Product descriptions:
Axpo Group, a major player in the Swiss energy sector, is on a mission to transform how energy is generated, traded and distributed. To accelerate innovation across their engineering organization, Axpo set out to implement knowledge-based search, modernize their API architecture and build an ML prototype to automatically label vendor and spending data. But legacy challenges — including data silos, fragmented tools and manual processes — slowed progress. With the Databricks Data Intelligence Platform, Axpo overcame these barriers, building GenAI solutions that boost the productivity of key power users by 30%, advancing their data and AI-driven approach to energy.
Confronting barriers to sharing critical knowledge
As the energy industry shifts toward decarbonization, utility providers face growing pressure to modernize, not just how energy is produced (e.g., hydro, solar and wind) but also how operations are managed. With over 7,000 employees across 40+ active markets in Europe, Asia and the U.S., Axpo Group set out to meet this challenge head-on by rethinking how knowledge, data and technology support the energy value chain.
In the spirit of continued innovation — particularly as GenAI and ML reshape business workflows — Axpo needed a faster, more efficient way for employees, especially engineers, to access critical internal knowledge such as technical standards, specifications and best practices. The legacy process relied heavily on manual keyword searches and frequent outreach to subject matter experts, leading to inefficiencies.
To prioritize future-facing initiatives like these, Axpo recognized the need for a reusable API architecture. “Without a scalable way to connect internal systems and data sources, we were forced to duplicate integration work, and this limited the speed and reusability of new GenAI applications,” Anaig Maréchal, Lead AI Engineering at Axpo Group, explained.
This effort is part of Axpo’s Generative AI Competence Center, a cross-functional team of 13 experts dedicated to pushing the boundaries of AI across the company and democratizing its adoption. Axpo saw opportunities to automate repetitive tasks and boost operational efficiency across the energy value chain. One initiative involves using machine learning to automatically classify vendor data and group suppliers into meaningful categories, improving procurement visibility and enabling smarter spend analysis. Another implements AI-driven document processing to extract insights from lengthy tender documentation files.
Despite their ambitions, Axpo faced several challenges in scaling AI. Critical knowledge was scattered across siloed systems, forcing employees to rely on point-of-contact experts and slowing down daily work. Disconnected tools and workflows hindered collaboration between data engineering and ML teams. Growing concerns around data residency and governance underscored the need for a centralized, compliant platform. That’s where Databricks came in.
Creating a shared knowledge layer for all teams
With Databricks, Axpo adopted the Mosaic AI framework to deploy GenAI solutions efficiently, enabling secure knowledge retrieval, unifying internal data and supporting seamless collaboration. Their architecture followed the medallion pattern: Raw data from sources like Confluence wikis entered the Bronze layer, was cleaned and chunked in the Silver layer and embedded into the Gold layer for downstream use. This approach ensured that clean, structured data flowed into GenAI applications.
Lakeflow Jobs orchestrated the pipeline, running daily to index new content, enrich it with metadata and keep the system up to date. Consolidating pipelines within Databricks gave teams a consistent environment, eliminating redundant workflows. At the core was an AI agent powered by retrieval augmented generation (RAG), the north star of Axpo’s internal knowledge base. Using Vector Search, Axpo indexed embedded Gold-layer content to enable fast, accurate retrieval of internal knowledge, enforced with row-level access controls to ensure proper permissions. OpenAI’s large language model GPT-4o generates conversational responses using the most relevant data. Axpo also integrated Mistral’s OCR model to extract searchable content from scanned PDFs and attachments, making the agent even more comprehensive.
Today, the team is advancing 12 active GenAI use cases — ranging from beta pilots to full-scale production — across all business areas. Databricks serves as the central pillar of this architecture and the primary RAG platform for all divisions represented within the Competence Center. “Now, we have employees — from engineers to project managers — able to access thousands of pages of technical knowledge without relying on keyword searches or subject matter experts,” Alberto Castillo Rodriguez, AI Engineer at Axpo Group, explained. Axpo deployed the agent using Mosaic AI Model Serving, exposing it as an endpoint that could be integrated into internal dashboards, chat tools and customer-facing portals that are fully integrated into their Axpo Insights product and used by numerous internal teams.
“This API-first approach allowed all of our teams to use the GenAI solution without duplicating integration work and made adoption seamless across departments, because employees interacted with the agent through familiar front ends,” Leiv Andresen, AI Engineer at Axpo Group, added. With unified data ingestion, secure retrieval and intelligent model orchestration, Axpo replaced a fragmented, manual information ecosystem with a centralized, intelligent knowledge layer — all powered by the Databricks Data Intelligence Platform.
Unlocking 30% more time for high-impact work
Axpo’s GenAI initiative is already delivering measurable value. Internal users report over 90% satisfaction with the knowledge assistant, citing faster answers, fewer interruptions and greater confidence in the information they access. Project managers and other power users estimate they’ve reclaimed up to 30% of their time.
The strong reception has fueled broader adoption of the Databricks Platform across the business. What began as an engineering-driven tool is now scaling company-wide, with Axpo seeing a 3x increase in users tapping into GenAI to unlock organizational knowledge. This evolution represents a fundamental shift in how the company collaborates, replacing siloed expertise and manual searches with shared intelligence that supports cross-functional alignment and removes operational bottlenecks.
Looking ahead, this approach will continue to support the broader mission of the Competence Center: integrating GenAI into Axpo’s products and processes to create an intuitive user experience, improve daily efficiency and drive future innovation. “The next step for us is making the agent useful beyond just search. We’re exploring how it can take action — like integrating with service systems — so it’s not just answering questions, it’s helping get work done,” Javier Fernandez de Alegria, Product Owner at Axpo Group, concluded. With Databricks supporting AI initiatives across the energy value chain, Axpo isn’t just improving internal productivity — they’re building the agility, intelligence and resilience needed to meet the energy challenges of tomorrow.