Our manufacturing expert's view on why data and AI keep failing food manufacturers, and what actually needs to change.
Ask Jacob Willems where the pressure is, and he'll give you a number.
~ 2,32%
The average operating margin erosion in food manufacturing since 2019.*
He's spent years inside this industry. Sat in the rooms where decisions get made. Watched pilots launch with fanfare and then disappear not much later. Seen dashboards built that nobody acted on. And he's arrived at a conclusion that cuts against the grain of most AI investment logic: the problem isn't the data. And it isn't the technology. It's what happens after the insight.
The Sector is Not Short of Ideas. It's Short of Impact.
To understand why this matters, you need to understand what this industry actually is.
3,160
Food Companies in Flanders, 95% SME
70,000
Direct jobs in the sector
€67.1B
Turnover, the largest industrial sector in Flanders
This is not a niche sector. It sustains farming communities across West-Flanders and Limburg, anchors regional economies, and keeps Belgium among the EU's top four food exporters. And since 2019, it has been losing ground.
Operating margins have dropped from ~7% to ~5% since COVID. The causes are structural, not cyclical:
• Raw material cost increases may be difficult to pass on to the customer
• Wages rising faster than productivity gains
• Energy prices that have swallowed efficiency improvements
• European compliance obligations: CSRD, CBAM, packaging regulation, behaving like mandatory costs with no revenue upside
Then came the external shocks. Supply chain disruptions after COVID. Energy price spikes after 2022. Geopolitical instability affecting commodity flows from Ukraine and beyond. Every time the sector absorbed one blow, another followed.
From 2020 onwards, food manufacturing went all-in on data and AI. Global agrifoodtech investment doubled in the post-COVID years, peaking at $33 billion annually. The potential was genuine.
And yet, by 2025, EBIT margins had fallen by nearly a third compared to 2019. The investment went in. The impact did not come out.
Jacob calls this the AI contradiction, and the reason it exists is more layered than it first appears. Food manufacturing is not data-poor. Most companies sit on mountains of it: production logs, sensor data, planning systems, ERP records, quality reports accumulated over decades. Some of it is clean and structured. Much of it is still scattered across systems.
The data exists, unevenly and imperfectly, but it exists. The models get built. The dashboards go live. Early results look promising. Leadership signs off on a rollout. Then the gap opens.
A planner ignores the recommendation because they don't trust the source. A supervisor identifies a predicted quality risk but lacks authority to stop the line. A maintenance alert arrives at the wrong moment, the planning window has already closed. A pricing signal comes through, but the approval process takes 3 days, and the margin window was open for 1 day.
The Pattern: the insight reached the dashboard. It never reached the floor.
Losing Faith is Expensive.
There's a cost to failed AI initiatives that rarely shows up on a balance sheet. Stanford's AI Index estimates corporate AI investment reached $252 billion globally in 2024. In agritech specifically, it surged to $33 billion post—COVID—then fell back to $15.6 billion in 2023, in large part because initiatives failed to translate into returns.
A pilot runs successfully in one plant. It doesn't scale. The next team is more sceptical. The sponsor is harder to find. People who changed their routines but saw no change in their KPIs are slower to engage the third time.
In a sector running on low margins, that credibility crash is one of the most expensive things that can happen because it slows the one type of improvement that can compound.
What Needs to Change
Most consultancies respond the way the industry always has: broader strategy, longer roadmaps, more use cases. Start from AI ambition. Build a transformation program and wait for adoption.
At Madison Partners, we start from the opposite end. Every food manufacturer already feels where the pressure is: a yield issue that won't resolve, an energy cost that keeps climbing, an OEE problem everyone acknowledges and nobody owns. Few can prove it precisely, quantify it across sites, or agree on it across the table.
Our Approach
We go into the data, find exactly where value is leaking, and turn that feeling into a fact. A number, a cause, and a clear next step. In six weeks.
Our diagnostic sprint is built around one margin-relevant business problem, not a use case catalogue or a transformation roadmap. We use the data the company already has. We define the decision owner, the decision moment, and what needs to change in the daily workflow for the output to reach the floor.
The result isn't a report that gets filed. It's a quantified value case and a practical path to action. The manufacturers who get this right don't just recover lost margin. They build a machine that keeps finding more.
That's the difference between a pilot and a profit engine. And right now, with margins where they are and the pressure still building, the window for getting this right is narrowing.
Jacob Willems | Food & agribusiness expert, Madison Partners
If this resonates, if you already know where the pressure is but can't yet prove it, a 30-minute conversation is enough to know whether there's something worth finding in your data.
*A flat 2,32% does not hold for all. Some larger manufacturers have managed to sustain margins by increasing their pricing power. But all of them are focusing on cost management.
