Industrial companies see artificial intelligence (AI) as a key enabler for profitable growth through increasing efficiency, flexibility and differentiation – however, insufficient quantity and quality of data, and lack of AI expertise are among the challenges that need to be overcome to allow for broader AI adoption. These are key findings of a survey of 858 predominantly European professionals and executives in the industrial sector1, conducted by Hewlett Packard Enterprise (HPE) and Industry of Things World, Europe’s leading Industrial IoT conference. The survey also revealed that industrial companies will deploy hybrid architectures, with AI infrastructure distributed evenly across the ‘industrial edge’ and data centres or the cloud. This will enable both real-time inference at the edge and data correlation and deep learning across locations.
“AI is at the heart of the fourth industrial revolution, a key enabler to take the step from automation to autonomy, create growth and competitive advantage,” says Volkhard Bregulla, Vice President Global Manufacturing, Automotive and IoT, Hewlett Packard Enterprise. “Our survey shows that the European industrial sector has clearly understood and embraced the strategic power of AI – but it also reveals that it will be essential we close the data and skills gap to fully unleash its potential.”
Double-digit growth on both revenue and margin – 95 percent success rate of AI projects
On average, respondents expect to grow their revenues 11.6 percent by 2030 as a result of AI adoption, while simultaneously increasing margins by 10.4 percent. AI is expected to yield benefits in virtually all activities along the industrial value chain, as well as creating differentiation for the companies’ products and services. This expectation is also fuelled by high success rates of completed AI projects: 95 percent of respondents who have already implemented AI in their company say they achieved, overachieved or significantly overachieved their goals. Accordingly, survey participants on average plan to invest 0.48 percent of their revenue in AI in the next 12 months – a significant amount considering that the average overall IT budget is 1.95 percent of revenue in the manufacturing industry2. In line with this positive outlook, two thirds of respondents expect that new jobs created by AI will balance or outweigh the number of jobs made redundant by AI.
The majority of respondents (61 percent), are already engaged with AI, with 11 percent having already implemented the technology in core functions or activities, 14 percent planning to do so within the next twelve months, and 36 percent evaluating the implementation. Survey participants’ AI use cases are spread across the entire value chain, including research and development (38 percent), demand forecasting (21), production planning (18), operations (32), maintenance (34), sales (20) and services (29), among others.
AI is seen as a tool to drive both efficiency and differentiation
When considering new technologies, companies tend to focus too narrowly on cost reduction while not paying enough attention to opportunities to increase revenue and differentiation. This is not the case for the respondents to this survey. While the goal “increasing efficiency in operations, maintenance and supply chain” received most votes (57 percent), many respondents also pursue goals like “improve customer experience” (45) and “enhance products and services by adding new features” (41). Other key goals are “quickly and automatically adapt to changing conditions” (37), “create new business models” (34) and “better match supply and demand by improved forecasting and planning” (32).
One of the most common AI use cases can illustrate the breadth of business goals that can be achieved with AI. Applying AI for predictive, or prescriptive maintenance can maximise the availability of production equipment and optimize maintenance processes (efficiency) – but it can also be sold by manufacturers as a value-added service to their products (differentiation). Moreover, the insights provided by AI on customers’ product usage habits can enable “outcome-as-a-service” business models where the customer pays for results instead of the hardware itself.
Respondent’s expectations with regards to AI-enabled business outcomes by 2030 reflect this balanced view, with, on average, 13.9 percent expected cost reduction, 11.6 percent expected revenue growth and 10.4 percent expected margin increase.
AI will be deployed in both the edge and the cloud
The survey also explored to which degree industrial companies will use AI intelligence from central locations – i.e. data centres or clouds – and from decentral locations – i.e. the edge. This is one of the key characteristics of an AI enterprise architecture. For example, some AI algorithms must run in close proximity to the industrial equipment, at the edge, because they control time-sensitive and critical processes – such as is the case with autonomous vehicles or robots. On the other hand, selected data from industrial equipment distributed across locations must be aggregated and correlated in central data-centre locations to enable advanced machine learning or deep learning to optimize the AI algorithms.
Survey participants were asked which percentages of AI are deployed in a data centre/cloud and at the edge today, and which percentages they expect for 2030. The current distribution of 39 percent for data centre/cloud and 32 percent for the edge reveal a slight tendency towards centralised deployments, but it also reflects the still relatively low adoption rates, with some companies deploying AI either in the data centre/cloud or at the edge. Respondents expect this to change by 2030, with 55 percent data centre/cloud and 52 percent edge deployments. One can conclude that hybrid AI deployments will become the norm over the coming years.
Data and skills are key challenges for scaling AI adoption
Even though survey participants report that completed AI projects have been very successful, this applies to only 11 percent of the sample who have actually implemented AI. Moreover, half of these respondents have either implemented AI in only one of their activities, or they have implemented a proof of concept. This confirms the fact that AI still is a nascent technology.
What are the key challenges that need to be overcome to further scale AI usage? The number one obstacle is data. 47 percent of respondents voted for “lack of data quantity and quality to feed AI models” and 34 percent for “lack of data governance and enterprise data architecture” as key challenges for AI adoption. In fact, the data available for feeding AI models is a key success factor for scaling AI adoption, because algorithms can only be as good as the data with which AI models are fed. This not only requires a data-capturing infrastructure, but also data governance and architecture to provide standardized, labelled, clean data across the value chain.
Another key challenge for broad AI adoption is the “lack of AI/analytics skills and knowledge” (42 percent). Despite this bottleneck, only 12 percent of survey respondents primarily focus on sourcing external AI expertise – a strong testament of the strategic importance AI in the industrial sector. The majority (55 percent) employs a mix of internal and external expertise and a third primarily focuses on building internal expertise by hiring from outside the company, and by training and developing own employees.
„Unfortunately, there are no shortcuts when it comes to implementing AI in the industrial enterprise to create competitive advantage,“ says Volkhard Bregulla. „Companies must define their AI strategy, identify promising use cases, source the data, buy and build technologies, and put the right people and processes in place. This will be a journey, and our survey results clearly suggest this is a journey worth embarking upon.”
Source: Hewlett Packard Enterprise
1 858 respondents (managers, directors, C-suite executives) from verticals including manufacturing, IT, transportation, chemicals, energy and consumer products participated in the survey “The Present and Future of AI in the Industrial Sector” via an online questionnaire between August 1 and September 20, 2018. 61 percent of respondents are based in Western Europe, 22 percent in Central/Eastern Europe, 7 percent in North America, 5 percent in Asia/Pacific, 3 percent in South America, and 1 percent in the Middle East and Africa.
2 https://www2.deloitte.com/insights/us/en/focus/cio-insider-business-insights/technology-investments- value-creation.html