AI Applied ConsortiumAI APPLIED
CONSORTIUMEnergy Conference Network
UPSTREAM AI AT WORK
UPSTREAM / AI & YOUR WORK

Scroll at your own pace. Use ← and → to move through the story.
An imagined Alberta prairie energy landscape in warm evening light

AI / UPSTREAM OIL & GAS

Staying above
the automation
line.

As technology takes on more of the work,
where will you make the difference?

Konrad KonarskiChairperson, AI Applied ConsortiumAI Applied Consortium × Energy Conference Network
Explore the journeySCROLL TO BEGIN

01 / DISRUPTIVE INNOVATION

The innovator’s dilemma.

Scroll to explore
TODAY
THE DECISION WINDOW
THE NEXT WAVE
Easy to
postpone.
The core looks healthy while the next approach improvesConceptual curves adapted from the supplied reference. The established business grows, peaks and then loses relative value. A simpler new approach starts lower, improves and overtakes it. The decision window comes before the crossover. The shape is illustrative, not a forecast. PERFORMANCE & CUSTOMER VALUETIME →
The core
The next wave
Still performing
Expectations shift.
LEARN EARLY
Your team
Core businessNew approach
THE CORE

The core still delivers.

Reliable work. Familiar customers. A proven way to operate.

TRY THE CHOICE
Conceptual illustration
The idea behind the animation

The dilemma is a resource-allocation problem: serving existing customers can make a smaller, less-proven opportunity difficult to prioritise. Disruption begins with simpler, more accessible offerings that can improve into the mainstream.

The declining core curve follows the supplied illustration. It is not a universal business lifecycle or a forecast for oil and gas. AI-enabled upstream workflows are a teaching application of the idea; not every AI improvement is disruptive innovation.

Christensen Institute ↗

01A / ENERGY THOUGHT EXPERIMENT

Unlimited energy
changes what companies sell.

Imagine energy available to everyone, at almost no cost.

Scroll through the change
Energy producerRevenue depends on selling fuel
CURRENT MODELSupply controlled centrally.

Customers buy fuel through the chain.

ENERGY AVAILABLE TO EVERY CUSTOMERCustomers no longer need to buy fuel.

Energy is available at the point of use.

THE OLD BUSINESS MODEL BREAKSSelling scarce fuel no longer creates the same value.
WHAT THE COMPANY NOW SELLSInfrastructure, integration and reliable outcomes.
01Extract fuel×FUEL STEP NO LONGER NEEDED
02Transport fuel×FUEL STEP NO LONGER NEEDED
03Sell oil & gas×FUEL STEP NO LONGER NEEDED
ENERGY CUSTOMER

Homes

∞Energy here

Direct energy access

ENERGY CUSTOMER

Industry

∞Energy here

Direct energy access

ENERGY CUSTOMER

Transport

∞Energy here

Direct energy access

∞ENERGY
∞ENERGY
∞ENERGY
01Build infrastructureEnable use of abundant energy
02Integrate systemsConnect equipment and operations
03Deliver reliabilitySell dependable outcomes
1 / CURRENT MODEL

The business earns revenue by selling oil and gas.

Production, transport and sales are organized around that revenue.

Unlimited energy requires a new business model.Energy thought experiment

01B / THE SAME CHANGE WITH INTELLIGENCE

Scalable intelligence
moves decisions to teams.

Give frontline teams AI, company knowledge and tools.

Scroll through the change
Leadership + central expertsKnowledge and decision authority at the top
CURRENT MODELKnowledge concentrated centrally.

Routine decisions require upward approval.

KNOWLEDGE MOVES TO THE EDGEEach team gains AI, approved data and tools.

Read → compare → draft → check at machine speed.

ROUTINE DECISION GATES COLLAPSEDecisions stop travelling up for routine approval.
HOW THE ORGANIZATION NOW WORKSLeaders set limits. Teams decide. Exceptions escalate.
03Senior approval×ROUTINE GATE REMOVED
02Manager approval×ROUTINE GATE REMOVED
01Information review×ROUTINE GATE REMOVED
FRONTLINE TEAM

Production

AIData + tools

Local decisions

FRONTLINE TEAM

Operations

AIData + tools

Local decisions

FRONTLINE TEAM

Drilling

AIData + tools

Local decisions

FRONTLINE TEAM

Maintenance

AIData + tools

Local decisions

AIKNOWLEDGE
AIKNOWLEDGE
AIKNOWLEDGE
AIKNOWLEDGE
EXCEPTIONS ONLYException reviewHigh-risk or out-of-policy
1 / CURRENT MODEL

Routine decisions move through several approval layers.

Frontline teams send information upward and wait for a decision.

Distributed intelligence requires a new operating model.Illustrative organization

02 / AI CAPABILITY & ENERGY VALUE

AI capability and energy-sector value.

Scroll to build capabilitySelect a stage

AI MODEL CAPABILITYCONCEPTUAL PROGRESSION
Model capability progression from the supplied presentationThe supplied illustrative capability index rises from GPT-4o at 200 through OpenAI o1, o3, GPT-5, GPT-5.2, Opus 4.6 and Fable 5.1 to a GPT-6 Astra scenario at 1600. It is not a validated benchmark.04008001,2001,600CAPABILITY INDEXGPT-4oOpenAI o1OpenAI o3GPT-5GPT-5.2AnthropicOpus 4.6Fable 5.1GPT-6 AstraScenario202420252026
Source: supplied presentation.
ESTIMATED ANNUAL VALUEUPSTREAM · US$ / YEAR
Three expanding operational-value horizonsThree bubbles are sized by area for annual value potential of about 65, 125 and 230 billion US dollars. They fill as the capability and organizational layers develop in this illustrative sequence.$65BProven AIProven applications$125BWider adoptionMore of the operation$230BAutonomous operationsFull operating potential
McKinsey estimates ↗
01

Model capability

Understand information. Work through a task.

02

Agent harness

Planning, tools, memory and feedback.

03

Organizational context

Data, domain knowledge, integration and governance.

MODEL CAPABILITY

Stronger models handle more complex work.

Useful assistance starts with stronger understanding and reasoning. People still supply context, check the work and decide what happens next.

Figures & sources

The model chart retains the supplied deck’s index, model labels and dates. These values are unverified illustrations, not measured performance or confirmed reproduction dates.

The approximately $65B, $125B and $230B estimates span multiple AI techniques across upstream energy. They represent alternative cumulative scenarios, not additive amounts or returns attributable to specific models or agents.

Separate market-revenue estimate: AI in oil and gas grows from $4.28B in 2026 to a forecast $7.91B in 2031. Mordor Intelligence ↗

03 / SKILLS IN MOTION · 2025–2030

Skills for working with AI.

Scroll to explore
AI workflow autonomyScope before human reviewIllustrative pathway · levels 0–5
A continuous workflow autonomy pathway from 2025 through 2030Six illustrative levels show increasing delegation: answer support, single tasks, linked steps, exception handling, workflow coordination and bounded autonomy. This is a presentation framework, not a measured benchmark or a forecast of capability by year.012345
Answer support
Single tasks
Linked steps
Exception
handling
Workflow
coordination
Bounded
autonomy

2025 · Level 0 · Answer support

AI WORK
ADOPTION SCENARIO
Draft
Compare
Chain steps
Check exceptions
Coordinate
Bounded workflow
Your skillsSpecialist knowledge → broader application
2025

Start with a skill you already use.

Evidence & assumptions

What the animation represents

The scale. Workflow autonomy describes the scope delegated to AI before a human review point: 0, answer support; 1, single tasks; 2, linked steps; 3, exception handling; 4, workflow coordination; 5, bounded autonomy. This is an ordinal presentation framework, not a benchmark score.

The timeline. The whole 2025–2030 line is an illustrative pathway. The year positions are not measured model performance, reproduction predictions or promised adoption dates. Higher autonomy still operates within defined permissions, checks and human escalation.

Skill paths. Illustrative development paths for upstream engineers, supervisors and operations teams. WEF expects analytical thinking, creative thinking, and resilience to remain important and grow toward 2030. These are not a ranked top three for 2030.

WEF’s top three core skills in 2025: analytical thinking; resilience, flexibility and agility; leadership and social influence. Employers expect 39% of core skills to change by 2030.

Dates synchronize the three views for discussion. They do not establish when upstream tasks become automatable. Domain expertise remains valuable.

Agents, workflows and human review ↗WEF skills outlook ↗Upstream operating-value context ↗

Specialist knowledge remains the foundation.

04 / STAYING ABOVE THE AUTOMATION LINE

Human judgment as AI takes on tasks.

Scroll to explore
YOUR CONTRIBUTION

Frame the
right problem.

Which production loss should we investigate first?

GoalConstraintsPeople THE RIGHT QUESTION TimingConstraints WEIGH THE TRADE-OFF The record The history ?The context LISTEN · UNDERSTAND · COMMIT
AUTOMATION LINE
AI PREPARESMore preparation. More room for judgment.
Trends · Reports · History
Find & organise
Ready for your review
Scroll down to advance. Scroll up to revisit.Illustrative workflowAAIC · The Spark ↗

05 / HOW AN AI AGENT WORKS

From a goal to completed work.

Scroll through the workflow

How it works

The harness runs the model–tool loop, maintains working context and records progress. The model chooses the next step; tool results inform its next decision.

MCP (Model Context Protocol) is a standard connection to tools and context. It does not supply reasoning or replace access controls. Tools may also connect through other APIs.

Guardrails are implemented checks and permissions around the workflow. This example pauses updates that require approval; missing evidence triggers another search or human review. Checks reduce risk but do not guarantee correctness.

Conceptual workflow. MCP architecture ↗ · Agent design ↗ · Harness design ↗

06 / USE CASE 01 · PRODUCTION SURVEILLANCE

Find the cause of lost production.

Scroll through the use case
HUMAN CONTRIBUTIONAUTOMATION LINEAI WORKFLOW ↓

Track: deferred production · diagnosis time

Illustrative upstream workflow, not a client deployment or measured result. AI supports preparation and analysis; authorized personnel retain operational decisions and required approvals.

The four human contributions remain relevant throughout. Their sequence highlights a different contribution at each stage.

06 / USE CASE 02 · DRILLING OPERATIONS

Carry the right context into the next shift.

Scroll through the use case
HUMAN CONTRIBUTIONAUTOMATION LINEAI WORKFLOW ↓

Track: handover preparation · missed actions · NPT

Illustrative upstream workflow, not a client deployment or measured result. AI supports preparation and analysis; authorized personnel retain operational decisions and required approvals.

The four human contributions remain relevant throughout. Their sequence highlights a different contribution at each stage.

06 / USE CASE 03 · ASSET RELIABILITY

Turn equipment signals into planned work.

Scroll through the use case
HUMAN CONTRIBUTIONAUTOMATION LINEAI WORKFLOW ↓

Track: unplanned downtime · planning time · rework

Illustrative upstream workflow, not a client deployment or measured result. AI supports preparation and analysis; authorized personnel retain operational decisions and required approvals.

The four human contributions remain relevant throughout. Their sequence highlights a different contribution at each stage.

07 / WORKSHOP DEEP DIVE · PRACTICAL EXERCISES

Five practical moves. Built around your work.

Scroll through the exercises ↓
START WITHShift notes
APPROVED AI TOOLSCapture
BUILD A REUSABLE RESULTHandover brief

Events · Open actions · Owners

Confirm the action owners
Purpose + context→Sources + checks→Human review

08 / WORKSHOP DEEP DIVE · FROM EXERCISE TO EVERYDAY WORK

A working example. A plan to put it into use.

Scroll to build your next step ↓
01A repeatable workflowInputs, output and review point
02A team working agreementData, permissions and escalation
03A measured trialOwner, baseline and success measures
04A shared learning patternFeedback and the next skill to build
YOUR NEXT 30 DAYS
BaselineTestReviewImprove

REGISTER · ABOVE THE AUTOMATION LINE WORKSHOP

Bring one workflow.
Leave with
a better way to work.

Register your interest ↗

Request workshop dates and details through the AAIC interest form.

Konrad KonarskiAI Applied Consortium × Energy Conference Networkinfo@aaiconsortium.org
QR code for the AAIC workshop interest formRegister your interestScan for workshop details