K'nexion Energy home

AI Solutions

K'nexion Energy is vendor-agnostic and human-in-the-loop. The data foundation comes first: OSDU, lineage and context, before any model is trusted. The goal is agentic AI that answers to the geologist.

Many noisy wavelet traces converging into one clean signal trace
The many traces: today's AI market noise.The one clear signal: K'nexion's recommended solution.

Why a geologist?

I've spent my career on the subsurface side of the decision. I know what the real question is, what data it needs, and what "right" looks like to the asset team. I speak both languages.

AI: noise vs. signal

The hype is loud. The value is real. The answer is connecting the data you already have to the decisions that depend on it, then pairing that with the right solution.

Vendor-agnostic, always

I work for the problem, not a platform. The roster is curated and growing. If the best fit isn't on it, I'll find who is.

From the NOVA 2026 talk: the business case

OSDU as the Bridge to AI-Enabled Subsurface Decisions — NOVA 2026, Benghazi, Libya, 7 Oct 2026

AI 'noise' vs. 'signal': the biggest bubble humanity has ever witnessed.

NOVA 2026 slide 3, Introduction: Business Questions and Problems. The slide text is given below on this page.
NOVA 2026, slide 3
NOVA 2026 slide 4, Introduction: Business Decisions. The slide text is given below on this page.
NOVA 2026, slide 4

Four questions that take profound human effort to answer fully, but that AI can be trained to answer 1000x faster:

  1. How do we consistently improve our chance of success when drilling new wells?
  2. Why are these wells performing poorly, and what can we do to improve them? Show me the detailed plan.
  3. How can we stay compliant with all our regulatory requirements? Where is our biggest exposure?
  4. Where in our field are the best upcoming well target locations?

Experience from a supermajor

Rebuild a full reservoir model
15–24 months
History-match that model
3–6 months
Make the decision
and then the decision…

When the work becomes the problem: too much time, money and stress.

Business decisions

  • Is AI making people's decisions worse?
  • More data and more knowledge often make decision-making harder, not easier
  • Why is technology making life harder, not easier?
  • Is technology creating a generation of bad decision makers?

A good decision is not one that leads to a good outcome. It is one that was the best choice given what was known at the time.

Technology reduces uncertainty about facts, but increases awareness of complexity.

The technological developments of the last decade have made poor decision making easier, more immediate, and more widely consequential.

Sustainable improvement comes from answering 'What did we learn?' and 'How should we act differently next time?'

The fix isn't more data. It's structuring, refining and connecting the data you already have to the decisions that actually depend on it.

A study of FID forecasts against actual production for 56 Norwegian Continental Shelf fields found forecasts were consistently optimistic. Source: Bratvold, Mohus, Petutschnig and Bickel, 'Production Forecasting: Optimistic and Overconfident—Over and Over Again,' SPE Reservoir Evaluation & Engineering, Aug 2020, pp. 799–810 (SPE-195914-PA).

Neurocraft Data Services

K'nexion Energy partners with Neurocraft (neurocraft.ai (opens in a new tab)).

NeuroEarthIQ

60,000+ Permian Basin sample-laboratory logs

A supervised multi-agent pipeline that extracts lithology intervals, stratigraphic tops, handwritten notes and header and temperature fields from scanned mud logs: the 30-minute-per-log manual task.

Demos with the USGS; intended to scale nationally.

NeuroMineIQ

Earth MRI mine-waste proof of value

Sites ranked from satellite, airborne, drone and hyperspectral layers plus federal records, with confidence gates and USGS SME review.

No unreviewed AI output in a scientific conclusion.

My role: principal geoscience director, providing domain-expert validation. The intellectual property is owned by Neurocraft.

Neurocraft orchestrated supervised agentic AI architecture: human expert supervision, specialist agents, subsurface intelligence layer, insights to action, built on Microsoft cloud
Image courtesy of Neurocraft, used by permission.

AI reselling & partners

K'nexion connects operators to AI solutions and partners. Each listing is a referral: I send operators where I'd send my own team.

Over the past year I worked as Director of Technical Business Development for MangoBytes AI and Neurocraft AI, and as President of RockNRG. That work built a large network in the AI space for operators, and it is why I can match an operator's problem to the vendor who can actually move it.

AI solutions

  • Noble.AI logo

    Noble.AI

    Surrogate reservoir modeling

    Science-based AI and surrogate reservoir models: physics-informed proxy models for fast scenarios.

    >1,000× faster reservoir model calibration Vendor-reported

    Partner case studies
    What it does
    • Physics-Informed Reservoir Analytics: Integrates field, geological, petrophysical, and production data with physics-constrained AI to validate models, estimate recovery, and assess injector–producer connectivity.
    • Static QC, Saturation & Reserves: Automates geological consistency checks, reconciles saturation vs. production, and generates P1/P2/P3 polygons for layer-level volume estimation.
    • Waterflood & Production Engineering: Identifies layer-specific connectivity, sweep efficiency issues, and bypassed zones; automates well diagnostics.
    Results
    • Cut reservoir model calibration time by >1,000×, enabling rapid scenario evaluation and more confident field development decisions. Vendor-reported
    • Improve asset performance and recovery through optimized materials, surfactants, polymers, and operating conditions, informed by physics- and chemistry-based modeling. Vendor-reported
    • Increase R&D productivity by up to 20×, unlocking greater throughput without expanding headcount. Vendor-reported
    Published work

    Gala, D., Becker, G., Kaul, K., DallAqua, M., Hegde, A., Moisselin, S., Fuda, C., Doraiswamy, S., Verma, V., and Wu, X. 2023. "AI-Powered, Lightning-Fast Production Modeling of Multi-Well and Multi-Bench Unconventional Development." SPE-214818-MS, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 16–18 October 2023. https://doi.org/10.2118/214818-MS (opens in a new tab)

    (Co-published by ExxonMobil and Noble AI.)

    The follow-on ADIPEC 2025 paper SPE-229440-MS reports a multi-scenario history match in ~2.5 hours versus ~160 hours with a conventional simulator. Vendor-reported

    Figures marked Vendor-reported are published by Noble.AI and not verified by K'nexion.

  • Neurocraft logo

    Neurocraft

    NeuroEarthIQ™

    Industrial and agentic AI on Microsoft Azure and Fabric: NeuroEarthIQ, NeuroMineIQ, NeuroVisionIQ.

    15 expert agents on one platform Vendor-reported

    NeuroEarthIQ and NeuroMineIQ, above

  • MangoBytes AI logo

    MangoBytes AI

    Part of Mango Group

    Agentic AI for E&P operations: energy AI agents on Microsoft Azure.

    <20 min workover planning, down from ~50 engineer-hours Vendor-reported

    Partner case study: Texian Operating
    Case study: Texian Operating
    • Well workover planning: ~50 engineer-hours down to under 20 minutes; approval in 15 minutes from the superintendent's phone (procedure, AFE, vendors, economics). Vendor-reported
    • Wells back online in 1–2 days instead of a week or two. Vendor-reported
    • Regulatory: a week's worth of TRRC filings done in 10 minutes; 1,200 wells handled by half of one person's time. Vendor-reported
    • Power management: $600K electricity savings in half a year in one South Texas field using agents to buy spot power. Vendor-reported
    • Texian Operating scaled from 600 to 1,200 wells (progressing to 1,800) across TX, LA, NM with the same back-office staff. Vendor-reported
    The platform
    • 75+ energy AI agents; Pixie super-agent; AI Analytics (dashboards, what-ifs); AI Studio (build agents in plain language). Vendor-reported
    • RRC production reports ~4 days → 10 minutes. Vendor-reported
    • Built on Microsoft Azure with zero data retention.

    Case-study figures are reported by MangoBytes AI and not verified by K'nexion.

  • i2k Connect logo

    i2k Connect

    Knowledge and unstructured data

    Knowledge and unstructured data: decades of reports turned into answers you can verify.

    4,000+ classes, at petabyte scale Vendor-reported

    What it does

    An AI platform purpose-built for geoscientists and engineers, 10+ years in the making. Vendor-reported

    • AI data readiness: crawl, enrich, standardize and index.
    • A unified knowledge graph, accessible via MCP or i2k agents.
    • Content Discovery across SharePoint, Teams, OpenText, databases and OSDU.
    • Document Enrichment: summaries and multi-taxonomy classification.
    • Oilfield Places entity recognition: basin, block, field, well, formation.
    • Petabyte scale, 4,000+ classes; on-premises or cloud. Vendor-reported
    • Developed the EnRG-LLM energy language model with Aramco support, licensed through SPE.

    Figures marked Vendor-reported are i2k Connect's own.

The data foundation

  • RockNRG Consulting logo

    RockNRG Consulting

    Data foundation and OSDU

    Data foundation and OSDU: trusted data ecosystems under every AI solution.

    Founded in 2023 by Ryan Jarvis, RockNRG is a cross-domain SME delivery team integrating geoscience, engineering and IT. It offers SME domain consultants, workflow optimization, modernization strategy (data management, OSDU industry standards, business and data-quality rules), AI-enhanced subsurface data ecosystems, and agentic AI automation with training. RockNRG is vendor-agnostic and operator-aligned.

    Company description from RockNRG's own materials.

The DIKW pyramid: Data at the base, then Information (+ meaning), Knowledge (+ context) and Wisdom (+ insight) at the top
Ryan Jarvis and Bjarne Bøklepp figure, used as drawn. Data: the measurement. Information: the measurement with a meaning. Knowledge: that meaning, with scale, uncertainty, and purpose. Wisdom: the decision, the outcome, and what we would do differently. The textbook pyramid assumes perfect information. A field does not.

How an engagement works

You bring the problem; I bring two or three real fits and a neutral read on each.

The MATCH process

  1. Map the problem
  2. Assess vendors
  3. Test with a Pilot Scorecard: success metrics, data, timeline and kill criteria, with an optional two-vendor bake-off
  4. Choose
  5. Hand-off

Four ways we work together

  • VNA Sprint (Vendor Neutrality Assessment)

    $500 single consultation meeting, or $165/hour across the sprint.

    It's a consultation fee only.

    Book a VNA Sprint

  • Operator / Vendor Roundtable

    $25 per seat monthly on the fourth Tuesday.

    Held live on Microsoft Teams at 6 p.m. CT, open to anyone interested. Operators and builders work through one real problem together. Presenters: $250 for 10 minutes on their platform and one case study.

    Request a seat

  • K'nexion Energy Podcast

    $1,000 Initiation Episode.

    Real conversations with people who've built working agentic AI stacks: a long-form episode, plus short-form clips across 5 platforms. Website listing after launch and a signed agreement: $500.

    Apply for an Initiation Episode

  • K'nexion Talent Bridge

    Free no fee.

    Between roles? AI companies need real domain experts to review and critique what their models produce. Register your interest for a referral link to Mercor, which pays experts for that work on contract.

    Register interest

    K'nexion Energy is independent. It is not Mercor, and it doesn't recruit or speak for them. Mercor decides who it engages. K'nexion may earn a referral reward under Mercor's published program terms (opens in a new tab).

Talks

  1. IMAGE 2026
  2. Oil & Gas Digital Transformation Conference Sept 2026
  3. NOVA 2026 Benghazi, Oct 2026 OSDU as the Bridge to AI-Enabled Subsurface Decisions: Accelerating Exploration, Development & Production
  4. VIP invitation to LEES 2027

Available as an in-house session.

Ben Hilton speaking into a microphone in front of a presentation screen at IMAGE
Speaking at IMAGE 2026.

Start with a short call.