The hype around artificial intelligence (AI) and machine learning (ML) has fuelled the perception of these technologies as a magic wand granting instant, transformational change to the UK’s manufacturing sector. AI is a capability, not a quick-fix product, writes Colin Elkins, Global Industry Director, Process Manufacturing at IFS. Yet with intelligent integration and buy-in from users, intelligent planning and forecasting tools can help optimise operations and inform more accurate business decisions.
It’s summer, there’s a heat wave, and England is in the final of the World Cup. The time of year, the weather, and the positive public mood (all of which can impact sales) have been acknowledged by manufacturers and distributors, and factored into their planning. As this is a repeat of a similar scenario that took place four years ago, forecasting is underpinned by historical data housed in spreadsheets and ERP (enterprise resource planning) databases, and manufacturers therefore prepare for a spike in demand for beer, burgers, England flags and the football strip of the team’s star player.
But what if during that same week, the star player’s shady past is revealed on social media, the horse meat scandal hits, and the heat wave is very short-lived. What if the past few months had seen health-conscious consumers shunning beer for low-calorie wine? What if environmentally-conscious venues had banned plastic flags, a new legislation had been passed restricting alcohol sales during the day, and major roadworks had been scheduled on key transport routes?
A three-minute TV commercial for a UK supermarket was all it took to inspire home cooks to attempt to replicate Delia Smith’s latest recipe, a rhubarb and ginger brulee.
Following the commercial, broadcast in 2010, the supermarket sold 12 weeks’ worth of rhubarb in four days – enough for an estimated 61,000 desserts. Sales of other ingredients used in the recipe, including Chinese stem ginger, organic ginger powder, Greek yoghurt and double cream also went through the roof. This isn’t the first time the chef has impacted sales and consumer behaviour, and the term ‘Delia Effect’ was even recognised by the Collins English Dictionary back in 2001 following similar events.
Planning for every eventuality
There are a huge range of factors and thousands of possible variables that must be taken into account when planning for the future and forecasting demand. Earlier this year, for example, a combination of a late bank holiday weekend, Mother’s Day, and warm weather boosted grocery spend by almost 6 per cent. Recording the impact of all of these factors upon sales, stock levels, demand etc. is a significant undertaking for any human. Keeping track of their impact in real time and, at the same time, using all of this information to predict the future, is near impossible.
Accurate forecasting therefore remains one of the most challenging and complex problems within supply chain management. Exacerbating this is the multi-faceted nature of supply chain management. Far from an isolated discipline, it’s part of a business as a whole, with multiple dependencies between finance, manufacturing, purchasing and sales – all of which rely on demand forecasting. Getting this right means knowing exactly what consumers or customers want, when and how much they want it. Get it wrong, and your business is at risk of overproducing, underselling, and having to shift excess stock and manage excess waste.
At the heart of accurate forecasting is robust data, both historic and sourced in real time from numerous internal and external sources. Businesses need to be sourcing information from social media, the local weather forecast, regional public and religious holidays, consumer demographics, political events, spending behaviour and a lot more. Even the most intelligent of human beings lacks sufficient brain power to aggregate, analyse and action this data. The answer is to augment human workforces with artificial intelligence, and adopt ERP tools with embedded AI capabilities. Driven by intelligent forecasting and material requirements planning engines, these solutions allow their users to extract real value from data.
- Five myths you have been told about industrial AI (opens in new tab)
Feeding the AI beast
There is massive potential for AI to reign supreme: in addition to processing and drawing insight from data, AI never forgets, works 24/7 and (provided its fed on a continuous diet of information) can learn in order to inform more accurate decision-making in the future.
Intelligent data management and accurate forecasting can result in less waste and more revenue in manufacturing, and the value AI technologies can bring to an organisation (and therefore guaranteeing ROI) has been widely acknowledged. By 2020, AI and ML in supply chain and manufacturing processes will create up to $2 trillion in additional value for businesses. Furthermore, AI technology is predicted to add €32 billion to the manufacturing output of Germany alone – a region already well-established as a manufacturing powerhouse. Can the UK – whose struggling automotive sector in 2018 contributed to the UK economy expanding at its slowest rate in six years – keep up?
On a global level, the automotive sector has in fact been an early adopter of AI, and is expected to account for the largest proportion of AI in manufacturing: a market predicted to hit $17.2 billion by 2025. Audi, for instance, is using ML software to detect tiny faults in sheet metal parts, automating and optimising the quality control process. Expect to see more big names in the industry adopting more AI-based technologies in coming years: according to a Forbes survey, 44 per cent of respondents from the automotive and manufacturing sectors said that AI would be ‘highly important’ to the manufacturing function in the next five years, while almost half said it was ‘absolutely critical to success.’
However, only 11 per cent of firms in the automotive and assembly sectors have actually adopted AI for supply chain management. AI works harder and can deliver better forecasts far faster than a human counterpart, resulting in stronger predicted growth for businesses. So why aren’t more manufacturing firms adopting and benefitting from AI solutions?
Don’t forget the data
The problem is two-fold. First, an organisation can be the biggest champion of AI and ML. It can be the most eager to integrate these into its ERP. But it won’t be successful if it hasn’t captured the volume and diversity of data to feed these technologies. Falling for the hype of AI is easy, but instead of taking a sudden leap into new AI territory, businesses must first lay the groundwork. Think about what parameters you want to take into account – what are the factors that have affected sales and forecasting in the past, are doing so now, and could potentially impact these things in the future?
Once the necessary data has been captured, organisations can then look to ERP solutions which integrate this from various different departments and sources, into a unified, centralised view. Finally, the solution must be able to gather (and its users access) data in real time from across every endpoint, device and sensor – something that’s becoming even more critical as the number of IoT endpoints within a business grows. By 2020, approximately 1.7MB of new data will be created every second – far too much for any human brain to process and capitalise on.
- Digital transformation readiness: 7 steps to prepare for IIoT and AI (opens in new tab)
The problem with humans
The second problem has nothing to do with the technology or the data – it has to do with us, the humans. Too many of us remain sceptical about the decision-making capabilities of AI and ML, whether that’s to do with a lack of understanding of complex algorithms enabling this, or the fact that we just don’t trust robots! According to a recent study on the use of AI for determining bank loans, only a quarter of consumers said that they’d trust a decision made by an AI system over that of a person. This is despite the fact that AI is capable of processing and using a far greater volume of data, far faster, and with far more accuracy, than any human counterpart.
The lack of confidence in AI technologies also extends to employees’ perceived ability to effectively use solutions with AI capabilities. According to one estimate, 69 per cent of respondents in the UK manufacturing industry rated the lack of internal skills as the top barrier to AI adoption. This is a fair bit higher than the 50 per cent of respondents in same sector, across the pond in the US. This stands to reason, as despite the hype in the UK around AI, as a nation we’re lagging behind regions including the US and China, which invest 50 times and eight times more than the UK, respectively.
Can we keep pace? Is AI the silver bullet to shoot profitability and productivity back into our manufacturing market? Will planners ever place their trust in AI predictions? Will England ever reach a World Cup final? We don’t expect a complete and instant reversal in attitudes towards AI. Nor do we expect businesses – in the manufacturing sector and beyond – to suddenly become data scientists. However, AI technologies which can be integrated across a company as part of a wider digital transformation strategy does exist today. There are ERP solutions readily available which can be configured to fit any business or process, and which are accessible and intuitive for employees of all levels. We might not be able to predict the future with 100 per cent accuracy, but we can learn to use and trust the tools that can help us humans get pretty close.
Colin Elkins, Global Industry Director, Process Manufacturing, IFS (opens in new tab)