AI Leadership Insights: AI and Processes

Mike Vizard: Hello and welcome to the latest edition of the Techstrong.ai video series. I’m your host, Mike Vizard. Today we’re talking with Marlon Dumas, who’s chief product officer for Apromore, and we’re talking about how AI is going to be applied to process mining. Marlon, welcome to the show.

Marlon Dumas: Thanks. Thank you for the invite.

Mike Vizard: What is the fundamental challenge with process mining? You would think companies kind of understand what processes they have in place that they are using to drive their businesses, but perhaps that’s not the case and maybe we need to start at the very beginning before we think about applying AI to figure out what is it we’re trying to automate in the first place.

Marlon Dumas: Absolutely, absolutely. We have numerous experiences going into companies that indeed know what their business processes are and why they are there, but they’re incredibly surprised when they actually see with the data how the process looks like. It’s just like, “Wow, I did not know we had so much variability. I did not know we had a standardization problem. I did not know that all sorts of non-compliant issues are happening in my process for various reasons.” And if you have not done a little bit of hygiene and cleaned up a little bit of all these deviations that are hanging around, then you can throw as much AI as you want. It is like throwing money at nothing. So you have to first clean up, get your house clean so that AI can magnify the efficiency.

Mike Vizard: I feel like in a lot of organizations there’s more exceptions than there are rules, and so that we need to just go back in and say, “Hey, here’s how this process is supposed to work,” and then clean up all the exceptions we’ve created over time.

Marlon Dumas: It’s a little bit more subtle than that. It’s not about cleaning everything. So when you look at a process with process mining, it looks like a spaghetti, there’s stuff all over the place. What you need is to understand what part of that spaghetti is good spaghetti, al dente, and what part of it is rotten? So what part of it is bringing down your performance? So the first thing we do in process mining is to define what are the KPIs that you want your process to achieve, to agree on those ones, to capture them. And then to perform something called root cause analysis to trace down which parts of your processes are contributing to you achieving your KPIs and which pathways are completely rotten and you need to eliminate, standardize, and bring up into your performance.

Mike Vizard: How does the process mining work? Am I applying some algorithms to go find out how my processes work? Or in effect, am I using AI to get ready for AI?

Marlon Dumas: It’s a combination of both. So it’s a combination, so the thing about process mining is that it’s really meant for humans. It’s meant for business users. It’s meant for business users to understand what they are doing. So a lot of it is visualization. It’s applying techniques that are simplifying a little bit the spaghetti to let you see different parts of it, a little bit like an X-ray machine in the airport if you have seen where people can look at different layers of it. And part of it is AI and it’s like trying to say, if you want to improve this KPI, what parts of your processes, which rework loops, which bottlenecks you need to tackle, and which other bottlenecks or rework loops are inconsequential? So it’s a combination of automated analysis and visualization.

Mike Vizard: Can I determine, through that process, which of my processes are kind of driving the most critical revenue streams, the highest amount of profit? I feel like sometimes a lot of businesses don’t understand exactly where the money they count on is actually coming from.

Marlon Dumas: Yes, indeed. So let’s say at the level of a given process, an end-to-end process, yes, process mining can tell you what are the parts of your process that are driving your margins, your top margins and your bottom margins. And from there, where do you need to work? What rework loops you need to smash, what over-processing you need to get rid of? And it’s a little bit more work to do it at the level of an entire process portfolio, but that’s where, let’s say process mining is combined with various other methodologies, process architecture, lean 6 sigma, et cetera. So process mining is not replacing the need for you to do proper enterprise architecture and to apply a disciplined approach to continuous process management, it’s rather trying to help you to become more efficient at analyzing individual processes and understanding what pathways in your process are right and which ones need to be revisited.

Mike Vizard: You mentioned compliance earlier. Will auditors be using process mining as part of their job to figure out what’s going on in an organization as they come in to do their evaluations?

Marlon Dumas: Two things here, so when we go and work with banks in particular and we show them there’s spaghetti and we start applying AI techniques to zoom in. And we start making them see certain parts of the process that are anomalous and that are associated with, let’s say, anomalous performance, sometimes they stop us and they say, “Stop there. I don’t want to hear anything else. We need to look at this specific case.” So they are finding terrible noncompliance issues in front of us the first time they see their processes, despite the fact that they are going through audits over and over again all the time. Why? Because the audits are focusing on you having in place certain controls and these controls are neither complete. And they don’t look at everything nor do they capture all the possible non-compliance situations in which you can be.
So process mining does not replace your audit, it is completely complementary and we are seeing a lot of interest actually from compliance managers, from process mining. It’s one of the big topics, I think, for process mining in the coming couple of years is how to magnify the effectiveness of auditing.

Mike Vizard: Who’s in charge of mining the processes? It seems like in a lot of organizations, not quite clear if it’s some sort of department level, C-level executive running a business unit? Is it a chief digital officer, is it somebody else entirely? I mean sometimes I wonder if this whole thing just falls through the cracks because nobody’s in charge of it.

Marlon Dumas: Absolutely. Things are currently falling into the cracks because process excellence initiatives are being taken from different angles, and sometimes we go and work with some business units and they’re very successful, but their cousins next door are not doing anything about it. Sometimes though, we work with digital transformation offices, so we work with chief innovation offices and there we managed to get process mining more widespread across the organization. But indeed, I think that companies need to put in place a stronger dedicated force to drive their digital transformations, to drive their operational excellence and their customer excellence under a common umbrella, coordinated umbrella. And it’s in that context that we see that the best use cases, the most powerful use cases are coming when we are working at the intersection between a IT digital transformation, operational excellence and customer excellence, all coming together to drive excellence across the organization.

Mike Vizard: Is process mining an event or is it something that we should be continuously doing because the processes are dynamic and they keep evolving?

Marlon Dumas: Yeah, it’s a never-ending stream of events, I would say. So in a way, it is a process, as you can talk about the process of process mining. Organizations that adopt process mining need to get very quickly to the point where they set up and they develop a center of excellence to spread this discipline across the organization and drive value at a higher level. But it takes time, it takes time, and while you get to the point where you have the maturity to have a center of excellence in process mining, you have to get the low-hanging fruits and demonstrate value. So I would say process mining starts with a few events where you analyze certain process and you discover you have some very powerful insights and you start driving transformation that generates value. And then what you need is to have a series of those events to generate a virtual circle where effectively the cost savings, the revenue increases that you are getting from process mining are being reinvested until the point where you build a center of excellence and then your little storm becomes a tornado of operational excellence going across the organization.

Mike Vizard: Do you think, as we go along here that it’s going to become easier to set this up? I think people have a perception in their head that process mining is challenging to engage in or is it getting any easier or where are we on this adventure?

Marlon Dumas: I think that sometimes what happens is that one company looks at another one and sees that what they have achieved with process mining. But maybe that other company has been doing it for five years, for six years, for seven years, and then they see what they are doing and they just want that and they want it tomorrow and that’s not how it works. You cannot take shortcuts with process mining, I will argue you cannot take shortcuts with operational excellence in general. So you have to set that as a goal, but you have to start small. You have to start analyzing some core processes, you have to start driving value in those core processes, you have to internally build the capacity to do process mining and then little by little build that competence and spread it across the organization. So I think when I hear process mining is hard is when I see an organization trying to go from zero to 100 in two months and that’s not going to happen.

Mike Vizard: So what’s that one thing you see people doing when it comes to process mining and just managing processes in general that makes you shake your head and say, “Folks, we’re better than this.”?

Marlon Dumas: When we see organizations trying to engage, keep engaging in some long running manual process modeling approaches, of course I shake my head. I think that process mining has completely overtaken the discipline of a manual process mapping. Even if you’re going to do process modeling, you should start with process mining and then take those models and refine them so that you start from reality and start modeling reality rather than modeling perception. So that’s definitely one area. Another area where I shake my head is when people start approaching process mining from an automation perspective. So we know Bill Gates saying that automation magnifies existing efficiencies, but it can also magnify existing inefficiencies. So I think we have to be very careful with the automation hammer and be very modest about how we apply it and start by having a very good understanding of the processes, using process mining to accelerate that understanding and of where can we inject automation to achieve the highest levels of efficiency.

Mike Vizard: All right, folks, you heard it here. This whole shift to AI may force us to, “Finally look under the carpet of our business processes,” and we may not like what we see, but at least we have some tools to figure out what’s going on and then maybe move forward from there. Marlon, thanks for being on the show.

Marlon Dumas: Thank you so much.

Mike Vizard: And thank you all for watching the latest episode of the Techstrong.ai video series. You can find this episode and others on our website. We invite you to check them all out. Until then, we’ll see you next time.