“The ability of real-time monitoring in determining how to solve a problem is magic"

'Big data' is here to stay, but the next step is to improve business performance through data analysis and instant reaction. The Chief Revenue Officer of Celonis, Miguel Milano, is committed to this in order to rectify the errors resulting from poorly integrated, overlapping systems.

Photo: The Chief Revenue Officer of Celonis, Miguel Milano. Credit: Courtesy of the interviewee.

By Patricia Ruiz Guevara

If we could observe the digital core of a company, we would see a jumble of sparkling, immediate and dizzying interconnections. Millions of interconnected layers that, with the growth of technological developments and the addition of devices and systems, have become increasingly complex.

In this mass, there is a risk of information loss or process failure. An invoice that does not reach the customer, an order that is not notified, a shipment that is sent twice. These small inefficiencies consume a lot of money and time. But there is a solution within technology’s reach: to analyze the data to react in real time to these problems.

At the multinational company Celonis, they have gone a step further and, far from simply storing customer data, they connect it to find hidden inefficiencies and apply intelligence to improve business processes. Miguel Milano, Chief Revenue Officer (CRO) at Celonis, explains how software as a service (SaaS) in the cloud is the future of business performance.

Cloud technologies are becoming increasingly sophisticated and specialized: they not only offer cloud storage, but also add new value. What do you want to add at Celonis?

Our goal is to become the execution data layer of business processes. Once we have that, we can apply artificial intelligence [AI] in different ways.

For example, with our proprietary process mining algorithms we can reconstruct a process: understand that here is a step, here is a manual intervention, there is a request or an input order.... These algorithms understand the process like X-rays looking through a system and figuring out what's going on.

We've found that putting the analysis capability on top of the information that the customer already has more easily reveals what the problems are with traditional processes.

What kind of failures do you encounter?

That companies pay twice, early or late, and reimburse invoices in the wrong way. Because of these kinds of things, they miss out on existing discounts on contacts that they are not aware of. Also, that most manufacturers send products to their customers and forget to enclose the invoice; this happens every day in every company.

"What companies want is to correct errors in real time, but, even if they understand what they should be doing, they can't."

In the end, companies usually know what kind of problems exist. It's like if you tell an oncologist that a patient has lung cancer; when you know there's something there, it's easier to find it. What companies want is to correct errors in real time, but, even if they understand what they should be doing, they can't.

What prevents a company from solving these process problems?

Companies have on average 450 IT systems that support business processes. It is not unusual to see a company with 3,000. The layers of these systems are rigid, they are fragmented and disconnected, they don't communicate with each other, and many are obsolete. Integrations are made, but the problem is that outside the ideal process path, there are thousands of variants. The process can go through different nodes or agents, and many of these variants are not supported by the systems.

These problems, the execution gaps, are the company's silent killers: no one sees them, no one knows that invoices are being paid twice or that a product is being sent to a customer without being charged, because it happens under the screens of the IT systems implemented in the companies.

Is this going to increase?

Yes, because there are more and more IT systems for smaller areas of the company and with more specific problems.

How can this be solved?

At Celonis we bought a technology integration platform a few years ago, platform as a service [PaaS], which allows us to make integrations and API calls.

We have already been able to capture transactional data and apply artificial intelligence, big data, and machine learning models to predict and analyze complexity. What is now a revolution is the ability to, from an event that we have real-time data on, perform an action or business logic.

In the case of orders that go out without an invoice, what we do in the action flow is that every time something is going to be shipped from the warehouse, Celonis triggers a wait order. That wait lasts 0.000000001 seconds. The AI analyzes whether there is an invoice ready to go with the order.

"The next disruption is the ability to link process mining with automation without changing systems."

In 99.9999% of cases, there is. There, Celonis does nothing. But one in 10,000 times it's not ready, the order goes out without it, and the customer is never billed anymore, because there are no people tracking backward billing flows. In that scenario, Celonis triggers automatic business logic and tells IP to create an invoice for that order. It attaches it and then it can be sent, all in a matter of seconds.

This real-time monitoring capability to decide when to act and what business logic kicks in to solve a problem is magic. The next disruption is the ability to link process mining with automation without changing systems.

How can these types of services be budgeted?

Our direct cost is the data in the cloud: the more we mine it, the more it costs us. That's why we tend to limit customers in the amount of data they can deploy.

In the past, we used to buy by process. Then we went to simplifying it and charged by data and users. The most expensive are the analytics, which analyze the process, and the business users, but here we charge less and less because we want people to consume the insights [key information of value]; that's what really attracts them.

"The future of process mining and the application of data intelligence will be an app store with hundreds and hundreds of apps to solve very specific problems."

That's why we have created a value-based contract model, where the client can deploy whatever they want and we help them find that information. What we are interested in is that companies adopt our strategies.

Can this apply to all sectors?

We are now developing use cases: banks, insurance companies, the telecommunications sector... We are collecting and customizing, and we take what we learn from one process in one company to another. When one company in an industry is in, they all have to be.

The future of process mining and the application of data intelligence will be an app store with hundreds and hundreds of applications to solve very specific problems: customers will be able to choose what insights they want to see from that process and what actions or remedies they can take to mitigate that system failure. It's going to be a wide range of options.