By Dr. Alicia Martinez de Yuso, ZLC Project Manager.
The European process industry sectors are behind the curve, relative to discrete manufacturing and to industry and commerce more generally, when it comes to the adoption of potentially game-changing developments in Artificial Intelligence, Machine Learning and Big Data technologies. That at least was the finding a couple of years ago of SPIRE (Sustainable Process Industry through Resource and Energy Efficiency), the public-private partnership which brings together the major European process sectors – water, non-ferrous metals, cement, minerals, ceramics, pulp and paper, chemicals, refining, engineering and steel.
That is particularly surprising given that process industries tend to be especially data-rich, the complexity of their upstream and downstream relationships, and their centrality to the quest for better use of energy and resources in a circular economy. SPIRE convinced the European Commission’s Horizon 2020 programme to fund a two-year project, ‘AI-CUBE’, in which ZLC has played a significant role, to examine the problem and how to resolve it. The project has now reported, in the form of a comprehensive ‘guidelines’ document, and though we say it ourselves, we believe this is essential reading for any process industry organisation considering how advanced digital techniques and the business models they enable can ensure their continued competitiveness, profitability, and contribution to a better future.
The guidelines come in six parts. The first analyses the current state of the art in the process industries, based on an extensive trawl through the academic, professional and trade literature, further informed by an in-depth survey of over 100 experts, both users and suppliers of AI-BD solutions, which yielded over 50 concrete use cases. The Cube concept itself refers to a tri-axial visualisation of the situation: the various process industry sectors, the different technologies that are becoming available, and the specific process areas in which these are (or are not) being applied. The ‘surface’ thus generated reveals peaks where adoption is relatively far advanced, and troughs where adoption is lagging. By application, AI-BD is most advanced, perhaps unsurprisingly, in process control and optimisation, and also in R&D and in predictive maintenance. By contrast, the clear potential in areas such as supply chain management, product customisation, innovation and market analysis appears to be underexploited.
In the second part this snapshot is enhanced through a Maturity Level Analysis developed through an online survey. This model looks at five key dimensions which influence an organisation’s ability to successfully adopt AI-BD – Strategy, Organisation, People, Technology, and Data – at four levels, from level 1 (little or no adoption or knowledge of the topic) to Level 4 (full adoption and optimisation, companies championing practices which are well-established and recognised as important). The guidelines offer a framework which companies can work through to assess their maturity levels – and thus to see where the likely barriers to improvement lie.
The paths and barriers to successful implementation form section three of the guidelines. Validated with our industry contacts, this allows companies to assess where the barriers (within the specific company, or more widely in the industry) lie, and leads readers to pathways, strategies and guidance on what to do about them. The enabling factors/barriers are found to fall into three broad categories: human related; technology and data related; and company/strategy related.
Implementation of AI-BD of course involves commercial investment decisions, so section four examines the Value Proposition, based on nearly 300 recent reports in the scientific and technical literature. This section enables readers to consider a target area for improvement – energy saving, for example – and see what AI-BD solutions are being found relevant or which make the biggest contribution, within the company’s own sector or across the process industries more generally. (This is also revealing sectors and issues for which there is little or no evidence of successful implementations, suggesting areas where knowledge transfer from one sector to another may be highly desirable).
If the firm has identified its goals, the barriers it faces, and the value proposition for AI-BD adoption, how does or should this affect the organisation’s business models? This is the subject of section five, which guides readers through the business models that will tend to maximise the benefits or realise the value proposition. The section identifies eleven current trends which can be grouped into seven AI-BD driven business models. (Above these can be discerned one overarching or mega trend, that of Hyperconnectivity, by which is meant the creation of agile, dynamic supply chains that flex to changes in demand through seamlessly integrated planning and execution). The different models may overlap, but address particular priorities. For maximum environmental impact, for example, a Sustainable, Circular, or Collaborative Symbiosis BM may be preferred (all these terms are well explained in the report). Where human impacts are the prime concern there is a Human First model. For maximal operational and logistics impact a Proactive & Predictive, or a Data Use/Control & Quality for Supply Chains model may be appropriate. Or where commercial considerations are the overriding concern, there is the AI-As-A-Service model. The guidelines introduce an approach to appropriate Cost Benefit Analysis techniques which can capture the benefits, value drivers, and costs: tangible, intangible and implicit as well as the potential risk costs, both short and longer term.
The final section of the guidelines looks into the future. It is inevitably rather more speculative, but is intended as a basis for discussions which are going to become increasingly important. In the inevitable 2×2 matrix (every report has to have at least one!) one axis shows the degree of implementation, from application globally across the organisation or sector, to a fragmented and silo-based approach. The other axis is more societal than technical – will implementation take a fully responsible approach to possible impacts on the workforce and wider society, or will there be unfettered implementation without any qualms over the effects on people? None of the four ‘destination boxes’ is without its potential problems, whether it be a ‘human free’ model where AI relentlessly replaces and devalues people, or an unsustainable (commercially and environmentally) ‘business as usual’ model of limited and fragmented applications. Our experts agree that, probably, all of the options will manifest at different times and different places – whether one model becomes dominant, and how we cope with that, depends on decisions we are making, and discussions we should be having, right now.
The AI-CUBE Guidelines cover a lot of ground, and are essential reading for strategists and technologists in the process industries. They can be downloaded from https://www.ai-cube.eu/news/ai-cube-guidelines-ebook/.
For more information please contact Dr. Alicia Martinez de Yuso, at [email protected].