Skip to content
Food & beverage industry

Catch critical defects before the lot
ever leaves the plant.

Computer vision, automated HACCP traceability and shelf-life prediction for food & beverage SMEs. A single product recall costs north of €500k on average — we install the AI guardrails that prevent them. Typical accuracy on pilot lines: >98 % depending on defect class.

Spacious and modern food production plant interior with industrial machinery and processing line.
Photo : Adrien Olichon / Pexels
What hurts at the end of the line

5 blind spots that turn a non-conformity into a reputation crisis.

>€500k

Average cost of a product recall

Reverse logistics, stock destruction, crisis comms, customer audits — plus the long-term margin hit on retail shelves.

12–18%

False negatives in manual visual inspection

Fatigue, shift turnover and throughput: a tired operator misses up to 1 defect in 6 by end-of-shift.

30 min/lot

Paper-based or semi-digital HACCP

Non-productive QA time and fragile traceability when an IFS/BRC audit or food-safety inspection lands.

Too late

Supplier non-conformities caught downstream

Out-of-spec raw material lands at receiving, gets processed, and surfaces three days later at a customer.

8–14%

Forecast-driven food waste

Overstock, slow rotation, mis-judged shelf life: pure margin evaporating and growing ESG exposure.

Primary use case

Automated quality inspection via computer vision

An industrial camera over your line, a vision model trained on your real defects (foreign bodies, labelling, fill level, sealing, calibre, colour). Sub-80 ms decision, ejector triggered, lot-level archive.

  • Multi-class detection: foreign bodies, visual defects, labelling, expiry coding.
  • Continuous learning on your real rejects — the model improves lot after lot.
  • Built for wet, low-temperature, high-pressure wash environments.
  • Edge computing: no cloud dependency, runs in shop-floor blind spots.
−85% false negatives, typical payback 9–18 months (typical results on comparable projects)
02

Smart lot traceability (automated HACCP)

OCR + AI agent that consolidates CCPs, temperatures, supplier lots and lab results into an audit-ready dossier.

−70% QA data-entry time (typical)
03

Demand forecasting & inventory optimisation

ML model blending history, weather, promotions and seasonality to tune production batches and raw-material purchasing.

−25% stockouts, −18% dead stock (typical)
04

AI-driven shelf-life optimisation

Per-lot shelf-life prediction based on production, transport and storage conditions. Auto-routing between retail, foodservice and anti-waste channels.

−30% food waste (typical)
The metrics that matter

What you'll measure after 90 days.

Ranges observed on comparable industrial-vision deployments. Typical results — to be validated during the diagnostic.

ppm
Product non-conformities
vs. % today
0
Recalls per year
zero-recall target
% revenue
Cost of poor quality
internal + external
%
Automated HACCP coverage
digitised CCPs
%
Waste reduction
French AGEC anti-waste law
Our 4-step method

From first line audit to series production.

Week 1 — Strategy AI (from €5K)

3–5 day diagnostic

On-site visit, video capture, critical-defect inventory, HACCP alignment, ROI scoring and quantified roadmap.

Week 2–8 — Acceleration AI (from €15K)

Vision pilot in production

Model trained on your defects, software integration (camera + ejector), real-time dashboard, go-live on one line, side-by-side validation with your QA lead.

Optional — Industrialisation (on quote)

Multi-line, multi-site, IFS/BRC dossier

Hardening (washdown, vibration, lighting), ERP integration, multi-site roll-out, governance artefacts that ease your IFS / BRC audits.

Optional — Maintenance (monthly retainer)

MLOps & continuous supervision

Re-training on new SKUs, drift monitoring, incident log, QA-lead support.

Lot traceability · auditor view

From supplier to retailer in 8 timestamped nodes.

Every lot becomes a queryable graph: raw material, CCP, AI vision result, shipment, retailer. When an alert fires, the recall scope is narrowed in seconds rather than days.

SSupplierMaterial lotRReceivingInbound checkPProductionTemp CCPQAI vision>98% catchAI · CCPLLabMicrobioCPackagingLot + expiryEShippingPallet trackedGRetailerReceiving

Curious what this would look like on your line?

30 minutes, your process, our first quantified recommendations. No commitment.

What you walk away with

An augmented quality system, not a demo.

  • Proprietary vision model trained on your defects (model weights are yours)
  • Specification + software integration on your vision hardware (Cognex / Keyence / Basler) or via referenced integration partner
  • Real-time dashboard: defect rate, drift, per-lot alerts
  • ERP connector (Sage 100 / EBP / SAP S/4 / Sage X3 Process)
  • Governance artefacts that ease IFS / BRC audits (training set, metrics, human-in-the-loop, drift log)
  • Model governance (MLOps) plan + re-training protocol
  • Operator + QA lead training (1 day)
We integrate with your existing stack

No ERP replacement. Your tools stay in place.

Sage 100
EBP
SAP S/4HANA · Sage X3 Process
TraceOne / CSB-System
Cognex / Keyence / Basler cameras
Compliance & regulation

HACCP, IFS, BRC and the EU AI Act: an audit-ready dossier.

Regulation (EC) 178/2002 (food traceability, art. 18), Regulation (EC) 852/2004 (hygiene), mandatory HACCP, Implementing Regulation (EU) 931/2011, French AGEC anti-waste law. On the AI side: a vision system on a food line may qualify as 'high-risk' under the EU AI Act if it influences the release-to-market decision. We ship the governance artefacts auditors expect by default — training data, model performance, human-in-the-loop, drift log — that ease IFS / BRC audits and EU AI Act compliance. Certification itself stays with your accredited body.

See our security & compliance commitments
Client proof

Already running in production for our clients.

Delivery notes, health certificates, supplier records: the document-extraction building blocks we deployed at Guenifey apply directly to your flows. Automated reading, structured data, zero re-keying.

GuenifeyHereditary genealogy · ~200 employees

Guenifey automates heir-file processing and saves 30 minutes per case.

−30 min
per processed file
~2 h
recovered every week
0
manual entry errors
Read the case study
WHAT MAKES OLIXID

An AI agency built by the engineers who ship the code.

Not a consulting firm. Not a digital agency. A team of AI engineers who scope, code and ship to production — for French SMEs and mid-market companies.

Meet the team and the method
01

Engineer-founded

From EPITA, ranked #1 computer-science engineering school in France.

02

AI student consultancy

Founded in 2024. Dozens of engagements shipped to production before launching OLIXID.

03

Scope · build · deliver

Same people from diagnostic to production. No disconnected pre-sales, no project tunnel.

04

Plugs into what you run

No ERP replacement, no big bang. AI sits next to your tools — your teams keep their interfaces.

Frequently asked

What QA leads ask us first.

Next step

Secure your next lot before the IFS audit.

A free 30-minute discovery call to pinpoint where AI vision and automated traceability will move the needle on your line. To go further, the Strategy AI diagnostic (3–5 days, from €5K) quantifies the roadmap.

Reply within 24 working hours · NDA available