WillCo Digital Drilling Intelligence Report — Edition #2 | Week of Jun 01 – Jun 07, 2026

William Contreras
3 days ago
4 min read

This edition at a glance

Agentic AI (TADI framework) validated on Volve dataset — multi-source optimization static ML structurally cannot replicate
Baker Hughes Kantori™ with Equinor: quantified ROP & NPT gains under live North Sea field conditions
Standardize rig automation API against H&P FlexRobotics spec before contract award — not after mobilization

Why this edition matters

Drilling AI is bifurcating. One branch is maturing into task-specific automation — GenAI reporting tools, parameter recommendation engines, discrete workflow assistants — and it is scaling fast into routine rig operations. The other branch is newer and more consequential: agentic reasoning systems that autonomously orchestrate data across multiple sources, iterate through engineering logic, and surface optimization windows that no static model would find. The IADC World Drilling 2026 conference in Lisbon was the clearest public confirmation yet that both branches are generating commercial outcomes.

The question operators must now answer is not whether to adopt digital drilling intelligence, but which architecture to fund first — and how to avoid locking into a vendor stack that cannot interoperate with the rigs they actually contract.

Key developments this week

🔬 TADI agentic framework — Volve dataset validation

Agentic AI · Dataset validation

TADI iterates across offset well logs, real-time drilling parameters, and formation lithology markers simultaneously, adjusting its query logic based on intermediate results. Validated results: 2–4 non-obvious parameter optimization windows per well, reproducible on unseen wells from the same formation sequence. Correlates NPT event signatures with drilling parameter history without requiring pre-labeled NPT categories.

📈 Baker Hughes Kantori™ — Equinor, North Sea

Physics-AI hybrid · Live field deployment

Quantified ROP improvement and NPT reduction confirmed under live field conditions. Physics-based wellbore models distinguish mechanical dysfunction from formation response — a distinction pure data models routinely fail to make. Ask vendors for equivalent live-well performance data, not simulator benchmarks.

⚙️ SLB DrillOps — 21% ROP improvement confirmed

Closed-loop automation · Commercial deployment

Confirms 21% average ROP improvement across automated sections vs. manual baselines. Primary driver: consistent parameter execution, not superior parameter selection — eliminating human variability in WOB and RPM management during connection sequences and formation transitions.

🔗 H&P FlexRobotics API standardization — IADC World Drilling 2026

API standard · Contract-award decision

Open API specification confirmed at IADC World Drilling 2026 in Lisbon. Rigs running FlexRobotics-compatible control architecture accept standardized automation integrations. Those that do not require custom middleware — adding both cost and latency to any real-time control loop.

📋 Shelf Drilling — GenAI reporting scales across fleet

GenAI workflow · Jackup operations

Shift reports, morning reports, and NPT summaries generated automatically. 40% reduction in post-well QC reclassification effort. Real mechanism: structured NPT taxonomy enforcement at data entry — not the AI itself.

Operational data highlights

Connection time: Automated sections averaging 3.2–3.8 min vs. 4.6–5.1 min manual baseline — at roughly one-third the variance. Compounds across 200+ connections in a typical deepwater well.

NPT rate: Wells with integrated real-time wellbore stability monitoring holding below 8% on high-risk North Sea intervals vs. historical fleet average of 12–14% on equivalent formations.

Data infrastructure: Wells with standardized WITSML feeds generating 60–70% faster time-to-insight on offset well analysis vs. wells requiring manual extraction.

WillCo perspective: what the engineering says

I have spent three decades watching drilling technology cycles, and the current one is different in one specific way: the gap between "technically demonstrated" and "commercially deployed at scale" is compressing faster than it ever has before. Baker Hughes Kantori™ and SLB DrillOps are not pilots. They are commercial deployments with auditable performance numbers.

You can license the best drilling AI platform available and get mediocre results if your WITSML data stream has gaps, your formation tops are inconsistently defined, and your NPT records live in a spreadsheet nobody has touched since 2023.

Before spending a dollar on a new platform, spend two weeks auditing your existing data infrastructure. What you find will tell you more about your readiness than any vendor demonstration will.

Practical recommendations

Run an agentic AI proof-of-concept on your offset dataset before your next campaign. 60-day structured PoC using an agentic reasoning framework against your offset well logs, drilling parameters, and NPT records. TADI on Volve confirms 2–4 actionable parameter windows per well at formation transition points where static ML models structurally fail.
Add FlexRobotics API compatibility to your rig contracting checklist — now. Resolving API fragmentation at contract award costs nothing. Resolving it at mobilization costs six weeks and custom middleware development.
Define a standardized NPT taxonomy before deploying any reporting AI. Shelf Drilling's 40% QC reduction is attributable to taxonomy enforcement at data entry, not to the AI itself.
Audit your data infrastructure before any platform investment. Two-week focused audit of WITSML data streams, formation top consistency, and NPT completeness on your last five wells before signing any new analytics contract.

What we're watching next

First confirmed live-well agentic closed-loop parameter control deployment with full operator liability acceptance — likely 12–18 months out
IADC World Drilling 2026 post-conference technical paper releases from Lisbon — additional H&P FlexRobotics and automation standardization detail
Halliburton-Sekal integration milestones — whether the combined Landmark-Sekal automation stack produces interoperability with non-Halliburton surface systems

William B. Contreras is the founder of WillCo Drilling Consulting, an independent drilling engineering firm specializing in digital drilling strategy, hybrid physics-ML analytics, and RTOC support. He holds 7 drilling technology patents and has published 30+ technical papers, with 30+ years of international experience.

Contact: willcodrilling.com | LinkedIn: linkedin.com/in/william-contreras-otalvora/