Request a demo
ArcelorMittal Cuts Patent Analysis Time by 90% with Axion™
By automating data extraction, ArcelorMittal reduced competitive patent analysis workload by over 90%, turning a weeks-long task into a rapid AI workflow.
( Scale )
120+ complex patents/month, 68 languages
( Results )
60Ă— faster data extraction, 94% accuracy
( Company Overview )
Industry
Steel & Mining
Company size
125,OOO+
Department
Patent and Global R&D
Use Case
Patent Data Extraction
More info
ArcelorMittal is the world’s leading steel and mining company, headquartered in Luxembourg. It operates in more than 60 countries and has an industrial footprint in over 15 of them, with a focus on producing safe, sustainable steel for various sectors including automotive, construction, and energy.
Close
( 01 )
Their Challenge
ArcelorMittal conducts ongoing competitive patent analysis to spot market gaps, uncover patentable solutions, and monitor for potential infringements. The team tracks over 120 competitor patents each month, but manually extracting critical data from each could take up to 4 hours. That’s up to 480 hours per month! This slow, labor-intensive process delayed innovation and reduced their ability to respond quickly to market changes.
( 02 )
Our Solution
Axion™ by Iris.ai automatically analyzes patents by identifying domain-specific entities and linking experimental data across text, tables and graphs. It transforms complex, unstructured patent files into structured, machine-readable outputs enabling:
- Faster decision-making
- Better resource allocation
- Shorter time-to-market
Our image captioning feature in action
( Input Image )
( Ground source and AI generated output )
Improved Thermal Processing of Polylactic Acid/Oxidized Starch Composites and Flame-Retardant Behavior of Intumescent Non-Wovens
"SEM micrographs of the intumescent char residue on fibers from PLA/OS-based non-wovens after flammability testing. The images (a), (b), (c), and (d) likely represent different formulations or conditions, showcasing variations in the char structure, coherence, and coverage on the individual fibers. The scale bar indicates 150 µm."
Commentary: This figure likely shows the microscopic state of the non-woven fibers after being subjected to a flame test. The intumescent system, incorporating oxidized starch (OS), would have foamed and charred, and these images show that char on the fiber surfaces. The differences between (a), (b), (c), and (d) could be due to different ratios of PLA/OS, presence/absence or type of other flame retardant additives, or different processing conditions.
Note: This output was generated using only the document title and image. Our AI does not rely on figure captions or surrounding text, which are often vague or missing in patents.
Figure 7. Scanning electron microscope (SEM) images of (a) PLA/OS1/APP10, (b) PLA/OS3/APP10, (c) PLA/OS5/APP10 and (d) PLA/OS7/APP10 multifilament fibers (PLA = polylactic acid; OS = oxidized starch; APP = ammonium polyphosphate).
( 03 )
The Results
With Iris.ai, ArcelorMittal automated competitive patent analysis with 94% accuracy in just 4 minutes per document – down from 4 hours. The team can now process 30× more patents per month, freeing up researchers and product development owners to focus on strategy and innovation instead of data entry.
Key performance indicators
94%
Precision
Accuracy of extracted patent data across graphs, tables and text, reducing the need for manual validation
6Ox
Faster Extraction
Time to extract the data from one patent from 4 hours to 4 minutes
68
Languages
Supports patent review across global markets, including English, Chinese, German, and more
REQUEST A DEMO
( 04 )
What It Means for You and Your Team
Manual patent review is slow, repetitive, and resource-heavy. Iris.ai automates the extraction of complex data — from chemical compositions to experimental outcomes — across PDFs, Excel files, and both structured and unstructured databases. You get faster insights without the heavy manual workload or reengineering your data infrastructure, freeing your team to focus on innovation and strategy.
“At ArcelorMittal, we are constantly looking for ways to optimize our R&D processes and this is what brought us to Iris.ai — Integrating Axion into our process has made the ingestion and processing of external data significantly easier. It has cut weeks and, in some cases, months out of our research and development timelines and gives us the capacity to review more patents. Iris.ai proves there is enormous potential when research and development meets artificial intelligence, and we are only witnessing the beginning of that journey.”
Sophie Plaisant
Head of the Intellectual Property
ACCELERATE YOUR AI DEVELOPMENT
Connect now and let’s push the boundaries of innovation together
Request a Demo
( QUESTIONS & ANSWERS )
FAQ
How does Axion™ by Iris.ai extract data from patents?
Axion ™ automatically identifies domain-specific entities, locates and extracts relevant tables, and links
data between text and tables. The output is generated in a structured, machine-readable format tailored
to the organization’s needs. You can read more about it here.
How does the Iris.ai engine ensure accuracy?
The engine combines advanced NLP and entity linking across text and tables. It was bench marked for
ArcelorMittal with 94% extraction precision. The accuracy is estimated through a self-assessment
module, meaning the machine assesses its own extraction results and gives researchers a confidence
level and requirement for additional verification.
Can Axion™ extract data for multilingual patents?
Yes. Axion ™ supports up to 68 languages (incl. Chinese), enabling organizations like ArcelorMittal to
extract and analyze patent data globally.