Knowledge Elicitation as a Service

Your Best Engineers Are Retiring.
Their Knowledge Doesn't Have To.

AI-native knowledge elicitation that captures 46 years of irreplaceable O&G expertise โ€” before it walks out the door.

231,000
Years of Expertise at Risk
15-Year
Experience Gap
Zero
Commercial Solutions โ€” Until Now
Schedule a Pilot Assessment โ†’ See the Methodology

The Retirement Wave Is Not a Future Problem

50% of the O&G workforce is retirement-eligible within a decade. The knowledge leaving with them took generations to build โ€” and cannot be Googled.

0
Estimated expert-years retiring from US O&G this decade
Based on 50% retirement-eligible ร— avg. 23yr tenure ร— ~20,000 senior engineers
โš ๏ธ
$2โ€“10M
Cost of a Single Well Failure
Drilling decisions made without the pattern recognition of a 30-year reservoir engineer. One wrong call can eclipse the cost of capturing that knowledge entirely.
๐Ÿญ
$50M+
Single Refinery Incident Cost
Catalyst management, crude slate optimization, turnaround execution โ€” all tacit judgment. When the expert retires, the knowledge gap becomes a liability event waiting to happen.
โณ
15 Years
The Experience Gap
The gap between your retiring experts and the next generation of engineers who can replace their judgment. You cannot accelerate it with training programs. You can bridge it with captured knowledge.
"Your competitors are losing the same expertise.
The question is who captures it first."

Traditional KM Captures What Experts Write.
We Capture How They Think.

Documents go stale. Chatbots hallucinate on edge cases. Knowledge Elicitation Architecture captures the decision-making judgment that makes your senior engineers irreplaceable.

Traditional KM
Document Management
Captures what experts write, not what they know
  • โœ—Procedures and reports โ€” not reasoning
  • โœ—Goes stale within 6 months of being written
  • โœ—Can't surface during real-time decisions
  • โœ—No edge-case coverage โ€” only documented scenarios
  • ~Compliance-useful, operationally limited
AI Chatbots
Surface-Level Q&A
Built on generic models, not your experts' brains
  • โœ—Surface-level answers from public data
  • โœ—Fails catastrophically on edge cases
  • โœ—No institutional memory โ€” just internet knowledge
  • โœ—Confidently wrong on domain-specific judgment
  • ~Useful for search, dangerous for critical decisions
KEA โ€” Our Approach
Knowledge Elicitation Architecture
Captures decision-making judgment, pattern recognition, tacit expertise
  • โœ“Expert interview trees designed for O&G domains
  • โœ“AI-assisted extraction โ€” structured knowledge graphs
  • โœ“Handles edge cases explicitly โ€” that's the point
  • โœ“AI-consumable, maintainable, queryable
  • โœ“Deployable as autonomous AI agents or decision tools
๐ŸŽ™๏ธ
Expert
Interview
โ†’
๐Ÿค–
LLM-Assisted
Extraction
โ†’
๐Ÿง 
Knowledge
Graph
โ†’
โšก
Deployable
AI Agent

The SEE Framework โ€” Four Phases to Decision-Ready Knowledge

Structured Expert Elicitation doesn't happen in a single interview. Each phase is purpose-built to extract, calibrate, and encode the probabilistic judgment that makes senior O&G engineers irreplaceable. Hover each phase for detail.

Phase 01
๐ŸŽฏ
Asset Identification
Define exactly what knowledge to elicit โ€” not everything, but the judgment that matters most. Pore pressure prediction, reservoir thickness uncertainty, drilling risk windows. Scope drives quality.
What Happens
Knowledge risk mapping ยท decision node identification ยท expert asset inventory
Who's Involved
Senior engineers + IIA knowledge architects

Output: Prioritized knowledge asset map with retirement risk scores by domain and individual expert.

๐Ÿ“„ Knowledge Asset Map
Phase 02
๐Ÿง 
Bias Training Workshop
Cognitive bias is the enemy of accurate elicitation. Before any interview, experts are trained on the three biggest sources of error in expert judgment โ€” anchoring, overconfidence, and availability bias.
Biases Addressed
Anchoring ยท overconfidence ยท availability heuristic ยท base rate neglect
Who's Involved
Expert cohort + IIA calibration facilitator

Output: Calibrated expert cohort with documented bias profiles and individual uncertainty baselines.

๐Ÿ“Š Expert Calibration Report

90% Confidence Interval Test

For each question, give a range you're 90% sure contains the correct answer.
If you're well-calibrated, ~4โ€“5 of 5 correct answers should fall inside your range.

Q 1 / 5
What is the average daily rate for a deepwater drillship operating in the Gulf of Mexico?
Enter your 90% confidence interval โ€” a range you're 90% sure contains the correct value.
USD / day
How many wells were drilled in the Permian Basin in 2024?
Enter your 90% confidence interval.
wells drilled
What is the typical pore pressure gradient at 15,000 ft TVD in a GoM sub-salt formation?
Enter your 90% confidence interval.
psi / ft
What percentage of well control incidents are attributed to inaccurate pore pressure prediction?
Enter your 90% confidence interval.
percent (%)
How many years of experience does the average retiring petroleum engineer have?
Enter your 90% confidence interval.
years

Your Calibration Results

How many correct answers fell inside your 90% confidence intervals?

โ€“
out of 5 answers captured by your ranges
Expected vs. Actual Calibration
Expected
(90% CI = 90%)
90%
Your Score
(actual)
โ€“
โ–ถ See correct answers
Experience the full elicitation process โ†’ โ†บ Retake the test
Phase 03
๐Ÿ’ฌ
Elicitation Interview
Structured questioning using the Stanford/SRI Protocol โ€” the gold standard for extracting probabilistic estimates. Real-time AI-assisted capture surfaces patterns and decision trees as the expert speaks.
What Happens
Stanford/SRI Protocol ยท structured questioning paths ยท real-time AI capture ยท probability extraction
Who's Involved
Expert + IIA domain interviewer + AI capture system

Output: Structured interview transcript + real-time knowledge graph draft flagged with coverage gaps.

๐Ÿ”— Structured Knowledge Graph
Phase 04
๐Ÿ“ˆ
Documentation & Encoding
Verbal estimates become quantitative distributions. P10/P50/P90 outputs. Decision-ready artifacts structured for AI consumption โ€” not PDFs, but machine-queryable knowledge that drives autonomous systems.
What Happens
Verbal โ†’ quantitative transformation ยท probability distribution fitting ยท P10/P50/P90 output
Who's Involved
IIA encoding team + domain peer reviewer

Output: P10/P50/P90 distributions + deployable AI agent knowledge base + digital twin calibration inputs.

โšก P10/P50/P90 Outputs
๐Ÿ”ฌ
The SEE framework is built on the Stanford/SRI Protocol โ€” the most rigorous methodology for eliciting expert probabilities, originally developed for nuclear safety and now applied to O&G uncertainty quantification. Combined with NTNU's Case-Based Reasoning research and 46 years of domain expertise from upstream exploration to downstream refining.

Built on 20+ Years of Peer-Reviewed Research

Our methodology isn't proprietary innovation in a vacuum. It's grounded in peer-reviewed decision science from the top institutions in energy engineering โ€” the work was already done. We operationalized it.

๐ŸŽ“
University of Stavanger (UiS)
Department of Energy & Petroleum ยท Decision Science
Research led by Bratvold & Morosov โ€” the leading academic work on probability elicitation specifically for hydrocarbon exploration. The project "Probability Elicitation in Hydrocarbon Exploration" directly informs IIA's Phase 2 bias calibration methodology and Phase 3 structured questioning design. Their debiasing framework is embedded in every SEE engagement.
Decision Science Probability Elicitation Hydrocarbon Exploration Debiasing
Morosov & Bratvold (2021). Probability elicitation in hydrocarbon exploration. SPE Reservoir Evaluation & Engineering.
Welsh, Begg & Bratvold. Debiasing oil and gas decisions. Journal of Petroleum Technology.
โš™๏ธ
NTNU โ€” Norwegian Univ. of Science & Technology
SFI DigiWells Centre ยท Autonomous Drilling Integration
The SFI DigiWells centre at NTNU develops the autonomous drilling infrastructure that IIA's Phase 4 outputs feed directly into. Case-Based Reasoning research by Aamodt & Skalle at NTNU forms the foundational architecture of our knowledge graph construction โ€” how cases are indexed, retrieved, and adapted in real-time operational contexts.
Case-Based Reasoning SFI DigiWells Autonomous Drilling Knowledge Graphs
Aamodt & Skalle. Case-Based Reasoning in petroleum drilling operations. NTNU / SFI DigiWells.
SFI DigiWells Centre for Research-based Innovation, NTNU Trondheim.
"Our methodology is built on 20+ years of peer-reviewed research.
The science was done at UiS and NTNU. We put it to work in the field."

Captured Knowledge Drives Autonomous Operations

Phase 4 outputs don't sit in a PDF. They feed directly into digital twin calibration and real-time autonomous drilling systems โ€” the same infrastructure NTNU's SFI DigiWells demonstrated across 500 meters of autonomous drilling.

Step 01
๐Ÿ’ฌ
Expert Knowledge Capture
SEE Framework extracts P10/P50/P90 distributions and decision judgment from retiring engineers
โ€บ
Step 02
๐Ÿค–
AI Encoding
Knowledge graphs and probability distributions structured for machine consumption โ€” not documents
โ€บ
Step 03
๐Ÿ”„
Digital Twin Calibration
Expert priors calibrate simulation models and uncertainty parameters โ€” replacing manual tuning
โ€บ
Step 04
โšก
Real-Time Drilling Decisions
Autonomous systems make real-time calls encoded with 30+ years of expert judgment
๐Ÿ›ž
NTNU 500m Autonomous Drilling Demonstration
SFI DigiWells demonstrated fully autonomous drilling across 500 meters โ€” the first of its kind at scale. The knowledge models powering those decisions required exactly the kind of expert encoding that IIA's Phase 4 produces: probabilistic, structured, machine-queryable.
๐Ÿ”—
From Expert Brain โ†’ Autonomous System
IIA's SEE output is designed to be DigiWells-compatible โ€” the same probability distribution format, the same Case-Based Reasoning indexing structure. Knowledge doesn't retire with the engineer. It becomes the operating logic of autonomous systems that run for decades.

Who This Serves

The knowledge retention problem is universal in O&G. Some markets have a narrower window than others. These five represent the highest-urgency opportunities today.

๐Ÿ‡ป๐Ÿ‡ช
Venezuela Return Operations
Operators returning to Venezuela need PDVSA-era expertise from the diaspora โ€” most of whom are now in Houston, aging out, and not coming back. The window to capture decades of Orinoco, Maracaibo, and eastern Venezuela operational knowledge is closing fast.
โšก Narrow Window โ€” 3-5 Years
๐Ÿ‡ฒ๐Ÿ‡ฝ
PEMEX & Mexico
15,000โ€“25,000 technical staff approaching retirement. Cantarell's decline was partly a knowledge failure โ€” engineers retired and judgment about that reservoir walked with them. The same dynamic is playing out across the entire PEMEX technical organization.
โšก 15Kโ€“25K Technical Staff at Risk
๐Ÿ›ข๏ธ
Permian Basin Independents
Drilling judgment from 35-year veterans who know the Wolfcamp, Bone Spring, and Delaware Basin from the rock up cannot be Googled or trained. That hyperlocal expertise โ€” when to push production, when to back off, how to read the pressure curves โ€” is uniquely valuable and uniquely vulnerable.
โšก Hyperlocal Expertise Irreplaceable
โš—๏ธ
Downstream Operators
Refinery turnaround expertise, catalyst management, crude slate optimization โ€” all tacit judgment built over decades. A retiring process engineer takes with them 20 years of hard-won knowledge about what happens when you push that unit to 108% capacity in August heat.
โšก $50M+ Incident Exposure
๐Ÿ‡จ๐Ÿ‡ฆ
Canadian Oil Sands
Cold-weather operations knowledge built over 40+ years of SAGD, CSS, and mining operations. The engineering solutions developed for the oil sands โ€” freeze-thaw cycles, bitumen handling, tailings management โ€” represent institutional knowledge that is genuinely irreplaceable without deliberate capture.
โšก 40+ Years of Unique Operations

What's the Knowledge Gap Costing You?

Estimate your organization's exposure and the ROI of capturing that expertise before it retires.

Your Organization

Retiring Experts (next 5 years) 20
Average Years of Experience per Expert 30 yrs
Estimated Cost per Knowledge-Gap Incident $2.0M

Your Exposure

Knowledge-Years at Risk
600
Estimated Annual Incident Exposure
$8.0M
Typical Capture Investment
$500Kโ€“$1.0M
ROI Multiple (first prevented incident)
4โ€“16ร—
Average engagement pays for itself within the first prevented incident.

Engagement Structures

From a focused pilot to a full value-chain engagement. Every tier includes ongoing maintenance to keep captured knowledge current as conditions change.

Entry Point
Pilot
$60Kโ€“$120K
8โ€“12 weeks ยท 3โ€“5 experts ยท Single domain
  • 3โ€“5 senior expert interviews
  • Single domain coverage (e.g., drilling, completions, or reservoir)
  • Knowledge graph construction + validation
  • Proof-of-concept AI agent deployment
  • ROI assessment and expansion roadmap
  • 1 year maintenance included
Start Pilot Assessment โ†’
Full Coverage
Enterprise
$500Kโ€“$1.5M
12โ€“24 months ยท Full value chain
  • Full value-chain coverage (upstream โ†’ downstream)
  • Entire at-risk expert cohort
  • Custom AI agent suite deployment
  • Training program integration
  • Continuous elicitation as expertise evolves
  • Executive knowledge risk dashboards
  • 5 years maintenance + continuous capture protocol
Request Enterprise Brief โ†’

Every engagement includes an ongoing maintenance subscription to keep captured knowledge current as conditions change, experts expand their understanding, and new edge cases emerge.

3 Generations. 46 Years. Upstream to Downstream.

The only firm combining deep O&G domain expertise with AI-native knowledge capture architecture. Not bolted-on AI. Not generic consultants. Built from the ground up for this.

๐Ÿ—๏ธ
46 Years of Domain Depth
Three generations of O&G expertise across upstream exploration, production operations, downstream refining, and industrial inspection. We know what questions to ask because we've lived the answers.
๐Ÿค–
AI-Native Architecture
Built on Claude's autonomous agent architecture โ€” not bolted-on AI. Our capture pipeline was designed for AI from day one: structured for machine consumption, queryable, and deployable as working agents.
๐ŸŽ“
Academic Foundation
Our methodology is grounded in 20+ years of NTNU Case-Based Reasoning research โ€” one of the world's leading programs for AI in engineering. Not a vendor with a chatbot. A methodology with a scientific foundation.
๐ŸŒ
Venezuelan Diaspora Network
400+ former PDVSA professionals in our Houston network. For operators returning to Venezuela or capturing PDVSA-era expertise before it's gone โ€” this network is irreplaceable and inaccessible to any other firm.
๐Ÿ”ง
We Speak Your Language
Cantarell decline curves. Wolfcamp proppant loading. SAGD steam-oil ratios. Catalyst deactivation kinetics. We're not learning your domain during the engagement. We walked in knowing it.
โšก
The Only Commercial Solution
There is no other firm offering structured knowledge elicitation as a service for O&G with AI-native deployment. This gap exists because it's hard. We built the capability anyway โ€” and you're looking at it.
Interactive Demo ยท GoM Sub-Salt ยท No signup required

Experience the SEE Protocol Firsthand

Step through a live pore pressure elicitation interview for a Gulf of Mexico Wilcox well. See your own bias in real time. Watch a full probabilistic distribution built from expert judgment โ€” and the $12M decision it changes.

Try the Simulator โ†’ 4 steps ยท ~5 minutes ยท No account needed
โœ“ Bias calibration exercise
โœ“ Live P10/P50/P90 chart
โœ“ Casing point recommendation
โœ“ Cost impact dashboard