Next article
Making AI work for business: Insights from BMW, IBM Consulting and more
Artificial intelligence took off in a blaze of hype and speculation around two years ago. Today, there are few company agendas that don’t feature AI. And it’s now being put through its paces as businesses assess its potential, test applications, and establish the most valuable use cases.
It’s a critical time where ROI can be won or lost depending on how much benefit businesses are able to uncover. At Celosphere 2024, Celonis’ annual user conference, we launched AgentC—a suite of AI agent tools, integrations, and partnerships that enables our community to develop AI agents in the leading AI agent platforms. It also allows them to use AI agents pre-built by partners. All these AI agents are powered by Celonis Process Intelligence, making them understand how the business runs and how to make it run better.
During a Celosphere session, leaders from major enterprises that have successfully scaled AI shared their insights to help accelerate businesses’ time to value.
Dr. Sven Jung, Head of Economic Intelligence at Handelsblatt Media Group’s economic research institute, chaired a panel with:
Stephan Bloehdorn: Executive Partner and Practice Leader AI, Analytics & Automation at IBM Consulting;
Andre Luckow: Head of Innovation and Emerging Technologies at BMW Group;
The Celonis CoE lead fromm a global leader in sustainable packaging.
They discussed the hype and practicalities of utilizing AI for the enterprise. Here are six takeaways.
1. What research tells us about enterprise AI implementation
The context for the panel discussion was a study of almost 300 decision makers conducted by the Handelsblatt Research Institute with Celonis, nearly 50% of whom believed they are pioneers in the field of AI. It showed that use of AI is currently most widespread in IT departments, where key use cases included software development and cybersecurity. After IT came marketing, with AI used less frequently in finance departments, despite lending itself to standardizing data for reporting.
The three most popular advantages that respondents identified were increased efficiency, increased productivity, and reducing errors. This emphasizes how companies can use AI to manage pressures around labor shortages and cutting costs.
In terms of challenges, most mentioned employees’ ability or willingness to engage, due to a lack of skills or acceptance of AI tools. More than half of the companies also face an issue from siloed data holding back AI effectiveness, as their processes are less flexible and lack transparency or a connected network.
2. Established AI use cases
In the 2010s, BMW Group began applying AI to identify frequent patterns in data across different company domains, such as production quality. The sustainable packaging leader’s early use case was purchase-order blocks, building a spare-parts intelligence robot with Celonis that analyzes material allocation.
IBM Consulting highlighted an HR solution that’s driven by AI assistants, agents and other integrations, through which 94% of all their HR transactions run automatically. Another platform is enabling every IBM consultant – developer, seller or architect – to be supported by AI, comprising 2,000 assistants.
Read: 6 ways to manage material shortages the smart way and meet supply chain SLAs
3. Areas where AI is working particularly well
The panelists explained that enterprises have many advantageous AI deployments to choose from, as Bloehdorn said: “AI is a very versatile technology, so there are, in principle, no boundaries where you can’t apply it. It depends on your strategic positioning, on your priorities, what makes more or less sense.”
He pointed to operational efficiency use cases as some of the most obvious to go after, such as quality inspection in manufacturing. But he also outlined how AI can help businesses create personalized customer experiences, along with research and development to unlock new products and services – from chemicals to coding assistance.
Product development is also a valuable use case for BMW Group, where Luckow states: “There's not a single part of our value chain that's not affected by AI.” Generative AI will accelerate vehicle design and engineering, for example, by summarizing vast volumes of research. But AI will also help predict outcomes of car crash simulations across a wider range of variables and scenarios, strengthening safety standards.
Read: AI and Process Intelligence: Insights from industry leaders at Celonis Day
4. How to drive AI buy-in and adoption
One of the biggest challenges for AI implementation is end-user adoption, as the packaging leader’s CoE lead explained: “We can deliver a beautiful co-pilot, but I need to deliver something that people are comfortable interacting with.”
He is an advocate for getting in front of the people who’ll use the tech, whenever you’re onboarding any IT solution. That means roadshows – not only to outline usage guidance and best practice, but to learn how users talk to the AI. You can then go back and teach the AI how to listen to the people ‘on the shop floor’.
The packaging leader is even experimenting with allowing end users to talk to the AI they’re developing, in any language. For example, the AI would understand how an engineer talks, such as how they describe a motor instead of an SAP part number. All the user would need to do then is simply tell the tool what they need and what it looks like.
Read: Responsible AI: How can new technologies respect data privacy?
5. How to decide where to start with AI
The panel then addressed one of the most common objections to AI. Like any new technology, implementation isn’t immediate, so it can take time to see a return on investment.
Luckow’s philosophy at BMW Group is simply getting the flywheel going. It’s a continuous process of achieving business value, then extracting new data from it, and improving the AI. The idea was echoed by Bloehdorn who advises enterprises to be deliberate in their focus areas, looking at your competitive positioning and which parts of the value chain are most critical for you. Then take it step by step: build a platform, examine the data, and the next use case will be easier to find.
The sustainable packaging leader uses process mining to make it easier to know where to start with AI. The technology can confirm an opportunity for AI by assessing whether the process is working sufficiently and with minimal human intervention.
The CoE lead also spoke about the importance of choosing the right AI partner. Investigate the AI vendor’s offering – what it does, how it works, and whether other customers have proven its value. They’ve partnered with companies they already work with and trust, such as Celonis, and started with processes they know well.
Read: The proof is in the pilot: Why AI process experiments can’t wait, according to futurist Steve Brown
6. The role of data in AI
The CoE lead also stressed that successful AI implementation comes down to the quality and availability of data. Their company has seen the potential of AI to clean master data.
They’ve built a layer of AI in the Celonis platform that automatically harmonizes data between their source system, end analytics, and co-pilots.
The IBM Consulting team is working with Celonis on making unstructured data available and included in the Celonis Process Intelligence platform. That could include pulling all the key information from emails or pdfs, then feeding it into the platform. As a result, no information or value is left untapped.
The session wrapped up with the CoE lead summarizing the essentials for any enterprise looking to get maximum value out of AI: “Really ensure that you have the base of the data and the processes there, and then start your AI journey.”
Read: How Process Intelligence helps companies build more effective AI agents at scale
