Case study · Food industry
AI came in through Quality. Today it operates across four areas.
The management of Lácteas García Baquero asked us to bring AI to the company's processes. Instead of a two-year plan, we started with the department carrying the heaviest documentation load, Quality, and in 90 days we shipped nine use cases to production, now spanning Quality, Export, Plant and Purchasing. It is our first major enterprise program, and it continues department by department.
First one department with measurable results. Then the whole company.
A small quality team, with the documentation scattered across the company.
Supplier certificates in network folders, technical sheets in Excel, audits in scattered PDFs, every customer reply written from scratch. Preparing an audit could take days of manual searching. The challenge wasn't “add AI”: it was governing years of documentation, wiring it into the real workflow, and doing it without AI making a single regulated decision on its own.
- 01Years of documentation in network folders, no reliable metadata, no single source.
- 02Days of manual prep for every food-safety audit.
- 03Critical knowledge tied to the person who knew where to look.
Nine use cases, four areas.
A real AI program in production. Exact figures and detailed methods are omitted for confidentiality.
Quality & compliance
Supplier assistant
Ask in natural language about any supplier —certificates, allergens, materials, validity— and get the answer with the source document cited. One search point for what used to be scattered.
Expiry monitoring
Every morning, the email with expired or soon-to-expire certificates, filtered and actionable. The end of manual tracking across hundreds of documents.
Document governance
Years of documentation migrated from network folders to a single repository, with structured metadata verified document by document. The single source of truth.
Audit preparation
The assistant cross-checks your Quality System against the IFS Food v8 standard and returns, point by point, whether each is Covered, Partial or Not covered, with the evidence cited. Days of prep in minutes.
Export & customers
Technical sheets & forms
Customer queries and recurring retailer forms answered from hundreds of indexed technical sheets, with the draft ready to validate. The inbox bottleneck, solved.
Multi-market labelling
Labelling questions per export market resolved citing the regulation, article and date. Every country has its rules; the assistant knows them.
Operations & plant
Plant bots
Operators check procedures, log daily reports and send photos from their phones. A vision model classifies product appearance, and the plant's process controls are one message away. Everything traced.
Traceability & complaints
Consumer complaints
The assistant reads the incoming complaint and proposes the reply letter; the team accepts, edits or rejects. With personal data handled under GDPR.
Batch traceability
Connected to the company's systems, it resolves traceability for thousands of batches in seconds, without leaving the chat.
See it working.
Forty seconds: each department's paperwork, solved in production and with the team always deciding. All on the Microsoft 365 the company already had, without a single data point leaving the EU.
Anonymized illustration. Specific figures and methods are omitted for confidentiality.
AI proposes. The team decides. Always.
In a regulated sector this isn't a slogan, it's the design. No regulated decision was made by AI alone, every answer cites its source, and the EU AI Act compliance framework was documented from day one.
Human oversight
Zero automatic regulated decisions. Every critical output goes through the quality team.
EU AI Act
Training (art. 4), transparency (art. 50) and governance committee, documented and ready.
Full traceability
Every agent answer cites the source document. No claims without evidence.
Is your company next?
Tell us which process is eating your team and we'll tell you how we'd solve it, with this case as reference.
664 24 13 28