Data & AI: why both are essential to the future of food security

Foods Connected's Head of AI Enablement, Dr Stephanie Brooks, explores why effective AI needs good data – and how together they are the key to future food security. 

Let me start with a question I put to a room full of food industry professionals at Queen's University Belfast recently: What is the single biggest data gap that stops your organisation from making better decisions about supply chain risk?

That silence tells you everything about where the food industry is right now. We are operating in one of the most volatile periods in living memory, and yet the systems we rely on to manage that volatility are, in many organisations, still built on spreadsheets, phone calls, and data that's days or weeks out of date. Something has to change.

The world has changed. Our data systems haven't.

The pressures bearing down on food supply chains are not abstract or distant. They are live, active and accelerating.

Extreme weather events – drought, flooding, unseasonal frosts – are disrupting harvests and ingredient availability with a frequency that historical data simply cannot predict. Supply plans built on last year's patterns are already unreliable. Add to this the persistent shock of geopolitical instability: conflict and trade policy shifts have ripped through commodity markets, from sunflower oil to cereals, creating sudden supply gaps that even the most experienced procurement teams couldn't have anticipated.

At the same time, the regulatory environment is tightening. EUDR, Scope 3 emissions reporting, CSRD, food safety legislation – these aren't future concerns. They're demanding deeper traceability and data transparency right now, from every tier of the supply chain. The compliance risk is real.

"Conflict and trade policy shifts have ripped through commodity markets... creating sudden supply gaps that even the most experienced procurement teams couldn't have anticipated.

And threading through all of these challenges is a single, consistent problem: data fragmentation. Most food businesses are making critical decisions without a complete, current picture of their supply chain. The inability to see, act and adapt in time is not a strategy failure. It's a data failure.

"Most food businesses are making critical decisions without a complete, current picture of their supply chain. The inability to see, act and adapt in time is not a strategy failure. It's a data failure.

Most food businesses are flying blind

Here's what the status quo typically looks like. Supplier risk is spotted after the disruption has already hit. Sustainability data is collected manually, making it incomplete and inconsistent. There is no single view of supply chain health across all tiers. And perhaps most worryingly – AI projects are being launched on top of data foundations that simply aren't reliable enough to support them.

This is the uncomfortable truth: you can't build good AI on bad data. Artificial intelligence is only as intelligent as the information it learns from. If your data is siloed, lagged or incomplete, your AI will confidently tell you the wrong thing.

What good looks like is the opposite of all this. Live data flows from suppliers, with exceptions surfacing automatically. Early warning of risk before it becomes disruption. Sustainability data embedded directly into procurement workflows. One platform connecting all supply chain stakeholders. And AI layered – carefully, deliberately – onto clean, connected, trusted data.

"This is the uncomfortable truth: you can't build good AI on bad data.

The gap between where most organisations are and where they need to be is significant. But it is absolutely closeable.

How AI actually adds value in food supply chains

When the data foundations are right, AI stops being a buzzword and starts being genuinely useful. Let me walk you through what that looks like in practice.

The inputs are the things your supply chain already generates – or should be generating: supplier performance feeds, weather and commodity signals, logistics and inventory data, regulatory and compliance updates. These are not exotic data sources. They exist. The challenge is connecting them.

Layered on top of that connected data, AI can do things that humans simply cannot do at scale or speed. It can detect anomalies – a supplier's delivery reliability dropping in a pattern consistent with operational stress, for example, before anyone has flagged it. It can score and cluster risk across your entire supplier base simultaneously. It can model predictive lead times, accounting for disruption signals that your team hasn't yet noticed. And it can translate all of that complexity into natural language summaries that a procurement manager can actually act on.

The outcomes are what matter: earlier warnings of disruption, proactive supplier conversations before problems escalate, more confident sourcing decisions, and auditable compliance evidence that satisfies regulators and investors alike.

This is not science fiction. It is available now – for organisations that have done the work to get their data in order.

"This is not science fiction. It is available now – for organisations that have done the work to get their data in order.

The technology isn't the hard part

This is perhaps the most important thing I want you to take away from this.

Organisations that thrive with AI are not the ones with the most sophisticated models. They are the ones with the clearest data, the most aligned teams, and the most honest view of where they actually are. The technology is, relatively speaking, the easy part. The hard parts are human.

First, data foundations. Before AI can help you, your data must be connected and trusted. That means auditing your current data quality and completeness to understand what you're working with. It means identifying the gaps: which signals are missing entirely, and which are still being collected manually? And it means investing in data ingestion and connectivity before you start building models. Too many organisations skip this step, and the result is always the same: "We have data but we don't trust it."

Second, people and skills. AI strategy must be business-led, not technology-led. The most unsuccessful AI implementations I've encountered are the ones where a technical team built something impressive that no one in the business actually uses or needs, or where the Return on Investment is either negligible or hasn’t been considered at all .  AI literacy needs to be built across your commercial, quality, procurement, and operations teams — not just in IT. Subject matter experts need to be embedded in the build process, because they're the ones who know what "good" looks like in practice. AI use cases need to be owned by the business. "We bought a tool but nobody uses it" is one of the most common — and most avoidable – outcomes in enterprise AI.

"Organisations that thrive with AI are not the ones with the most sophisticated models. They are the ones with the clearest data, the most aligned teams, and the most honest view of where they actually are.

Third, process integration. AI that sits outside your workflows doesn't change your decisions. It generates reports that get filed and forgotten. For AI to drive real change, its outputs need to be embedded in the daily operating rhythms of your teams. You need to define clearly who acts on which signals, and when. And you need to build feedback loops so that your AI can learn, improve, and remain relevant over time. The failure mode here is familiar: "It gives great insights but nothing changes."

ESG is not a reporting problem. It's a data problem.

I want to address sustainability specifically, because I think it's often misunderstood.

Most organisations can articulate their sustainability commitments. Fewer can evidence them with data. The gap between ambition and proof is almost always a supply chain data challenge – not a strategy one.

In its rawest sense, EUDR requires commodity traceability to plot level. Many organisations struggle to get visibility past their tier 1 supplier, let alone down to farm level. Scope 3 emissions data is fragmented, inconsistently measured, and often estimated rather than evidenced – which is becoming increasingly untenable for auditors and investors. CSRD requires mandatory supply chain disclosures across social, environmental and governance criteria, all of which depend on supplier-level data that typically lives in spreadsheets, if it exists at all.

And then there's traceability for food safety and provenance. Consumer and regulatory expectations are rising. End-to-end traceability requires data connectivity across every single handoff – from farm inputs through processing, logistics, and all the way to shelf. That connectivity doesn't happen by accident. It has to be built.

Sustainability is a data mandate. Traceability is now table stakes.

Where to start: three practical steps

If all of this feels overwhelming, let me offer a more grounded way in.

Audit before you build. Map your data touchpoints. Understand what you collect, how often, and how reliably. Identify the decisions in your business that still depend on a phone call or a manually updated spreadsheet, rather than live data. Prioritise the highest-risk gaps — you don't need to fix everything at once. And set a data quality baseline so you can measure improvement over time.

"Audit before you build. Map your data touchpoints. Understand what you collect, how often, and how reliably.

Start with a signal, not a system. One use case, done well, builds more trust than a platform nobody uses. Pick the single most painful supply chain blind spot your team faces today. Find or build the data feed that addresses it — keep the scope tight. Surface the output in the workflow where the decision actually gets made. Then measure and share the result. That proof point is how you build the internal mandate to go further.

Collaborate to scale. No single company has all the data it needs for true supply chain resilience. Consider sharing non-competitive signals with trusted supply chain partners. Engage with industry platforms working on pre-competitive data standards. Use forums and peer networks to benchmark and learn — you are almost certainly not alone in the gaps you're facing. And build the case internally: resilience is a board-level strategy, and the data infrastructure that enables it deserves board-level investment.

The question worth sitting with

I'll close where I began. What is the single biggest data gap that stops your organisation from making better decisions about supply chain risk?

If you can answer that question clearly – and honestly – you already know where to start.

Visibility is the foundation. Without connected data, AI is guesswork. Readiness is organisational, not just technical – people and process must come first. And sustainability is now a data mandate: the gap between ambition and evidence is a supply chain problem that only better data can solve.

"Visibility is the foundation. Without connected data, AI is guesswork.

The food system is too important, and too fragile right now, to navigate on incomplete information. The good news is that the tools, the platforms, and the knowledge to do this differently already exist. The question is whether your organisation is ready to use them.