Talking to the Ground

The Business Context: Unifying the Subsurface and the Surface

Drilling operations are complex, involving geology, mechanics, and business performance. Most organizations improve these areas independently (e.g., OSDU for subsurface, rig IoT, modern ERP systems) but lack a unified data platform for combined analysis, security, and metrics. This makes cross-domain analysis a series of custom, one-off projects.

Operational excellence now requires correlating these distinct datasets; knowing why subsurface conditions caused equipment failure, not just that it failed. Historically, this was difficult, requiring extensive coding and time.

The Databricks Data Intelligence Platform and natural language analytics change this by unifying data and democratizing access to complex insights. Users can now ask simple questions, such as identifying that the Travis Peak formation causes 50% of pump failures, lowering the barrier to entry. This shifts data from a retrospective record to a real-time operational partner providing audit trails and actionable recommendations quickly.

As margins tighten, the real-time ability to correlate subsurface conditions, equipment performance, and operational outcomes is essential. Systematically reducing NPT, recovering fleet capacity, and avoiding millions in costs makes timely analytics a key driver of EBITDA, capital efficiency, and asset utilization, transforming data into an operational asset for smarter, faster decisions.

Put simply, analytic competency is profit.

The Challenge: Unanswered Questions Cost Millions

Every drilling operations manager faces the same daily frustration: critical insights buried across siloed systems, equipment failures that go undiagnosed for days, and root cause analyses that take weeks instead of minutes.

The operational toll is significant:

The result? Equipment failures and challenges related to formation lead to unplanned downtime, costing drilling operators millions in NPT each year. This figure does not even account for the additional expenses incurred from deferred production, repair costs, and supply chain disruptions.

The Solution: Conversational Analytics on Unified Data

Operations managers ask the Databricks Genie Research Agent questions and get a multi-step analysis linking IoT sensor data, OSDU well logs, and ERP systems.

Research Agent extends Genie's capabilities to help you uncover deeper insights and tackle complex business questions using multi-step reasoning and hypothesis testing.

What Genie Delivers

Introducing AI-Powered Operational Intelligence for Drilling Operations

Built on the Databricks Data Intelligence Platform, this solution transforms raw operational data from multiple sources into actionable insights through natural language conversations. The solution brings OSDU well logs, rig IoT streams, and ERP maintenance/financial records together into a single, governed lakehouse, so every team, from drilling to subsurface to finance, works from the same source of truth.

The Demo Scenario: A Day in the Life of an AI-Augmented Operations Manager

A drilling operations manager at DeepCore Energy begins their day by opening Databricks and asking Genie Research Agent a simple question. Unlike traditional dashboards that show only pre-configured views, Genie creates a research plan, runs multiple SQL queries against the unified lakehouse, and delivers a comprehensive operational picture.

Outcome Snapshot

• Fleet-level NPT visibility across 118 wells and multiple formations via a single natural language query.
• Rapid root cause analysis correlating pump failures with formations and mud weights, saving weeks of manual work.
• A quantified action plan recovering 64–91 days of fleet capacity and avoiding $1.6–2.7M in costs through formation-aware maintenance.

Question 1: “What’s our current average NPT and why?”

What Genie Does Behind the Scenes:

Genie doesn't execute a single query. Instead, it generates hypotheses, runs multiple analyses (see Fig. 1 in the right sidebar), and synthesizes findings:

• Fleet performance across all 118 active wells
• County-level NPT breakdown (Loving, Midland, Ector, Ward, Reeves)
• Formation-level assessment (Bone Spring, Spraberry, Delaware, Wolfcamp, Avalon)
• Root cause categorization: equipment, geological, procedural

This is where the architecture becomes truly transformative. The operations manager's questions don't return simple query results, they trigger comprehensive multi-step analysis that correlates data across the entire unified platform.

The Response:

DeepCore Energy's 118-well Texas fleet is operating with stable baseline performance (6.88% average NPT). Performance is remarkably uniform, county-level NPT ranges from 6.33% to 7.21%, less than 1 percentage point variation.

The root cause breakdown of NPT reveals that equipment issues, especially related to mud pumps, are the dominant constraint on fleet efficiency, accounting for almost half (47.52%) of all NPT minutes.

Question 2: "Why are my mud pumps failing so often?"

Traditionally, reliability engineers and subsurface teams would each run separate analyses, then try to reconcile findings manually. With all the data unified on Databricks and exposed through Genie, the system correlates failure modes, MTBF, formation exposure, mud properties, and maintenance history in a single multi-step analysis.

What Genie Analyzes:

• Failure mode distributions across pump units and component types
• Mean time between failure (MTBF) patterns
• Correlation with operational parameters (depth, mud weight, flow rate)
• Equipment alarms and their relationship to geological formations
• Work order patterns and maintenance sustainability

The Response:

The analysis reveals a systemic reliability crisis: mud pumps are failing at 8.5 work orders per day (765 total in 90 days), affecting all 118 wells. Genie lists three primary failure modes: Liner Wear, Seal Leaks, and Bearing Failures, indicating simultaneous degradation across multiple components, not isolated part failures.

An analysis correlating pump failures with OSDU geological data revealed that the Travis Peak formation, which requires a 6% heavier mud weight, accounts for 50% of pump alarm events due to increased hydraulic pressure and abrasive forces accelerating mechanical wear.

Question 3: "Tell me about the formation causing these issues"

The Response:

Travis Peak is a fractured, vuggy carbonate reservoir spanning 9,600-10,049 ft TVD with geological characteristics that create the conditions driving mud pump failures. It presents significant drilling challenges due to highly elevated average pore pressures (up to 10.62 PPG) and a high risk of fluid loss, indicated by a Loss Risk Index of 0.70 and affecting 84% of wells.

Question 4: "What can I do to reduce NPT?"

The Response:

The Genie Research Agent offers a dual approach to well optimization. Immediate actions (1-2 weeks), such as specific mud pump maintenance like liner replacement intervals, are provided alongside a set of long-term strategies (6-month horizon). These long-term initiatives include automated torque limiting, mud weight optimization, and other related actions.

Because the action plan is driven by the same unified dataset and modeling, operations managers can see not just what to do, but how much NPT and cost each intervention is likely to recover, helping prioritize work across rigs and partners.

Well Path Reference Architecture on Databricks

Architecture: How It Works

The Databricks Lakehouse, structured as a Medallion architecture, is ideal for analytics, organizing data across three layers. The Bronze layer contains raw data like OSDU well logs, IoT streams, and ERP records. This data is cleaned and enriched in the Silver layer with standardization, formation metadata, and equipment ID mapping. The Medallion architecture replaces scattered integrations with a unified foundation. Instead of each team building its own NPT or MTBF logic, the Gold layer standardizes these metrics and makes them accessible to Genie, BI tools, and predictive models.

Data Sources & Integration

The new solution can significantly boost business value by delivering faster insights in minutes using natural language queries instead of weeks of manual analysis, correlating root causes across previously siloed data (operations, equipment, and geology), enabling proactive and predictive actions, and democratizing data access for all stakeholders through simple queries, eliminating the need for specialized SQL.

Quantifiable Business Outcomes

Unified data platforms with AI-powered analytics drive significant improvements for organizations, including:

• Reduced NPT: By proactively addressing formation-specific obstacles, organizations minimize NPT before issues escalate.
• Minimized Equipment Downtime: Predictive maintenance, which correlates potential failures with geological conditions, leads to reduced equipment downtime.
• Accelerated Decision-Making: Critical insights are delivered in minutes instead of weeks, enabling faster decisions.
• Optimized Capital Allocation: Data-driven prioritization, based on quantified Return on Investment (ROI), ensures capital is allocated more effectively.

For a personalized demo and discussion on transforming your drilling operations with AI-powered natural language analytics, contact your Databricks representative.