Section 01
The Challenge
The operator had a problem that didn't show up on any dashboard — until it was almost too late. One person held the institutional knowledge that kept their Mississippi Canyon operations safe. And they were leaving.
🧠
The Expert
Chief Drilling Engineer
32 years of sub-salt pore pressure expertise across the Gulf of Mexico. Deep knowledge of carbonate formations, pressure window estimation, and wellbore stability in the Mississippi Canyon protraction area.
⏳
The Timeline
8 months to retirement
Standard knowledge transfer — shadowing, documentation, mentoring — typically requires 18–24 months to be effective. The operator had less than half that window.
👷
The Team
3 junior engineers
All with under 5 years of experience. Each had worked offshore, but none had developed the intuition for sub-salt pressure prediction that takes a decade of deepwater exposure to build.
⚠️
The Warning Signs
2 NPT events in 18 months
Both traced back to pore pressure misinterpretation during the planning phase. Each event cost the operator between $800K and $1.4M in non-productive time. The chief engineer was consulted after the fact — not before.
🔴
The real risk wasn't the retirement. It was the 6–18 months after retirement — when the junior engineers would be making pore pressure decisions on their own, without knowing what they didn't know. The NPT events were the early signal. The operator recognized that the next event could be significantly worse.
Section 02
Timeline & Approach
We deployed the SEE Protocol — Structured Expert Elicitation — across four phases. Each phase built on the last. The chief engineer participated in structured sessions, not open-ended interviews. Every session was calibrated to surface tacit knowledge that couldn't be retrieved through documentation alone.
1
Weeks 1–2
Asset Identification
Before a single elicitation session, we mapped the full landscape of what we needed to capture. Domain mapping with the engineering team identified 14 critical knowledge areas. Pore pressure prediction was ranked highest on both criticality and risk of loss.
- Mapped 14 critical knowledge domains across drilling, completions, and geomechanics
- Prioritized sub-salt pore pressure prediction as the highest-risk knowledge gap
- Identified 6 analog well datasets critical to the expert's mental model
- Established ground truth: 73% of documented procedures lacked the "why" that made them work
2
Weeks 3–4
Bias Training Workshop
Expert knowledge elicitation fails when the expert isn't calibrated. Overconfidence, availability bias, and anchoring systematically distort even genuine expertise. We ran a two-day calibration workshop before any technical elicitation.
- Calibration exercise revealed 73% overconfidence across the junior engineering team
- Chief engineer showed well-calibrated uncertainty ranges — a key signal of true expertise
- Identified 4 specific cognitive biases active in the team's current decision-making
- Established probabilistic framing: P10/P50/P90 vocabulary used in all subsequent sessions
3
Weeks 5–8
Elicitation Sessions
Six structured elicitation sessions, each 3–4 hours. Sessions were built around specific well scenarios drawn from the expert's actual project history — not hypotheticals. Every decision point was probed for preconditions, thresholds, and the analog experience behind it.
- 847 discrete decision points captured and categorized across pore pressure prediction workflow
- Recorded 94 distinct "rules of thumb" — each traced to a specific well or formation experience
- Captured 12 failure modes the expert could predict but had never written down
- Documented uncertainty bounds for 6 different geological sub-environments in the Mississippi Canyon area
4
Weeks 9–12
Documentation & Encoding
Raw elicitation output was structured into a knowledge graph, validated by the expert, and encoded into a deployable AI agent. Peer validation by two external pore pressure specialists confirmed fidelity. Junior engineers participated in the final validation — their confusion points refined the agent's explanation logic.
- Knowledge graph built: 847 decision nodes, 1,200+ edges encoding conditional relationships
- AI advisory agent trained, deployed on Slack — accessible during well planning meetings
- Training curriculum built for junior engineers: 8 modules, structured around captured decision framework
- Quarterly refresh protocol established for incorporating new well data post-retirement
"
I've been drilling these wells for 32 years. I never thought I could explain what I know — it just lives in my head. These sessions forced me to slow down and articulate decisions I make automatically. The agent they built doesn't just remember what I said. It reasons the way I reason.
Chief Drilling Engineer — Mississippi Canyon Operator (name withheld)
Section 03
Results
Measured at 12 months post-engagement. Baseline figures drawn from the two-year period before the engagement. All NPT figures independently verified against drilling reports.
NPT Reduction
0%
Non-Productive Time
12 days/yr
→
4 days/yr
Pore Pressure Accuracy
0%
Prediction Accuracy
71%
→
94%
Engineer Ramp-Up
0 mo
To full pore pressure competency
6 months
→
2 months
First-Year ROI
0×
Return on engagement cost
$600K cost
→
$2.8M savings
Section 04
What the Client Received
At the end of the 12-week engagement, the operator had four concrete deliverables — not a report, not a workshop summary. Working assets that the engineering team used starting on day one after retirement.
🕸️
Pore Pressure Knowledge Graph
847 structured decision nodes with conditional logic, uncertainty bounds, and formation-specific context. Covers the full sub-salt pore pressure prediction workflow for Mississippi Canyon geology.
847 decision points
🤖
AI Pore Pressure Advisory Agent
Deployed on Slack. Junior engineers tag it during well planning meetings. The agent queries the knowledge graph, surfaces relevant analogs, and returns calibrated P10/P50/P90 estimates with supporting rationale.
Live on Slack
📚
Junior Engineer Training Curriculum
8-module training program built directly from the captured knowledge graph. Each module focuses on a specific decision category. Includes worked examples from real well histories (anonymized).
8 modules
🔄
Quarterly Refresh Protocol
A structured process for incorporating new well data into the knowledge graph each quarter. Prevents knowledge decay over time. Includes a bias-calibration check for the engineering team every 6 months.
Evergreen system
Start Here
Ready to Capture Your Team's Expertise?
Every organization has a knowledge cliff. The question is whether you see it before the retirement letter lands. Our assessment takes 5 minutes and identifies your highest-risk knowledge domains.