JBT Marel makes knowledge accessible with an AI Agent in Teams
How JBT Marel and Ciphix built an AI agent that makes documentation instantly accessible for field service engineers
A fragmented landscape of systems and documentation
JBT Marel is part of JBT Corporation, a global technology company active in the food and beverage industry. Years of mergers and acquisitions brought significant growth, but also increased complexity. Each business unit maintained its own systems, folder structures and methods for storing technical documentation, ranging from SharePoint and Confluence to local network drives.
For field service engineers, the technicians who carry out maintenance and repairs worldwide, it has become increasingly difficult to quickly locate the right manual or maintenance instruction.
The knowledge existed, but accessing it was the real challenge,” says Salvador Gimeno Marin, Global Field Service Director at JBT Marel.
This fragmentation became even more apparent after the merger with JBT. Engineers who previously worked only with JBT machines now also need to support Marel equipment. They had plenty of technical expertise, but not always direct access to the correct documents or versions. The organisation decided this needed to change: information had to be accessible within seconds, with a single question.
- Instantly accessible knowledge
The right documentation within seconds. - One reliable source
Always work with the most up-to-date version. - Smarter everyday work
Less searching, fewer errors, faster onboarding.
An adoption-focused and step-by-step approach
JBT Marel engaged Ciphix for its combination of integration expertise, AI capabilities and adoption-focused way of working. The project deliberately started small with a proof of concept, which delivered a working version in a short period of time.
“We did not want a large IT project that would deliver results only after a year. By building step by step, users could start immediately, and we could feed their feedback straight back into the next iteration,” says Francis van der Giessen, Software Engineer at Ciphix.
Because engineers already use many tools, both teams chose Microsoft Teams as their interface. The AI agent had to feel like a colleague you could simply ask a question, without extra software or login steps.
Meanwhile, Ciphix built a Power BI dashboard to provide insight into:
- The types of questions engineers ask
- where information is missing
- where confusion arises
- How the agent performs per source
These insights allow for targeted improvements and better prioritisation.
The TechDoc Finder: a single AI agent for all documentation
The result is the TechDoc Finder, an AI agent within Microsoft Teams that functions as an intelligent search assistant for technical documentation.
A field service engineer can simply ask:
- “How do I clean the MEYN 6500 line?”
- “Where can I find the user manual for machine X?”
The agent searches SharePoint, Confluence and other internal sources and presents the right documents within seconds, including direct links to the original source. This is essential, as engineers must always work with the correct and most up-to-date version.
“The AI supports you, but the documentation remains the single source of truth,” Francis emphasises.
Users can provide feedback on each answer. This is analysed in Power BI, giving the team insight into missing content, inconsistent naming or topics that require additional training. This way, the TechDoc Finder grows with the organisation.
The impact on engineers’ daily work
The first division went live in September 2025. Within a month, more than 120 users had already submitted over 1,000 questions. Adoption was high and feedback was positive.
“It feels like having a new colleague who knows all the documentation,” Salvador says. “You ask a question and immediately receive the right documents. It saves an enormous amount of time.”
Key impact areas include:
- Search time reduced from minutes to seconds
- New engineers onboard faster
- Fewer errors thanks to always using the correct versions
- Clear insight into missing documentation or training needs
The match rate of around 80 per cent for connected sources indicates that the agent aligns well with engineers’ daily needs.
JBT Marel also discovered additional value. By analysing frequently asked questions, the organisation now identifies knowledge and training gaps, which are used to create targeted learning pathways.
The next step: more sources and a broader rollout
Following the initial success, the TechDoc Finder will be expanded with new content sources, including Health and Safety documentation and legacy JBT technical knowledge. A broader rollout to more teams and regions will follow.
“We started with something small,” Francis concludes, “but it has grown into a scalable solution that is fundamentally changing how people find and use information.”
- More sources, more value
Health and Safety and legacy JBT documentation are being added for full coverage. - Broader organisational adoption
Rolled out step by step to more teams, departments and regions. - Continuous learning and improvement
Feedback and usage data help the agent become smarter and more relevant for everyone.
We will help you further!
Looking for a partner who understands your organization and translates this into smart solutions with results?
Discover
Define
Design
Do
