AI in manufacturing – how ready are we?

Manufacturers are moving beyond curiosity when it comes to artificial intelligence (AI), but for most, the journey is still only just beginning. The Manufacturer AI Readiness Index has recently been produced in partnership with Oracle and IBM. To discuss some of the findings, we spoke with Robert Garratt, Partner at IBM, and Dominic Regan, Senior Director Logistics Solutions at Oracle. 

Deployment of, and discussions around AI in manufacturing is certainly nothing new. Our research does indicate something of a shift however, particularly this year. Manufacturers appear to be moving from exploration to implementation, and from potential to proof.

The landscape still reflects one of caution, however. The AI Readiness Index found that more than 70% of manufacturers remain at an early stage of AI adoption. The figure is a striking one, though perhaps not entirely surprising.

Manufacturing is often viewed as a sector that approaches new technologies with caution, particularly when change could affect operational continuity. And indeed, Robert pointed out that adoption is more readily seen in other industries.

“In the current pace of change in AI, it’s fair to say that everybody is at a relatively early stage,” he said. “But in relative terms, I do agree that manufacturers are perhaps not as advanced as some.

“A lot of organisations with a large consumer-facing presence are already using generative AI for customer interaction, chatbots and similar use cases.”

He continued: “We also see heavy adoption in businesses with large financial processing or procurement operations, where agentic AI can simplify processes and take cost out. Those areas are often more prominent in other sectors than in manufacturing.”

Although Robert argued that the picture is more complex than simple risk aversion – this doesn’t give us a full explanation.

“Engineers and manufacturers are scientific, experimental people who like trying things out,” he said.

“I suspect the bigger reason is that many of the functional tasks of manufacturing and design have already been heavily supported by technology for a very long time. There may simply be less obvious room for AI in some of those areas, at least for now.”

Moving the needle

That helps explain another notable finding in the report. Nearly one in five manufacturers remain unsure of the value AI can bring to their business.

For Dominic, that uncertainty reflects both the relative immaturity of AI adoption and the lack of manufacturing-specific proof points.

“We often forget how recent extensive AI adoption really is,” he said. “Most large language models have only appeared in the last two or three years. There also aren’t yet many publicly referenceable use cases showing exactly how AI is delivering bottom-line value in manufacturing.

“Much of the adoption we’ve seen so far has been in front-office functions such as sales and marketing, which may offer less direct value to manufacturers than some of the back-office or operational use cases.”

He went on to say: “I think it’s a combination of relative infancy and a lack of proven examples that explains why some organisations are still asking where the value will come from.”

Even so, both Robert and Dominic agreed that the conversation has shifted markedly over the past 12 months. As stated at the beginning, AI in manufacturing is moving beyond experimentation for experimentation’s sake and towards a more serious focus on practical impact, measurable outcomes and long-term value.

“The needle is certainly moving,” Robert said. “There can’t be a professional in the world who isn’t testing the potential uses of AI in some way. As time moves on, we’re all starting to work out where it might properly be applicable and where it might properly deliver value.”

He said the biggest challenge in his mind is the pace of change, “AI itself is moving so fast that the needle has to shift quite quickly to keep up with it. A year ago it was all about generative AI. At the moment we’re talking about agentic AI. Who knows what it will be in another year?”

For now, the most common use cases remain focused on internal operations. As found in the reports, this is around reducing cost, improving productivity and streamlining existing workflows.

Dom said that is entirely logical. “If you’re taking existing processes and trying to apply AI to them, looking for efficiency is the low-hanging fruit.“

Using AI to drive process change or new business outcomes is much more challenging, however. Dom pointed out that most organisations are still at the stage of improving their existing processes.

He said: “Questions centre around how to automate what they have currently and how they can make it more efficient.

“To move beyond that, you need a different skill set. You have to ask what AI can do for the business, what it can do for the end customer and how it could help redesign the process itself.”

‘Data is everything’

If one theme underpinned the discussion more than any other, it was data. The report identifies data quality, governance and trust as prerequisites for successful AI adoption. This certainly reflects our constant conversations on AI, it does appear to be the case that this is well recognised among manufacturers.

Robert was in agreement that manufacturers are increasingly aware that without strong data foundations, even the most promising AI project will struggle.

“Data quality feeding into AI was the dominant discussion at a recent manufacturing roundtable Dominic and I hosted,” he said.

“With traditional analytics, if the data wasn’t available or wasn’t right, the system would simply error and not compute. The danger with AI is that if the data is insufficient, it may hallucinate.

“That means data quality and trust in AI are very closely linked. If you don’t have the data quality, you risk hallucination, and then you lose trust. AI is not a sticking plaster that resolves the need for good quality data. In manufacturing, which is such a numeric, science-based and precision-led industry, data is everything.”

One of the biggest barriers

The report also highlights ROI, security concerns and skills shortages as major barriers to adoption. Dominic believes the skills challenge is especially important, not because manufacturers lack technical talent, but because AI requires organisations to think differently about process design and business outcomes.

“There’s been a tendency to treat AI as a technology decision, which it really isn’t,” he said. “So much of it comes back to looking at business processes, understanding the required outcomes, and then seeing how AI can feed into that.”

Dominic is concerned by the challenge he is seeing around a lack of process skills and believes it could be a major blocker to progress. “Being able to dissect a process, understand the preferred business outcome, and then look at it from an AI perspective – if you don’t do this it’s one of the biggest barriers to not just to adopting AI, but scaling it.”