haynesville Basin

See the recorded source reviews

The percentage is not an annual result. It concerns liquid-loading wells, not all gas wells.

EXHIBIT 03 / HOW MUCH DO WE KNOW?

Operating use is easier to see than net return.

In our 220 company-use profiles, 151 describe operating or scaled use. 55 of those also report a numerical result. The figures describe this selected collection, not the industry’s adoption rate.

One mark = one profile. The four groups are mutually exclusive and sum to 220. Profiles can overlap within a wider program.

All 220 profiles shown.

Inspect the profiles behind these counts

No counted result does not mean failure. A numerical result does not establish net return. The register has 104 named companies or groups with company-use profiles; some programs have several profiles.

THE NEXT CONVERSATION

Use the evidence to choose your next question.

Choose the work you want to change.

Start with a decision someone owns today. Compare the proposed system with the current process. Measure operating results and the work needed to keep the system useful.

Name the current workflow and its owner.
Agree the assets, baseline and measurement period.
Set approval, override and fallback rules.
Decide in advance what would justify stopping or expanding.

Find the work the installed system leaves behind.

Investigate unresolved reading, coordination and follow-up around existing tools. A public case does not establish an unmet need, a budget or willingness to buy. Treat the proposed openings as questions for discovery.

Establish what the incumbent system already does.
Find a recurring exception or review burden.
Test the model’s incremental contribution.
Include integration and human review in the economics.

Compare twelve projects and open an evaluation brief

ORIGINAL SOURCES

Follow the claim back to the disclosure.

Open a company for publication dates, source locations and recorded review status. Some publishers restrict access. An access limitation is different from evidence that a claim is wrong.

Devon Energy · AI-006

Devon Energy — deployment evidence
2026-08-05 · Operator
Location: Devon’s operator-hosted Q2 2026 transcript (2026-08-05) says its closed-loop AI system autonomously optimized 1,000 wells in real time, 24/7; management said 200 wells in March and a path to broad deployment across >2,000 additional Permian wells, including first legacy Coterra wells. This supersedes the Q1 count of >850.
Recorded access: 2026-10-04. Relevant source passages reviewed

Devon Q1 2026 earnings call — Smart Gas Lift
2026-05-06 · Issuer-hosted earnings transcript
Location: Prepared remarks p. 4; Q&A pp. 12–14; 21 October additional locator: PDF and printed p. 4; Clay Gaspar prepared remarks, three waves of AI
Recorded access: 2026-10-21. Issuer-hosted corrected 20-page transcript retrieved HTTP200. Editor read relevant p.4 and visually inspected it; management claims, not independent performance validation. Earlier claim reviews retain their original scope.

Occidental · AI-215

Occidental Q2 2025 Earnings Call Transcript
2025-08-07 · Issuer-hosted earnings-call transcript
Location: Page 3, Vicki Hollub prepared remarks
Recorded access: 2026-09-29. Full 17-page company-hosted transcript reviewed.

ADNOC · AI-226

ADNOC and SLB Deploy AI Platform Across Over 120 Drilling Rigs to Strengthen Upstream Performance
2026-08-04 · Operator press release
Location: Release body: paragraphs on RTOC deployment, workflow, operating scope and reported metrics
Recorded access: 2026-09-30. Full primary release opened and read; release body reviewed in full.

QB Energy / Quantum · AI-054

Geospatial field planning for an unnamed Quantum portfolio operator
2026-04-30 · Microsoft customer story
Location: Field development planning solution; Improved speed, scale, and precision; updated algorithm paragraph
Recorded access: 2026-10-21. Full original Microsoft customer story read again on 21 October. Updated-version comparison is vendor-reported; no independent validation.

Quantum 2026 Stakeholder Report
2026 · Sponsor stakeholder report
Location: Pages 11–15; page 68 disclaimer
Recorded access: 2026-10-09. Relevant primary sections reviewed

Vickery Energy / Quantum · AI-276

Land-title research at an unnamed Quantum portfolio company
2026-04-30 · Microsoft customer story
Location: From the field to the back office and beyond — land-title research paragraphs
Recorded access: 2026-10-10. Original land-title passage reopened 10 October 2026; prior full-document review 6 October 2026. Results remain vendor-reported.

Quantum 2026 Stakeholder Report
2026 · Sponsor stakeholder report
Location: Pages 11–15; page 68 disclaimer
Recorded access: 2026-10-09. Relevant primary sections reviewed

Aramco · AI-141

Aramco AI and Big Data
Undated; accessed 2026-09-27 · company announcement
Location: Fadhili Gas Plant section
Recorded access: 2026-09-27. full text reviewed

Burk Royalty · AI-285

Field Optimization Reimagined: full paper
2026 · Operator/vendor-coauthored technical paper
Location: PDF pp. 3, 7–9; figure on p. 8
Recorded access: 2026-10-10. Relevant original source text reviewed; performance remains author-reported, not independently validated.

Pump-by-exception monitoring and control at a West Texas lease
2026 · Operator/vendor-coauthored technical paper
Location: SWPSC 2026052 abstract; linked full paper PDF pp. 3, 7–9 (site configuration, eight-month results, economics and projections)
Recorded access: 2026-10-10. Relevant original source text reviewed; performance remains author-reported, not independently validated.

BPX Energy · AI-284

Daily lift recommendations across a 441-well Permian pilot
2026 · Operator/vendor-coauthored technical paper
Location: SWPSC 2026051 abstract: method, 441-well asset and reported results
Recorded access: 2026-10-10. Relevant original source text reviewed; performance remains author-reported, not independently validated.

Expand Energy · AI-286

Autonomous liquid unloading on nine Haynesville gas wells
2026 · Operator/vendor-coauthored technical paper
Location: SWPSC 2026001 abstract; JPT 1 July 2026, Haynesville Field Results
Recorded access: 2026-10-10. Relevant original source text reviewed; performance remains author-reported, not independently validated.

Edge Computing and IIoT Enable Autonomous Artificial Lift and Well Optimization
2026-07-01 · Supplier-authored technical account
Location: Field Results: Haynesville Basin
Recorded access: 2026-10-10. Relevant original source text reviewed; performance remains author-reported, not independently validated.

Scope, dates and how to read this briefing

Sunya’s synthesis of selected public-source profiles. Workflow layers and proposed next steps are editorial interpretations, not a technology inventory or company plans. Counts describe the selected register, not worldwide adoption or a success rate. Source statements do not independently establish performance or net return. This briefing reorganizes existing evidence; it does not re-audit every source.

Initial screen: 26 September 2024–26 September 2026. Broad classification review: 8 October 2026. Selected register updates: 14 October 2026, including CNX safety and Aker BP’s implementation announcement. Individual sources retain their own dates. This briefing’s edition date does not mean every original source was reviewed again.

The evidence counts include the 12 October Flywheel platform and 14 October CNX/Aker BP additions. The six selected workflows remain unchanged. Microsoft’s Quantum story and ADNOC’s rig release were reopened while preparing the 10 October brief. Other cases use the register’s recorded source reviews.