–––––– EXECUTIVE SUMMARY ––––––
Domain expertise and AI combined
Public sector tenders are high-stakes, high-volume, and relentlessly time-pressured. For Bechtle, one of Europe’s largest IT service providers, they’re also a core part of the business, accounting for roughly 40% of group revenue. So, when tender managers were spending hours manually extracting information from 600-page documents, the business case for a better approach was clear.
PLANET AI built a purpose-specific solution that automates the extraction work and presents every result with a direct link to its source. A conversational interface lets teams ask their own questions against the documents and tap into years of archived tender knowledge. The outcome: specialists spend their time on win probability, risk assessment, and bid strategy, not on data entry.
“PLANET AI’s solution has made an enormous difference to our daily work. It handles the repetitive tasks and frees us up for the decisions that matter. The tailored data extraction combined with an interactive chat interface means we can work through even the most complex tenders faster and more systematically than before.”
Dr. Alexander Fink, Head of Public Sector Innovations, Bechtle AG
–––––– THE CHALLENGES ––––––
Complexity meets time pressure
Bechtle operates across 14 European countries and generates roughly 40% of its revenue from public sector clients.
The initial assessment of incoming tenders is where the pressure is highest:
Volume and diversity
600+ pages per tender in mixed document formats
Go/no-go under time pressure
Specialists spending hours on data extraction instead of strategic assessment
Critical requirements
Manual extraction where a single error means exclusion
Fragmented knowledge base
Experience from previous projects hard to access
–––––– THE SOLUTION ––––––
A purpose-built AI assistant for tender management
Working closely with Bechtle’s tender management team, PLANET AI developed a solution that goes well beyond generic tools and integrates directly into existing workflows. At its core are three functions.
Intelligent data extraction
Automatic identification of standard information with over 90% accuracy
„Talk to Tender“
Chat interface for individual questions against tender documents with source references
„Talk to Archive“
Access to historical tenders and accumulated experience
Intelligent data extraction
Users upload tender documents directly into the interface. A workflow built on Large Language Models (LLMs), Retrieval-augmented Generation (RAG), and specialized logic components analyzes the documents and identifies predefined information automatically. Every result is presented not as a given fact, but as a suggestion with a direct source reference. An integrated review function lets users preview and verify each value individually.
Automatic checklist with review feature
“Talk to Tender” & “Talk to Archive”
A chat interface lets users ask their own questions directly against the tender files, backed by the JAIDE architecture with purpose-built LLM agents and RAG tools that go well beyond a generic setup. Every response is backed by the relevant source passages, including the originating file, which is especially valuable when a colleague picks up a tender later in the process. “Talk to Archive” extends this to previously processed tenders, turning years of accumulated institutional knowledge into an active resource for new projects
Conversational document analysis with source references
–––––– RESULTS ––––––
Measurable efficiency, greater process reliability
Tender managers at Bechtle now have a tool that doesn’t just make their work easier — it changes how they work. Instead of working through documents to extract standard information, they can focus on what actually requires their expertise: assessing win probability, identifying risks, and developing a compelling bid strategy. A real asset to the daily work of the Bechtle teams.
