AI in practice - Celonis’ VP of Digital Transformation shares how AI can support system and process change right here, right now

(Image by Gerd Altmann from Pixabay )

It goes without saying that at this moment in time, the hype surrounding AI - particularly generative AI - is at its peak. Following the launch of ChatGPT at the end of 2022, consumers, enterprise buyers and vendors spent a lot of time last year talking up the possibilities of how Artificial Intelligence could change the way that we live and work. And whilst future gazing has its place, it’s not hard to imagine that the swathe of AI announcements over the last 12 months, and the ongoing discourse around its impact, will have left technology buyers a bit bewildered. 

Which is why this year, in 2024, diginomica is looking to get a deeper understanding of how AI in all its forms is practically being used on the ground, right here, right now. As the hype dissipates, what buyers need is insights into how other organizations are applying these technologies - and how these technologies can be deployed successfully - at this moment in time. 

This was the focus of a discussion I had with Kerry Brown, VP of Digital Transformation at process mining vendor Celonis, earlier this week. Brown has a long history in the world of enterprise technology, having previously also worked at SAP for 17 years (in roles that were similarly focused on transformation change and future of work), and prior to that being Global Director of Change Leadership at Coca-Cola Enterprises. 

Celonis has of course been considering its own role in how enterprises will use its process mining capabilities in an AI-driven world, most notably with the announcement of its Process Intelligence Graph late last year. At the time, Celonis said that the Process Intelligence Graph could enable enterprises to have better control, and understanding, of what data they use for their AI models, providing greater governance. Equally, a recent Celonis survey of over 1,200 global business leaders highlighted the connection between process optimization and successful AI adoption - with 72 per cent of respondents concerned that process shortcoming could hinder their success. 

But again, for a lot of organizations, they don’t know where to start or understand how AI can help them right now. Brown had some useful insights into how the technologies associated with AI - particularly generative AI - can be used to support system transformation, especially with regard to how content can be generated for process owners. So, let’s get into it. 

What Brown is seeing right now is that a lot of the work being carried out at the moment by enterprise buyers is in terms of how generative AI can be used to support documentation. She explained: 

One focus is the visibility that [buyers are] getting across their business execution. Instead of looking at just how might a company better train, or better do workflow, it's looking at where AI can generate more substantive content?

For example, in system transformation we are having this dialogue of AI generated documentation and content for system transformation programmes. 

Think about a programme, for example, the order to cash process. I need to do business process procedure documentation across the whole thing, by virtue of having visibility into the process, we now have the ability to automate the documentation of that process by looking at that process, versus asking somebody: ‘what are you doing? And please write it down’. 

This feels like a helpful example of work that enterprises typically rely on human knowledge and man hours to carry out, which is prone to error, dull and doesn’t add a great deal of value to the outcome. It’s necessary work - documentation is important - but it’s more often than not drudge work. Using generative AI to accurately build that out with human oversight seems like a good way to put an organizations’ people to better use. Brown added: 

Content generation with a data driven lens, of that actual process, allows for a capturing that reflects reality, not wishful thinking, or best memory, but rather a view of ‘this is what's actually occurring’. And then that can become a source for a data driven dialogue on design change or configuration or user specifications. 

Indeed. If processes are critical to the successful adoption of AI more broadly, not relying on siloed voices or knowledge, but data-driven insights, to understand those processes could be helpful. 

Work that is prone to human error

Brown said that, at the moment, Celonis is seeing AI being used to expedite often tedious work, or work that often is prone to human error. Looking back at the adoption of previous general purpose technologies, this makes sense. More often than not the tools that are adopted early on are applied to use cases that take time, don’t add a significant amount of value, and where mistakes are easily made by people. 

Another practical example provided by Brown focuses on customizing software training assets, specific to the unique capabilities of an organization. She said: 

If somebody's doing a process, but they're failing at it, we can trigger them with an action flow back to prebuilt assets and training assets inside SAP and Enable Now, for example. 

One of the things historically that a customer would need to do with Enable Now is buy it and then edit it to add in their own policies and procedures and information, so that it's unique to them.

With AI we have the ability to have that deconstructed, or have a broader view of all of the assets that are distilled to that user, so that as a company you don't need to edit that content - because you will have access to the unstructured information related to that role.

AI has the intelligence to collect that, so that when I'm executing against my work, I can source to that, versus editing it and continuing to edit it. You're able to collect that information in a way that the user is getting what they need, curated for them by AI.The evergreen nature of keeping content current for the user using AI, is also supplementing that for sure. 

These examples may seem like low-hanging fruit - and in many ways they are - but I like the way that Brown is providing a practical, helpful use case that many organizations and users would look at and think ‘I could actually do with that’, rather than a superlative AI scenario where the workforce is entirely reimagined. In most cases enterprise buyers will be starting small and scaling up gradually over time. 

Change management

Brown also had some thoughts regarding how enterprises should consider their approach to AI adoption, with a focus on not isolating people away from the technology - keeping them close to the change and bringing them along on the journey. Firstly, Brown acknowledged that this is going to be challenging, given the tendency for employees to ‘build empires’ within enterprises and protect them at all costs. She said: 

I'll go back to a phrase I used for a long, long time and I still use: people don't hurt what they own. So if I'm invested in it, and it's part of what I care about, I'm going to protect it and grow it. 

If I boil down change management into one sentence, it’s about expectations and accountability. So, what can I expect to be different and what do I need to do differently? 

The fear of the unknown can cause innovation to be stifled, but it can also lead to people actively resisting change. We are going to see a lot of this with the proliferation of AI, particularly given how much of the discussion has centered around job losses. Brown said: 

Going back to 30 years ago, computers were going to take all of our jobs away and in fact it created many, many, many jobs. If I look at AI, it’s the same reaction. ChatGPT has certainly magnified what AI means, but AI isn't new, and that's probably the biggest misnomer. 

AI is ever present in everything we're doing all the time and it has been for probably about 10 years. And so the best opportunity is really how organizations can invite and involve people in the identification of opportunities for where AI and ChatGPT can be used to make work easier, smarter, etc. 

For Brown, this comes back to an age-old problem when it comes to change management: getting IT and business users to work closely together to bring about successful change. She added: 

Where we're seeing success is where the business and IT together are doing proofs of concept. Rather than IT saying ’we've got a wizard to make your jobs disappear’, where the reaction is going to be ‘it's a threat’ - there’s an opportunity to say ‘no one likes writing business process procedures, it's a sucky job’. 

The fact that there's a way to do that, and then you as a user can edit it, or I can clarify that it’s correct - that’s far more interesting to me than writing it. So when you're looking at the opportunities, the win is: how to get the people who are going to be part of it, playing in it, together.

My take

A specific and useful conversation with Brown - one that can hopefully help buyers consider not only how AI could be applied today in their organizations, but also how they can successfully manage that change. I hope for more similar conversations this year that look at the specifics and realities of AI for buyers in 2024, rather than simply the possibilities of what the technology could do at some point in the future. Possibly. Maybe.