AI is expected to improve food manufacturers’ productivity and efficiency but as with the development of any new technology there is an element of apprehension.

Much of the anxiety seemingly comes down to a lack of understanding of the technology itself, how it works and the myriad of AI systems that could be employed for different applications and problem-solving. And then there’s the question of labour, tech-skilled labour rather than the manual kind.

Think about the evolution of phone technology – moving from analogue to digital devices and hefty back-of-the car handsets to those that now fit in a pocket – electric vehicles and their slow development due to a lack of infrastructure and the introduction of hybrid models as a half-way-house to address bottlenecks.

Food manufacturers are adopting AI but reservations might come down to investing too much too quickly when the tech is advancing at pace, capital on hand and assessing the return on investment.

Abhinav Agrawal, the co-lead of AI and data monetisation at consultancy AlixPartners, argues food manufacturers also face a bottleneck in that they “are not exactly the desired location for data scientists, AI programmers and developers” to attract the desired talent pool.

“I don’t recommend our clients make big bets right now,” Agrawal tells Just Food. “One of our clients launched a $55-60m AI initiative about two years ago but there’s no measurable ROI right now. There’s no success that we can point to that $60m was worth investing.

“Instead, launch $5m projects and see. Maybe two of them fail but three of them produce a ROI and you gain a little more experience in which type of skills you need to hire.

“Then you launch another five, or maybe next time you can launch ten, so in the end you may end up spending $60m over three years, for instance.”

“Strategic advantage”

It’s a bit of a juggling act for food manufacturers. There’s almost an obligation to invest in AI now or risk being left behind and losing a competitive edge.

Or as Tom Clayton, the CEO of Sheffield-based IntelliAM, puts it, gaining a “strategic advantage”.

Clayton, the head of the tech company specialising in AI software and machine learning for manufacturers, believes productivity improvements through the tech are a way to address the increasing demand for food as global populations rise.

“In the food and beverage world, where you’re talking about high volumes and low margins, it’s really interesting because, if you can make an extra 100,000 SKUs a day, that’s direct profit to the business. Similarly, if you can avoid wastage, you’re moving costs from the bottom line.

“We’re talking considerable amounts of money. Essentially, the driver of these organisations is to be able to improve capacity of the plants by being more intelligent and doing more with less.”

AI not yet “transformational” to profits

There’s an obvious buzz around AI, not just in the food industry but outside as the benefits (or otherwise) of the technology are debated across media.

Clayton says there’s been an “immense increase” in interest over the last two years as boardrooms wake up to the fact the tech might give them an advantage.

“In any new technology convergence, there’s always people who want to embrace it immediately. As your competitors do it, you’ve got to do it. Unfortunately, I think there’s going to be a bit of first-mover advantage over the AI element,” he adds.

Agrawal presents an example of how AI can work its magic in improving productivity and efficiency in terms of a salty snacks manufacturer for which AlixPartners acted as an adviser – assessing the size and shape of corn used in the production process to ensure quality and consistency.

The production line’s operating efficiency rate went up from the 80s percentage range into the 90s from employing an AI system, he says, adding each percentage point increase “translated into a couple of million dollars of bottom-line impact”.

However, Agrawal suggests the tech is not yet “transformational” when it equates to output and profitability.

“For me, transformative means you’re going after 30-40-50% plant P&L improvement, that type of thing. I would not say it is possible right now. It is in maybe the 20s with these technologies,” he argues.

“But I do think that in a year or so, the technology can improve, that it can produce like an 30-40-50% improvement.”

Labour leverage

Nevertheless, AI can translate into savings on the labour front, Agrawal says, proposing “what was not possible even five years ago is possible now”.

A bakery manufacturer in the US needed a lot of temporary, hourly paid staff to help run its 30 or so production plants, with managers having to assess shift arrangements and working patterns on a daily and weekly basis, he says.

“Our algorithms recommended how much labour they needed by the day, by zone and by type, and whatever. We tracked it for six months and those predictions saved them on average 20%,” Agrawal explains.

For those food and CPG manufacturers already “leveraging” AI tech, Chris Ashley, the vice president of strategy at Peak, a unit of software company UiPath, says the “advantages compound over time”.

He believes the “more sophisticated they become, the more they can leverage these systems”.

Ashley adds: “They will accrue more working capital efficiency, more operating margin, more productivity and that will just continue to compound over time. We think this needs to be a topic of conversation in every boardroom currently on how to educate the teams.

“I think the impacts can be transformational. The augmentation of these teams with AI-driven systems is driving more productivity and driving more economic upside for the business, which means more jobs.”

Level playing field

Meanwhile, Clayton at IntelliAM says AI helps improve the reliability and life of production components by identifying problems before they arise, reducing maintenance times and the potential for lines going out of action.

AI does that by “ingesting millions of data points from the brains of the machines in a factory and contextualising” them to “understand tolerance settings to get maximum throughput speed settings, reduce bottlenecks, to understand the blend of different ingredients from different suppliers to achieve quality improvements, and to reduce waste streams”, he explains.

IntelliAM advised an unnamed dairy producer trying to make the best use of hygiene requirements between the production of different SKUs on the same line.

“If you’re running ten SKUs a day, that might save an hour’s production a day and an hour’s production turns into hundreds of thousands of bottles of milk,” Clayton suggests.

AI is impacting snacks much more than proteins or meat

Abhinav Agrawal, AlixPartners

However, it’s not a level playing field in terms of benefits when it comes to assessing the AI productivity impact on different food categories, Agrawal proposes.

In areas like meat and pastries, for instance, AI-driven robots don’t yet have the “delicate” touch to avoid damage, although developments are being made there.

“My experience is that for the protein manufacturer, the ROI from AI so far is still there but it is modest compared to snacks and things like that, largely because there are some regulations around humans doing certain inspections.

“In a nutshell, AI is impacting both sides but it is impacting snacks and those things much more than the proteins or meat and those types of products.”

“Secret sauce”

Food manufacturers that Just Food approached to comment for this article were generally disinclined to chat or did not respond – Tyson Foods, Nestlé, Mondelez International and Kellanova fell into the former camp, for instance.

It’s a “sensitive” topic, Agrawal says, because some food manufacturers are now using custom solutions rather than the “out-of-the-box software” in the past – their “secret sauce”.

Cargill was, however, more willing by identifying its “patent-pending” computer vision technology CarVe, which measures red meat yields. It was rolled out at the company’s beef facility in Friona, Texas, in 2024 and additional plants will be added in a “phased rollout”.

Through cameras, CarVe uses AI-inputted models to assess human cutting and trimming techniques to ensure consistent quality and to reduce waste.

“Those gains, together with steadier throughput and better use of labour and materials, strengthen overall plant economics and help us stay cost-competitive for customers,” Jarrod Gillig, senior vice president for food at Cargill’s North America beef unit, says.

“Our philosophy with CarVe is to test, learn, refine then scale. We pilot first, capture lessons quickly and scale once the business case is clear. This approach lets us invest wisely, stay nimble and roll out proven innovations across our facilities when they’re ready to deliver value for producers and consumers.”

Cargill’s comments were echoed by Roger Gaemperle, the head of industry strategy and marketing in EMEA markets at US-headquartered automation and tech firm Rockwell Automation.

“There is a lot of potential or high impact on margin” through increased yield output when AI assisted robots can replace the human role in meat processing, which is often not “nice work” conducted in a cold environment, Gaemperle says.

“If you really want to optimise the margin and also productivity or the output, I think these are the systems that producers need to look at. And, over the next few years, the systems will get smarter and smarter,” he envisages.

Another AI benefit is the reduction of waste in raw materials, Gaemperle adds, using the example of milk powder production and the need to have a consistent milk fat percentage to ensure equal quality by production unit.

It’s becoming transformational because of the speed of the algorithms and computing power

Roger Gaemperle, Rockwell Automation

More generally, Gaemperle adds: “It’s really becoming transformational because of the speed now with the algorithms and computing power. That’s really what has opened up many new opportunities that were not feasible in the past.

“Look at it holistically, the whole production process, and identify first where there are potential opportunities, and then prioritise those opportunities. If the return on the investment can be proven first with one line, then additional use cases can be implemented over time.”

Implications for jobs

It would be amiss here not to address one of the biggest concerns over the implications of AI technologies across industries – jobs.

Opinions varied for this article but the underlying theme to emerge was new types of jobs will be created.

“If you take a modern food and beverage organisation, whilst there is arguably a high degree of automation, there is still a hell of a lot of human interaction and judgement. AI will replace all that,” Clayton opines.

However, he throws in the caveat: “There’s an argument that jobs will change, that there will be a transfer of skills because of AI. But there will be better jobs, it’s not about job reduction.

“People are intrinsically part of the solution because a) you need the automation engineers to support obtaining the data to create the insights in the first place and then b) they need to look at the people on the output to react to these or manipulate data, like data scientists.”

Agrawal suggests the food industry will eventually replicate the semiconductor sector, where companies such as Nvidia and TSMC are using AI to the “fullest”.

“What I see as the biggest change in four or five years is that we will not need as much human labour in manufacturing. I don’t think it’ll lead to mass unemployment because there’ll be other manufacturing opportunities,” Agrawal suggests.

For Ashley at Peak, boardroom education has a key role to play.

“Those jobs might look different going forwards in terms of individuals that can map decision logic and map processes and help architect AI systems. The nature of those jobs and the skills required may shift slightly but I think there is huge upside economically and for every workforce in the AI era,” he argues.

“I think that for the manufacturing industry it basically means that there’s incredible opportunity to deploy pilots and capabilities that you can test very quickly to see if they drive business uplifts but it also means there’s an education challenge in boardrooms.

“I think it is a watershed moment but it is evolving continuously and businesses need to lean into that and try and figure out how to harness it as best as possible.”