The capabilities of Artificial Intelligence (AI) are undeniably exciting. But the transformative impact it could have, both inside and outside of manufacturing facilities, is yet to be recognised. We explore some of the current industry use cases, as well as the barriers and potential costs of not adopting.
The fully autonomous factory is a wonderful vision, but for the most part it remains in the realms of fiction, for now. Yes, AI-powered robots are increasingly utilised in assembly lines for repetitive tasks, leading to higher precision and faster production rates. But in an industry where practical planning underpins every approach, we’re unlikely to see AI employed in this way anytime soon.
Some users may expect to find the loftiest AI-driven outcomes, but like any aspect of digital transformation, it comes down to finding where the solution adds value. In this article, that point is made several times by Matt Walsh, Managing Director Manufacturing UK & Nordics at Microsoft.
The effective use of AI looks more like a collection of applications for compact, discrete systems that manage specific manufacturing processes. The use cases may be narrower than first thought, but this makes them more scalable.
And ultimately, they are still remarkable. They will operate more or less autonomously and respond to external events in increasingly intelligent and even humanlike ways. So, what’s stopping adoption?
The current landscape of AI and cloud within manufacturing
There are a few absolutes within a manufacturing facility; efficiency, productivity, and quality control, to name a few. AI isn’t a brand new solver in this regard. It has been extensively employed in predictive maintenance, with algorithms analysing equipment sensor data to anticipate potential breakdowns. Vitally important, as this enables proactive maintenance to prevent costly downtimes.
Quality inspection processes also use machine learning algorithms that can quickly identify defects in products with greater accuracy than human inspectors, ensuring higher-quality outputs.
And within the recent, and indeed, persisting concerns over supply chain management, AI-forecasting is optimising inventory management, reducing waste, and improving resource allocation.
These examples led Matt into admitting, the current landscape “Looks exciting.” He continued: “With the amount of potential there is, within a fairly slow adopting and risk adverse industry, it’s really exciting.”
As well as factory use cases, Matt identified some other starting areas where manufactures are starting their AI journey; within functional areas like finance, HR, or procurement. Some are also starting from the customer facing endpoint.
“What we’re seeing is the companies with clear governance and structure going from a pilot, proof of concept (POC) or minimum viable product (MVP), to scaling – these are the companies that are jumping ahead” he said.
“Everyone who’s in POC purgatory are just running another POC with AI. But some of the key customers who are finally stretching away are showing clarity on their decision making and on their process. They’re clear on how AI can be used, how AI won’t be used and the role of the human in the loop.
“We’re seeing people being busy on AI, but not getting value from it. These are the companies saying ‘here’s another problem, here’s another use case, here’s another thing we could do.’ But they haven’t finished doing last week’s problem.”
From Matt’s experience, it sounds like discipline is required when it comes to scaling AI. There are companies playing with its capabilities, and teams who are running a POC with no intent on scaling. The companies that are taking it beyond that stage however, are starting to drive some competitive advantage, regardless of what area of the business it’s being deployed.
As mentioned, the industry is already using AI across a number of functions, but it was with an air of excitement that Matt said, “We’ve only just scratched the surface.”
He continued. “There are two real underlying technology breakthroughs with this new generation of AI.
“The first is we now finally have a new natural interface. And it starts with language, but it’s going to quickly go beyond that, to see, to hear, to interpret and make sense of our intent and the world around us.
“And secondly, a new reasoning engine, which helps us make sense and find patterns and all that is digitized. And together, they have created what is essentially a new category of computing. This is as significant as the PCs in the ’80s, the web in the ’90s or the cloud in the 2010s.
“Just like you boot up an operating system to access applications or use a browser to navigate to a website, you will invoke a copilot to do all these activities and more to shop, to code, to analyze, to learn, to create.
“I think the entire industry just needs to have some confidence and urgency around pushing it. Microsoft Copilot, for example, is a good place to start. It’s like the Excel for AI. We don’t give people who have Excel a list of use cases to go and use Excel. We just say, ‘here’s Excel, here’s the tool, away you go.’
Barriers to adoption
But in many instances, the question, ‘what’s stopping adoption?’ remains unanswered. Typically, the people closely associated with manufacturers, innovators and disruptors, are the people that fail fast. Whereas the manufacturing industry is terrified of failure, as a result, it fails slowly.
And given the very nature of manufacturing; calculated and process driven, its perhaps unsurprising. We should also consider the backdrop of economic uncertainty over recent years, and the catalogue of world events that are impacting businesses to this day. Can they afford to take the loss of another failed digital transformation project?
Manufacturing is more risk adverse compared to other industries. “I think we’ve probably seen that, to some extent, with AI adoption as well,” admitted Matt.
But manufacturers are also problem solvers. Where value lies in AI driven solutions, they will sniff it out eventually. Matt believes organisations still need help in that regard though, they need to a develop a greater understanding to enable them to identify the value. In some instances, another barrier to adoption comes from not understanding the boundary of the use case and not managing expectations.
“One of our customers recently asked, ‘Is that all it’s going to do?’ Matt explained. “By looking at the value it could create, the answer was yes. I think some people expect it to be this huge panacea, or some magical golden egg. Whereas others are really clear that some of the use cases are narrow, but they’re scalable and they can be used to drive adoption.”
He continued, “You can drag in 70,000 contracts, in multiple languages and an AI tool can arrange them all for you in a consistent way and save you an hour per contract. And the response we get is, ‘only an hour per contract of 70,000 contracts!?’ That expectation management is quite an interesting piece. I think it’s another big adoption barrier.”
Interestingly, the barriers exist in organisations of all sizes. And similarly, in Matt’s opinion, the adoption approach should be no different whether you’re an SME or large, global manufacturer.
Often the SMEs of the world see advanced technology as something unattainable, but the differentiator doesn’t lie in the size of the company, it comes from being clear on what problem you’re trying to solve, having the conviction to execute it, and then the conviction to scale it, according to Matt.
“Every company shares that problem,” he explained. “Large global companies often have a legacy mindset, they’ve operated a certain way for 100 plus years. The smaller SMEs of the world that have been around for 50 years have also operated the same way. There’s an inertia to change from everyone.”
He continued, “One of our customer’s biggest question was, ‘have we got enough data?’ This company had 35 years’ worth of IP data. They wanted a way of ingesting all of this into their AI models, thus breaking down the silos of IP and understanding how they can increase their speed to market.
“You could spend ages questioning if you’ve got enough data, or you could just get going. Don’t come up with another list of another list. It’s vital that you just get started, then have a plan to scale.”
Dangers to lack of adoption
Having already touched on the various global macro trends that have impacted the sector, we could easily flip the ‘See why we can’t adopt?’ to ‘See why you must adopt.’ By that we mean that instead of using the unrest of recent years as a reason to scale back, and retreat on AI adoption, it should be used as a prelude to action. The cost of doing nothing is too great, as Matt explained:
“Take the Suez Canal blockage and the disruption within the global shipping lanes we had a customer telling us about the ripple effect that this is still having around 18 months on. I really do sympathise with businesses, because it’s been one thing after another. But now, the cost of not doing anything is showing itself more often.”
AI could be used to predict these potential disruptions and analytics can provide insights into demand fluctuations and identify alternative sourcing options. With predictive modeling and scenario analysis, businesses could even minimize the impact of global shifts such as geopolitical tensions or trade disputes.
The other risk of not doing anything is that other countries are. The UK already lags behind it’s global counterparts when it comes to industrial automation. The UK has an average manufacturing robot density of 111 robots for every 10,000 employees, which The IFR World Robotics report notes is “very low for a Western European country.” The UK sits 24th in the world in robot density rankings, the only G7 country to sit outside the top 20.
This trend can’t continue, and Matt warned against allowing global competitors to pull away, saying: “It can’t just be a UK challenge, it’s a global challenge. There are manufacturers out in the Far East that are embracing AI at a far more rapid pace.”
Current and future benefits
As Matt mentioned, AI as a technology is only just getting started. It’s future role in manufacturing is well poised to expand even further. As just mentioned, the UK will need to move fast on increasing industrial automation, because advancements in robotics, machine learning and autonomous systems will become more prevalent.
The fully autonomous factory was referred to as fiction at the start of the article, but with future AI capabilities it could be moved a step closer to reality, with factories operating with minimal human intervention. AI-driven adaptive manufacturing processes could enable real-time adjustments in production based on demand fluctuations, market trends, and resource availability.
We could also see further advancements in AI-powered digital twins, which are already revolutionising product design and prototyping.
Matt was also keen to touch on the people side of AI. Not just through the exciting human augmentation that AI can generate with intelligent robots and cobots. (collaborative robots) It could also assist greatly with knowledge transfer, given the ageing workforce in UK manufacturing.
“There are a lot of people leaving the sector in the next five years,” said Matt. “There’s a huge amount of tribal knowledge that could just walk out the door. How do we capture that tribal knowledge, ingest it, and pass that down into the next wave of workforce coming in?”
He continued: “There’s a two sided coin to knowledge management; on one side you have the new people coming in, receiving training and embracing the new ways of working. But on the flip side, we don’t want to lose 25 years of knowledge.
Matt also touched on the interconnected approach across businesses that AI can assist with.
“I think the other current benefit is around the overall value chain. With AI, rich data can flow through, allowing a tier two supplier to help a tier one supplier, then the tier one supplier can help the OEM. I think the watermark could all be raised. It’s easier said than done, we’ve always been trying to crack the data in the value chain. There’s a trust challenge on that.”
He concluded: “But I think it’s the most exciting thing to look at; the overall value chain and sharing of data.”
Can manufacturing learn from other sectors?
AI is, of course, a global fascination at the moment. With Generative AI tools such as Chat GPT being used in every-day life, for a multitude of tasks, it’s no wonder that other customer centric industries are trying to utilise AI capabilities.
Within healthcare, AI is revolutionising diagnostics, treatment planning, and patient care. The finance sector provides insights into how AI can be harnessed for risk management and decision-making. There are also use cases in transportation, where AI is being integrated in autonomous systems and logistics optimisation. These are just a few examples.
“There’s a lot of activity in the direct to consumer (D2C) market,” said Matt. “Gen AI is used in both the sales and customer journey. How can we take those into a business to business (B2B) context? Generative design of product; how can we reduce the time to design and time to market of large manufactured goods?
“The finance, HR and procurement use cases are cross-industry, they’re not industry specific, he continued. “But there’s a lot of rich value being generated from customer interaction within the D2C space. There’s probably some potential cross pollination there, but at the same time, the manufacturers are looking at things from an R&D, product development and lifecycle management perspective, whereas clothing retailers are probably more focused on the customer side.”
Again, Matt urged businesses to be targeted: “I would say go after the value pools,” he said. “Getting a customer that would be happy buying more and increasing their basket size, is probably going to drive a bit more value than reducing the time to market of a new white t-shirt from two days to 1.75 days.
The current and future possibilities for AI in the manufacturing industry is thrilling. However, as businesses embrace AI technologies, it’s vital to adopt a targeted approach to implementation. Like any technology, rather than pursuing it for its own sake, companies should focus on specific areas where AI can deliver tangible benefits. A misunderstanding of where and how it will add value only puts you further back in your adoption.
By strategically leveraging AI to address specific pain points and capitalise on opportunities, businesses can maximize the value of these technologies while ensuring alignment with their broader goals.
For more information, visit Microsoft Cloud for Manufacturing: Tackling data accessibility in manufacturing alongside partners
