Nearly Half of Food and Beverage Employees Are Using AI Their Leaders Didn’t Approve

Every generation of food and beverage leaders reaches a moment where it must decide what to trust. A generation ago, trust relied on gut feel. Businesses were small enough for one person to just walk the floor and see everything that was happening. As operations grew, trust moved to systems, reports, and dashboards. The business had outgrown what any single person could hold in their head.

Today those decisions are harder than ever. And in many organisations, employees are already making some of them for you.

In Aptean’s 2026 Artificial Intelligence Research, conducted by independent research firm Vanson Bourne across 300 food and beverage decision- makers, 44% admitted to using AI tools without formal approval, which was above the cross-industry average. Adoption of AI is happening. It just isn’t happening where leadership expects.

Your Employees Aren’t Waiting for Permission

Shadow AI (the use of unapproved tools) is now common in planning offices and on plant floors. The motive is almost always practical. People reach for whatever helps them work faster, cut manual effort, or make a call with more confidence.

Here’s why that matters more in food and beverage than almost anywhere else. The sector runs on sensitive, high-stakes data: supplier records, batch and lot information, formulations, allergen segregation rules, and traceability records. When any of that finds its way into a public AI tool, even by accident, you’re risking confidentiality breaches, compliance gaps, and outputs nobody can stand behind in an audit.

Accuracy is the second exposure.

General-purpose AI tools have no model of how a food plant runs. They have no concept of a changeover, a catchweight, or a lot code, so they fill the gap with assumptions and present them fluently.

Picture a planner at a bakery running 40 SKUs with two allergen changeovers a shift, asking a public chatbot to help sequence next week’s production. The answer is structured, confident, but quite possibly incorrect in ways only someone who knows the allergen matrix would catch. That’s the real exposure - an AI tool working well outside its competence.

The Gap Is Between Wanting AI and Redesigning Work

Appetite is not in short supply. Among the food and beverage organisations surveyed, 84% say they are motivated by what AI makes possible rather than by pressure from competitors. Leaders can see a way to cut waste, improve on-time in-full (OTIF) performance, strengthen compliance efforts, and protect margins as operations grow more complex.

Some of that is landing: food and beverage organisations reported results above the cross-industry average on operational efficiency, on-time delivery, and forecast accuracy. Almost a quarter (23%) now treat AI as essential to how they work and make decisions.

Underneath that progress sits a widening gap. More than one in five food and beverage organisations (22%) have either no AI strategy at all or are still exploring and piloting, the highest of any industry surveyed. That sits awkwardly against the 83% of food and beverage leaders who expect their business to fall behind quickly without AI.

So, the gap isn’t really about appetite. It’s about whether businesses have redesigned a core workflow yet.

Part of the problem is simply recognition. Plenty of organisations already run AI inside graders, sorting machines, planning tools, and quality systems, and think of it as automation. Others are still working out what it means for their business, which makes governing AI close to impossible.

Industry-Specific AI Outperforms General-Purpose Tools

One of the clearest findings in Aptean's research is that AI built for the industry delivers more than AI built for everyone.

Today, 63% of food and beverage organisations use general-purpose AI and 46% have implemented AI designed for their industry; many run both. The performance gap between the two is wide. Organisations using industry-specific AI were more likely to report forecast accuracy improving by 10% or more over the past year (29% against 19% for general-purpose tools). They were also more likely to name stronger competitive positioning as their biggest gain (39% against 32%), and improved workforce morale (39% against 33%).

Anyone who has spent time in a plant will recognise why. Shelf-life constraints, lot-level traceability, batch records, allergen segregation, production scheduling, and cold-chain dependencies all shape how a decision must be made. Perishability changes everything. Inventory doesn’t just cost you money, it expires, which makes first expiry, first out (FEFO) a constraint rather than a preference. Yield variance eats margin in ways that compound across high-volume production. And traceability runs in both directions, within hours.

A general-purpose model has no way to account for any of that. A model built for the industry starts from it.

Budget is not always following that logic. When food and beverage organisations secure AI funding, 54% say strategic alignment is not an influential factor in the decision. The money is moving and some of it is moving towards work that will never shift an operational number.

Data and Governance Come First

The research is unequivocal about what’s holding organisations back. 81% name data quality and access as their greatest challenge, 78% cite a lack of governance, and 85% say connecting AI to their existing business systems is harder than getting the AI to work.

Most food and beverage organisations only find the gaps in their enterprise resource planning (ERP) or supply chain data when they start connecting AI to it. AI is simply the first thing to read the data closely enough to show them.

There’s an advantage hiding in that reality for food manufacturers. Batch records, lot genealogy, date codes, allergen declarations, and QA hold histories - data you already maintain because an auditor might ask for it - are exactly the clean, governed inputs an AI system needs. They're complete, because gaps become findings. They’re timestamped and attributable, because traceability demands both. It’s the most underused asset in food manufacturing.

Most food and beverage businesses are therefore closer to ready than they think. They have spent years building a high-quality evidence trail and only ever used it to prove what happened after the event. The same records will support a forecast, a schedule, a quality prediction or an exception alert. Nothing new needs to be collected. The work is connecting what is already there.

North America Is Pulling Ahead

The research shows a clear regional split.

Among food and beverage organisations already using or implementing AI, 57% in North America have adopted industry-specific tools, 38% in Europe. On custom-built AI the gap narrows: 50% against 42%. The outcomes follow the same pattern. Nearly half of North American organisations report faster product delivery (48%, against 31%) and improved revenue growth (45%, against 33%). Europe here means the UK, France, Germany, and the Netherlands.

North American food and beverage organisations are more likely to use industry-specific AI, more likely to build custom AI, and more likely to report faster product delivery and improved revenue growth.

For European manufacturers, that's a commercial question with a clock on it. The pressures are the same on both sides of the Atlantic with thinner margins, more SKUs, and constant audit exposure, but North America is further ahead in building the data foundations and governance that let AI act on them. Closing the distance is a matter of sequence more than spend: the right layer first.

The Five Practices That Separate Real Programmes From Pilots

1.    Use AI built for food and beverage. Generic tools don’t recognise lot codes or catchweights, and they can’t be trusted with a decision that touches allergen segregation.

2.    Start with the data you already keep for audits. Batch records, lot traceability, and date codes are the raw material. Clean and connect them before you introduce anything new.

3.    Tie every AI project to an operational number. Giveaway, OTIF, forecast error, QA hold time. If a project moves none of them, it’s a pilot, not a programme.

4.    Govern what’s already happening. Map where your teams use AI today. You can’t set direction for something you’ve never measured.

5.    Aim at decisions, not dashboards. Another screen showing yesterday’s problem adds nothing. Value comes from surfacing an exception early enough for someone to act on it.

The Moment of Trust Has Arrived

The organisations getting the most from AI are the ones combining industry-specific tools with strong data foundations and a clear business objective.

Trust once meant the plant manager’s eye, and then it meant the dashboard. The next version will be earned by organisations that can say exactly what their AI is working from, who approved it, and which decision it made better. That work starts now.

The full study – including the industry and regional breakdowns behind every figure here – is available at https://www.aptean.com/en-GB/resources/industry-insights/report/food-and-beverage-ai-report?utm_source=article&utm_medium=web&utm_campaign=fnb_ai_report_2026

Bron: Aptean - partner FoodPro Network