# Adoption Map: Now / Next / Later / Never A deployment guidance framework that sorts energy positions into four maturity buckets based on the scoring axes: compressibility, decision criticality, reasoning demand, and company control. All 404 positions carry compressibility scores; 304 of 404 carry the full augmented axes (criticality, reasoning demand, control). The 100 positions added after the augmented scoring pass are not included in this framework. Examples below are drawn from the scored 304. --- ## Now: Deploy Immediately **Criteria:** High compressibility + high or semi-controlled company control + low regulatory friction These workflows are ready today. The company controls the prep stack even if the final deliverable goes to an external counterparty (e.g., a lender or buyer). The AI handles prep and assembly; humans retain signoff authority. **Risk profile:** Low. Upside is measurable (cycle time, error rate). Downside is contained (model outputs are reviewed before they leave the company). **Deployment pattern:** Pilot team (3-5 people), bounded scope, weekly validation against the old path, measurement dashboard, then roll out to parallel operations before retiring the manual path. **Implementation examples:** 1. **Production accounting close support** — Variance reconciliation, accrual mapping, close workbook population. Fully company-controlled. No external approvals. High template density. *Compressibility: 8.7, Criticality: 3.52, Control: Company.* 2. **JIB reconciliation** — Compare operator and non-op JIB statements, flag discrepancies, draft response packets. Non-op workflows are among the most standardized in hydrocarbons. *Compressibility: 8.8, Control: Company.* 3. **Borrowing base packet assembly** — Source document gathering, covenant workbook update, exhibit population, Q&A response drafting (with human review). Treasury owns the final signoff. *Compressibility: 8.3, Reasoning demand: 7.93, Control: Semi-controlled.* 4. **CIM (Confidential Information Memorandum) production** — Extract deck structure, populate exhibits from data room, cross-check assumptions against source docs, flag inconsistencies. *Compressibility: 8.6, Control: Semi-controlled.* 5. **Trade capture, confirmation, settlement** — Booking compliance checks, settlement exception routing, counterparty rec support. Trading ops is high-recurrence, document-heavy, semi-controlled (settlement involves counterparty reconciliation and external clearinghouse rules, though the booking and exception-routing workflow is internal). *Compressibility: 9.6, Reasoning demand: 9.3, Control: Semi-controlled.* 6. **Division order processing** — Owner name standardization, AFE election tracking, curative issue triage, exception queue management. Highly standardized, high document density. *Compressibility: 8.8, Control: Company.* 7. **Well-completion package support** — RFI response drafting, submittal routing, change-order evidence packs, schedule variance tracking. *Compressibility: 8.9, Control: Company.* 8. **Vendor invoice and contract compliance** — Clause extraction from procurement docs, invoice reconciliation against contract terms, exception routing. *Compressibility: 8.7, Control: Company.* --- ## Next: 6-18 Month Horizon **Criteria:** High compressibility but moderate external dependency. The company controls the analytical and document work (CSV: Company-controlled), though the ultimate regulatory or counterparty decision is external. The control zone label reflects who controls the prep stack, not who controls the ruling. These workflows can be compressed aggressively on the internal side, but the external approval clock won't accelerate. The company gets speed and documentation quality; regulators and lenders move at their own pace. Value is partly internal (cycle time, rigor) and partly external (credibility, exception resolution). **Risk profile:** Moderate. The company can move faster, but must signal that to external parties early (don't surprise them with a 48-hour package when they expect two weeks). The reputational risk is low if the external party sees the package as higher quality, not just faster. **Deployment pattern:** Compress the internal prep stack while documenting the external approval gates. Build handoff protocols. Set expectations with counterparties about what's changing. Measure internal cycle time and external resolution time separately—they will diverge. **Implementation examples:** 1. **Rate case evidence assembly** — IRP modeling support, cost-of-service exhibit assembly, discovery response drafting, testimony package support. Company controls the analytics; PUC controls the timeline. *Compressibility: 8.5, Decision criticality: 6.87, Reasoning demand: 8.27, Control: Company.* Expected timeline gain: prep cycle time drops from 8 weeks to 4 weeks; PUC approval timeline remains 6-12 months. 2. **Interconnection analysis** — Facility study evidence packing, constraint analysis, cost-allocation support, interconnection agreement (IA) draft exhibits. Company controls the model; FERC controls the queue. *Compressibility: 8.3, Control: Company.* Internal analysis of FERC interconnection timelines suggests 60-70% of queue time is document/regulatory work (modeled estimate, not observed project data); AI compression buys internal speed, not queue position. 3. **Reserve audit support** — Source data consistency checking, audit workbook assembly, auditor Q&A drafting, reserve-definition compliance verification. Audit firm controls the final audit; company controls the response pack. *Compressibility: 8.4, Criticality: 5.65, Control: Semi-controlled.* 4. **Environmental permitting packages** — EA/EIS exhibit assembly, regulatory requirement cross-check, stakeholder comment response drafting, revision tracking. Company controls the analysis; agency controls the permit. *Compressibility: 8.4, Control: Semi-controlled.* 5. **Debt amendment support** — Covenant waiver analysis, amendment impact modeling, lender Q&A prep, term-sheet comparison. Company controls the analytics; lender controls the decision. *Compressibility: 7.5, Criticality: 7.4, Control: Semi-controlled.* 6. **Environmental compliance due diligence** — Phase I/II report assembly, liability mapping, regulatory baseline documentation, insurance package prep. Company controls the doc work; external counsel/insurers control signoff. *Compressibility: 7.1, Control: Semi-controlled.* 7. **M&A data room assembly** — Workbook standardization, cross-ref consistency, missing-doc flagging, title and regulatory document tagging. Sell-side controls the room; buyer's counsel controls the process. *Compressibility: 7.7, Control: Semi-controlled.* 8. **Acreage disposition analysis** — Acreage value calc support, divestiture candidate ranking, executive summary drafting, board-package support. Company controls analysis; board/partners control decision. *Compressibility: 7.3, Criticality: 6.9, Control: Semi-controlled.* --- ## Later: 18-36 Month Horizon **Criteria:** High value + either low compressibility or high external governance + evidence maturity still building These workflows matter enormously but require either: - Field validation (reservoir engineering, well engineering) - Real artifact volume (nuclear licensing workflows, LNG shipping) - Competitive capability (trading decision support, rig operations) - Regulatory clarification (what counts as "AI-assisted" vs. automated signoff) They belong in the backlog now because the infrastructure (eval frameworks, domain models, regulator guidance) is still being built. The company should start problem definition work and pilot data collection now. **Risk profile:** Moderate to high. These are the high-criticality, high-reversibility decisions. Mistakes are expensive. The model maturity must exceed internal expectations, not just be "better than a spreadsheet." **Deployment pattern:** Start with composite proof rooms and simulation environments. Run side-by-side comparisons on 6-12 months of historical cases. Build domain-specific eval frameworks (e.g., reserve estimate accuracy, well-completion timeline prediction, trade P&L attribution). Get regulator feedback on automation boundaries early. Then pilot on prospective cases with heavy human review. **Implementation examples:** 1. **Reservoir engineering augmentation** — Production forecast support, pressure/depletion modeling, well-performance analysis, decline-curve analysis. Low compressibility (real physics, site-specific constraints); high criticality (capital allocation, reserve booking). Company controls; SEC defines reserves rules. *Compressibility: 7.6, Criticality: 5.71, Control: Company-controlled.* Timeline: needs 18+ months of field validation against actual well performance. 2. **Nuclear licensing support** — Technical specification analysis, safety case assembly, NRC interrogatory response, design documentation. High prep-stack compressibility (document assembly, cross-referencing, interrogatory drafting are highly automatable), but regulatory signoff remains the bottleneck — NRC controls the approval timeline and AI-assisted submissions are uncharted. Ultra-high criticality (safety, license retention). *Compressibility: 8.7, Criticality: 9.08, Control: Company-controlled.* Timeline: NRC guidance on AI use in licensing still emerging; requires explicit regulator engagement. 3. **Real-time trading decision support** — Volatility modeling, position-sizing recommendation, curve-shape analysis, market-event triage. High reasoning demand; high recurrence; moderate compressibility. Desk controls execution; market controls outcome. *Compressibility: 6.1, Criticality: 8.3, Reasoning demand: 8.6, Control: Company.* Timeline: needs 12-18 months of live P&L attribution to validate. 4. **Well intervention decision support** — Workover candidate ranking, intervention type recommendation, prognosis modeling, rig logistics optimization. Low compressibility (field conditions, operator judgment); high criticality (CAPEX, safety, production). *Compressibility: 4.7, Criticality: 8.9, Control: Company.* Timeline: needs field validation, rig-level side-by-side testing. 5. **Rig operations support** — Drilling parameter optimization, wellbore-stability modeling, casing-design assistance, stuck-pipe risk assessment. Low compressibility (real-time field data); high criticality (safety, NPT, well quality). *Compressibility: 4.3, Criticality: 8.7, Control: Company.* Timeline: requires downhole sensor integration and 12+ months of offset-well validation. 6. **Transmission operations support** — SCADA data anomaly detection, outage prediction, switching-order support, islanding-scenario modeling. High reasoning demand; high criticality (grid stability); moderate compressibility. *Compressibility: 6.3, Criticality: 8.8, Reasoning demand: 8.4, Control: Semi-controlled.* Timeline: NERC approval for automation boundaries; 12-18 months field testing. 7. **Subsurface characterization for carbon storage** — Geologic risk assessment, cap-rock integrity analysis, migration modeling, corrective-action scenario analysis. Low compressibility (site-specific geology); high criticality (permanence, compliance). *Compressibility: 4.1, Criticality: 8.6, Control: Semi-controlled.* Timeline: EPA/state CO2 storage rules evolving; needs regulatory alignment. 8. **Investment committee memoranda** — Project ranking, NPV sensitivity, scenario modeling, board-readiness assessment. High value; moderate compressibility; humans retain decision authority. *Compressibility: 6.8, Criticality: 8.4, Control: Company.* Timeline: 12-18 months to build enough case studies for confidence. --- ## Never Alone: Always Human-in-Loop **Criteria:** High criticality + high irreversibility + safety or regulatory signoff required These workflows will not be automated. They are decision nodes where human judgment and accountability matter. The role of AI is to compress the prep stack, surface risks, and let the human decide faster. The human retains authority and bears the liability. **Risk profile:** Ultra-high. These are the decisions where the company can be sued, the executive can be prosecuted, the license can be revoked. AI is a tool for analysis, not a substitute for judgment. **Deployment pattern:** AI handles evidence gathering, scenario assembly, and impact modeling. Humans handle judgment, negotiation, and final signoff. The system should make it harder to hide evidence, not easier to dismiss it. Auditability is not optional. **Implementation examples:** 1. **Board of directors signoff** — Strategic decisions, major M&A, capital allocation, executive succession. High irreversibility; human accountability mandatory. AI supports the memo (scenario modeling, precedent analysis, risk assessment); humans decide. *Criticality: 10.0, Compressibility: 2.1, Control: Company.* Success metric: board confidence in evidence, not speed of decision. 2. **NRC licensing authority** — Operational authority, emergency procedures, license-amendment decisions. Ultra-high regulatory bar; safety is non-negotiable. AI handles document assembly and precedent search; NRC staff makes the judgment. *Criticality: 10.0, Compressibility: 1.8, Control: Externally governed.* 3. **Well intervention decision** — Whether to kill a well, sidetrack, or attempt repair. CAPEX at risk; safety implications; irreversible. Operator decides; AI assembles the prognosis and risk case. *Criticality: 9.4, Compressibility: 3.2, Control: Company.* 4. **Trading position limits and hedging strategy** — CEO and CFO set the appetite; risk committee approves. AI models the exposure; humans decide the posture. *Criticality: 8.7, Compressibility: 3.8, Control: Company.* 5. **Environmental compliance commitment** — Regulatory agency decides compliance path; company commits to it. AI supports the compliance memo; regulator and company counsel decide the commitment. *Criticality: 9.1, Compressibility: 3.6, Control: Externally governed.* 6. **Lender covenant waiver decisions** — Lender decides whether to waive; company negotiates. AI models the impact; treasury and counsel negotiate the outcome. *Criticality: 7.8, Compressibility: 3.4, Control: Externally governed.* 7. **Emergency response authorization** — Spill response, blowout response, operational shutdown. Safety-critical; time-sensitive; irreversible. AI assists with risk assessment and procedure routing; operators make the call. *Criticality: 9.8, Compressibility: 2.7, Control: Company.* 8. **Operator/non-op dispute resolution** — JOA interpretation, force majeure claims, penalty clauses. Legal judgment required; financial stakes high. AI helps with contract clause extraction and precedent search; counsel and executives decide the settlement posture. *Criticality: 8.3, Compressibility: 2.9, Control: Semi-controlled.* --- ## How to Use This Map **For strategy:** Pick one "Now" workflow per business line. Pilot for 12 weeks. Measure cycle time, error rate, and confidence. Roll out to parallel operations. Then move to "Next" tier. **For risk:** Flag the "Never alone" workflows in compliance and audit planning. These require explicit AI use policies and human signoff protocols. Do not attempt to automate them. **For vendor evaluation:** Score vendors on their ability to handle "Now" tier workflows first. Ask for specific artifacts, not generic case studies. Ask how they handle the human signoff gate. **For regulatory engagement:** Get ahead of the "Later" tier by engaging regulators now (NRC, NERC, FERC, state PUC). Ask what "AI-assisted" means to them. Build the approval path before you need it. --- ## Appendix: Deployability Gaps by Workflow Class **Highest deployability readiness** (ready in 0-3 months): - Production accounting close support - JIB reconciliation - Trade capture/confirm/settle - Borrowing base packet assembly **High readiness** (ready in 6-12 months): - CIM production - Division order processing - Rate case evidence assembly - Environmental permitting packages **Moderate readiness** (ready in 18-24 months): - Reservoir engineering augmentation - LNG shipping decision support - Nuclear licensing support - Transmission operations support **Lowest readiness** (requires 24-36+ months + regulatory alignment): - Well intervention decisions - Board-level strategic decisions - NRC licensing authority - Rig operations optimization The gap between "Now" and "Never alone" is the gap between compression (making humans faster) and automation (replacing human judgment). Compress the prep stack. Respect the decision gate.