Complete, actionable product intelligence – powered by AI automation

The challenge: Progress with automation, but 80% of data still out of reach

Our client operates in an industry where competitor manuals and product data are constantly changing. Staying current means tracking information across hundreds of models, always using public, legally approved competitor documents in line with company compliance. Manual review limited their intelligence team to analyzing only about 20% of the market.

To ease the burden, our client had built an internal pilot AI assistant designed to answer questions about competitor products. While this pilot helped with a few models, every document still required manual upload and setup. As a result, most competitor information remained out of reach and the team lost valuable time to repetitive searching and compliance-driven fact-checking.

Strategic goal: Automate extraction and analysis for every approved competitor manual – so our client could base decisions on complete, up-to-date intelligence, stay compliant, and scale without adding manual workload or extra headcount.

Our role: From POC to full-scale automation

The client turned to Proekspert to bridge the gap between a working concept and a dependable, scalable platform. Our job was to deliver automation, compliance, and scalability for real operational impact. This meant tight coordination between our cloud, data engineering, AI, and digital services teams. 

Here’s what we delivered: 

• Automated the end-to-end processing pipeline, so every competitor manual uploaded to SharePoint is automatically parsed and made available via Databricks – no manual intervention. 

• Integrated Microsoft Power Automate, Databricks, and cloud AI models to keep document flow seamless, error-free, and always up to date. 

• Developed advanced extractors for complex tables and technical data, unlocking information that used to be inaccessible. 

• Built deduplication logic to skip unchanged manuals and cut cloud costs. 

• Embedded full traceability – every AI answer links directly to its source and includes image snippets for rapid verification. 

• Used a metadata-driven architecture so new document types or sources can be added quickly, without code changes. 

We’re especially proud of delivering a system that processes every approved competitor manual automatically – without any manual steps remaining. Despite data complexity, extraction accuracy improved by 20%. The platform is robust, fully traceable, and easy to extend for future needs. 

The outcome: Instant, accurate answers across all competitor manuals

Previously, our client’s teams spent days manually combing through competitor PDFs, often missing most of the market. Now, every approved manual is automatically processed and searchable through the AI agent. Answers arrive in seconds, every datapoint is instantly traceable to its original source, and intelligence remains current as new documents are added. With higher extraction accuracy, teams work faster with more reliable information.

Impact for the client’s organization

hese changes arm our client to move quicker, win more business, and operate efficiently as the market expands.

• Full market visibility: our client now has a complete competitor view, enabling faster, better-informed decisions in product strategy, sales, and R&D.
• Manual research slashed: time spent on competitor document analysis has dropped from days to minutes – reducing cost, speeding up proposals, and freeing experts for high-impact work.
• Compliance and scale by design: every answer is documented and auditable, reducing regulatory risk and making it easy to adapt the system for new markets as the business grows.

Technologies

Python 3.12, Databricks, PySpark, Microsoft Power Automate, SharePoint, Azure OpenAI, Google Gemini