Customer Stories

How Geologic Cut Document Processing Costs 73% Across 24 Drilling Report Formats

Client
Geologic
Industry
Oil and gas
Partners
Foundation model
Product Types
Applications
Drilling report analysis, data extraction, document classification

57,000+ pages processed

in August 2026, across 1,741 documents

91.9% job completion rate

in August 2026

73% lower processing cost

per document

Geologic
Multimodal has played a key role in advancing our ability to scale complex document-to-data workflows, significantly improving throughput and processing efficiency. They have enabled the extraction of complex information from unstructured documents into structured, decision-ready data—supporting more scalable operations and better business outcomes. The team has been highly responsive and consistently focused on delivering practical, high-impact results.
Stephen Werny
Director, Data at Geologic

Challenge

Solution

Results

The Problem

Twenty-Four Ways to Write Down the Same Well

Every day a well is being drilled, someone writes a report about it. Depth reached. Time logged. Bits used and bottom hole assemblies. For Geologic, the Calgary company that supplies well data to operators, analysts and service companies across Canada, those reports are a key part of the business. For most of the company's history, turning them into data meant having a person read them and type them, then pivoting to rudimentary automation on a limited number of format types.

A drilling report is not a form. The same fact sits in a different place depending on who filed it and what software produced it. Geologic sends 24 distinct layouts into this pipeline. Some run past 200 pages. Some wrap one reporting format around another inside a single file.

That variety defeats conventional automation. OCR can extract text from a page, but it cannot tell you which number represents the depth at the start of the tour and which represents the total drilled. Rules break the moment a table splits across a page. And the output has to be exact, because it feeds Geologic's data model and, from there, their customers. A value in the wrong column is worse than no value at all.

So the work remained a combination of manual processes and limited automation for a handful of formats, and the limitations in those processes set the ceiling on how much data the company could publish.

The Solution

One Format at a Time, Inside Geologic's Own AWS Account

Multimodal deployed the pipeline inside Geologic's own AWS account, so no document ever leaves their environment. Each layout gets its own extraction logic, built and tuned against real reports until the output matches Geologic's data model expectations, and then released into production.

Through 2026, the extraction pipeline was re-engineered, cutting processing costs per document by 73% and improving reliability across all formats.

A Rollout the Customer Controlled

The hard part was never the engineering. Formats already running in production could not be allowed to regress, and only Geologic's own developers could confirm that. So the work was structured around their review rather than ours.

Formats were released in batches so testing could start on day three rather than at the end. Each one went live only after Geologic's data team validated the output against the source PDFs, and anything their QA surfaced went back into the queue until it was right. Page-level logging was added so their team could see for themselves which pages hit limits and which were skipped, rather than having to file a ticket and wait.

That cadence continues today. An issue raised in the morning is usually deployed within a day or two.

The Results

More Pages, Lower Cost Per Document

In August 2026, the pipeline processed more than 57,000 pages across 1,741 documents and completed 91.9% of jobs. This is a significant increase in the number of pages processed before the rebuild and an improved completion rate.

As of September 2026, 15 of the 24 formats are in production, with the remainder in testing or queued.

Why it worked

The interesting thing here is not that AI read some documents. It is that the document set was genuinely hostile: two dozen layouts, no shared schema, hundreds of pages, exact output required, and a customer who could verify every field and would. What made it work was building one format at a time, on the customer's own infrastructure, with the customer holding the release gate.

Key results

Volume

  • More than 57,000 pages processed across 1,741 documents in August 2026

Reliability

  • 91.9% of jobs completed in August 2026

Cost

  • 73% lower processing cost per document after the 2026 rebuild, compared with the original pipeline

Coverage

  • 15 of 24 drilling report formats in production as of September 2026
  • The remaining 9 formats are in testing or queued
  • Every format goes live only after Geologic's data team checks the output against the source PDFs

Turnaround

  • Issues raised by Geologic are usually fixed and deployed within a day or two
  • Page-level logging shows Geologic's team which pages hit limits or were skipped, without filing a ticket
About Geologic

Geologic brings clarity and precision to the world's most complex energy decisions. For over four decades, it has provided decision-ready energy data, market intelligence, and software that improve performance, optimize operations, and unlock opportunity.

Client
Geologic
Industry
Oil and gas
Foundation Model
Product Types
Applications
Drilling report analysis, data extraction, document classification
Use Case
Drilling report analysis, data extraction, document classification

Book a
30-minute demo

Explore how our agentic AI can automate your workflows and boost profitability.

Get answers to all your questions

Discuss pricing & project roadmap

See how AI Agents work in real time

Learn AgentFlow manages all your agentic workflows

Uncover the best AI use cases for your business