FSSC 22000 version 6: a smart way to record quality data and traceability with AI

What makes version 6 more demanding

FSSC 22000 version 6 is more than a textual update. Two elements demand demonstrably more than the previous version did. First, food safety culture: the scheme expects a specific plan for it, with attention to communication, training, feedback and measurability. Culture is thus no longer an abstract concept: at an audit, you must be able to show that you are working on it and following it up.

Second, traceability and the associated management of quality data and records. The chain must be traceable in both directions, and the data that supports this must be up to date, complete and retrievable. That is exactly where things get stuck in practice: the information often does exist, but is spread across paper, separate files and systems that are not linked. In a recall or an audit, pulling that data together then takes far longer than it should.

Where the manual work is now

Before you look at AI, it pays to take an honest look at where the record-keeping burden comes from. In most food companies it sits in a few regular places, and those places determine whether digitalisation and AI support deliver anything or merely add another system.

• Retyping: measurements and checks that are on paper or in a machine and are later entered into a system by hand, with a risk of errors.
• Searching during an audit or recall: data that is correct but scattered, so that compiling a file takes hours.
• Incomplete or inconsistent entry: fields that each employee or shift fills in differently, so that reports do not tally later.
• Signals noticed too late: a deviation that only becomes visible at the monthly review instead of at the time.

Where AI lightens the load of quality data

AI does not change the requirements of FSSC 22000, but it can make recording and retrieving the supporting data less labour-intensive. The gain rarely lies in anything spectacular, and almost always in removing repetitive record-keeping and in making deviations visible sooner.

• Structured recording: converting spoken or written notes from the shop floor into consistent fields, so that checks do not have to be retyped later.
• Making documents searchable: opening up specifications, supplier certificates and batch records so that, when a question comes up, you find the right document in seconds instead of searching through folders.
• Flagging deviations sooner: recognising patterns in measurement and inspection data that point to a trend, so that an employee can step in sooner.
• Speeding up audit preparation: bringing the requested evidence together into a draft file that a human checks and completes.

Traceability: from paper to traceable data

The core of traceability under version 6 is being able to trace everything: from raw material to finished product and back again, quickly and completely. AI can help here by linking fragmented records together and making them searchable, but the foundation remains the quality of the underlying data. A model running on incomplete or incorrect records gives a quick answer that happens to be wrong, and in a recall that is more dangerous than slow manual searching.

So start with the data, not the model: which links in the chain do you record now, where is the connection between batch, raw material and check, and where is it missing? Only when that foundation is in order does AI support add speed you can rely on. In addition, always ask a vendor to show how an answer can be traced back to the source record, because without that traceability a result is of little use to you at an audit.

Food safety culture remains human work

Food safety culture cannot be outsourced to software. AI can support parts of it, for example by tracking training attendance, helping to record deviations or making recurring themes in reports visible. But culture takes shape on the shop floor, in how managers respond to a report and how seriously suggestions for improvement are taken up. A dashboard is no substitute for that.

The sensible division of roles is therefore this: use AI to lighten the record-keeping and the overview work around culture, and leave the culture itself to people. That way, what you show at an audit remains a reflection of what actually happens, and not merely of what a system has recorded.

The human decides and signs

At AIBA, our starting point is that the human decides and signs. For quality and food safety, that means in practice: an AI system can prepare a record, flag a deviation or compile an audit file, but the quality manager reviews, corrects and signs. Responsibility under FSSC 22000 remains with the company and its people, regardless of which tool organised the data.

Article 4 of the EU AI Act (Regulation (EU) 2024/1689) has also applied since 2 February 2025. Since the amendment by Regulation (EU) 2026/1744, in force from 27 July 2026, it has asked organisations to take measures that support the AI literacy of employees who work with AI systems. They do not have to guarantee a specific level. For quality and production staff, what matters in practice is that they understand how a system arrives at a signal or summary, where the assumptions and limitations lie, and when a result must be checked or adjusted by hand. That is exactly the literacy an auditor may expect of you as soon as you use AI in the quality process.

Where to start in practice

Start small and make it measurable. Choose one record-keeping or traceability process that currently takes the most time or causes the most frustration, for example audit preparation or retrieving batch data, and first record how it works now: who does what, how long it takes, and where it goes wrong. That baseline measurement shows afterwards whether AI support really saves time there, so that you do not have to rely on a promise.

Be critical of savings claims from vendors. The effect varies widely by company, by product group and with the state of your data, and a general percentage says little about your situation. Ask for a trial with your own records. Our free Practice Scan helps you choose that first process and sharpen the baseline measurement.