Various flavors of decision modeling, analysis, and automation have been around for decades. However, recent cloud, data management and AI advances are planting the seeds for a new era of more integrated, granular, and capable decision intelligence platforms. This mirrors how process modeling, capture, and mining tools have coalesced into modern process intelligence platforms.
I recently caught up with Aera CEO Fred Laluyaux to discuss how this new category of tooling works and why it has reached a new stage of maturity. We previously covered Unilever's journey to develop a self-driving supply chain on top of the Aera platform. It seemed helpful to help pick apart this new category of enterprise infrastructure in more detail.
One early signal of maturity is that the depth and breadth of decision coverage have increased significantly over the last year. Laluyaux explains:
It was kind of limited in my mind to supply chains, and now it's going across all lines of business. We're under tremendous pressure from our clients to help them scale out this thing. The second is the same clients deploying Aera for supply chains are now pushing beyond. It's going to be in finance with PNL forecasting or cash management. It's going to be in revenue management with digital media spend optimization. So, you're seeing going across the value chain, which we predicted, but it's starting to really now happen.
Another major milestone was when Western Governors University, the largest online education institution with 176,000 students, used the Aera platform for scalable, personalized student support. The new use case assists faculty in improving course completion rates for struggling students. Laluyaux says:
It's exactly the same technology and the same platform we've been deploying at Unilever, Dell, and Exxon. It has now been applied to other use cases, and we didn't have to tweak anything. It's just the core of what we do. So, for me, this is super interesting because you're seeing the applicability of DI moving beyond the traditional world of enterprises that are trying to improve performance in supply chain and finance into other areas.
For background, Laluyaux cut his teeth as general manager at SAP, where he first wrote about the need for a new type of decision management platform to manage the explosion of data and the acceleration, granularity and complexity of decision cycles. After another three years as CEO of Anaplan, an enterprise performance management platform, he decided to strike out on his own in 2017 to realize his original vision for a new kind of decision orchestration system at Aera. He explains:
That we had to start from scratch was a little bit difficult. And if you want us to do what we did, you literally cannot retrofit an existing engine to do what Aera does. You have to start from scratch and tackle four main pillars: the data, the intelligence, the automation, and the engagement. And it literally needs to be designed from the ground up."
People have talked about decision support systems and business analytics for decades, but this feels like something different. Defining Decision Intelligence, Laluyaux says:
I frame Decision Intelligence as the digitization of decisions. People are making decisions every day. I need to decide how to ship my inventory from this place to that place. I need to decide how to load the trucks. I need to decide what to buy. I need to decide what the right forecast level is. I need to decide which campaigns I'm running. How do you make those decisions? You identify a problem, or you work on a schedule. You need it to react, so you get the data.
Then, you deploy the logic, which is your experience plus the decision flow. You leverage technology to help project, predict, optimize, and allocate, and then you have a series of authorization levels. I can make that decision up to that level. Otherwise, talk to my boss. And then you make a decision, you capture it, and then you have to execute it. So, decision intelligence is the process of digitizing that decision-making. When the velocity of decisions is coming too fast, it's too complex for a brain to do it. Sometimes, decisions have to be made in real time to have an impact.
One of the big challenges is that the information needed to make better decisions can be scattered across multiple enterprise apps. Also, it needs to be collected, harmonized, and processed in a way that does not slow down transactional systems of record like ERP, supply chain, and order processing. So, an essential element was developing a technology that could pull the required information from these systems using agents deployed on the online system.
But they could not just pump it all into a data lake as is. Laluyaux explains:
I think companies have had the strategy to just dump everything into a data lake and it was going to save their life. And it's not the case. The issue is people don't know how to use the data because it's too complex. And we, as humans, can't. We're not computers. There's a point where we can't process all of this. So, resolving the data problem for us was paramount. And this is why we had to build it from the ground up because the data is the connection between the brain and the nervous system.
Also, the data needed to be organized to help understand decisions and measure their success. Aera developed a decision data model containing financial and operational metrics and the decision data. It provides a permanent memory of the decisions made on a given specific problem, such as inventory optimization, order management, logistics, transportation, and procurement, stored in a particular way. These are intertwined and logically structured with the operational and financial data.
Another big challenge lay in paving a foundation for trust. More focused machine learning models can achieve considerably higher accuracies than gen AI ones, but users must understand how and why they make decisions. And this can take time as the models improve and humans build trust. Laluyaux says:
Trust in the world of digital decision-making is actually not that different than trust between humans. So, what Aera does is basically replicate or mimic the work that a human operator, an experienced analyst, or a planner would do either on schedule or on an event. I look at the data, which is sometimes very complex. Sometimes, it can take seven days to resolve a problem manually because we have to gather the data. You have to deploy the decision logic, come up with the right options, document it, then go to someone to get it approved. And then you have to execute it. They need to trust the person who's coming with the report.
Those are the steps that basically build trust between a manager and a director. After a while, they know they can trust each other because the logic is always sound, and the data source is solid. So, in the platform that we built, we need to be able to demonstrate being one click away from the source of the data and the logic of the decision. When we start deploying Aera in a big company and a large enterprise, there is always that first phase of challenges.
It took a while for some of their early customers to accept the Aera recommendations. For example, in the beginning, Unilever only saw about 20-30% of recommendations approved by humans. However, that has improved significantly over time as the system has learned what humans approve.
It's also important to appreciate that the nature of decisions is becoming more challenging over time. A decision that improves sales may come at the cost of customer ratings, carbon footprint, or profits.
Decision Intelligence feels like it's just starting to hit its stride as a new category of platforms. Major enterprise vendors, including IBM, SAP, and SAS, have begun rolling out decision intelligence platforms. In addition, several legacy business rules management system platforms are modernizing for a fresh take on the technology. Also, other vendors, like Actico, Cogility, FICO, and Quantexa, are developing more targeted platforms to improve decision intelligence for anti-money laundering, credit scoring, fraud management, and cybersecurity.
There seems to be a fundamental shift in how a fresh approach to managing decisions might help support the rise of the autonomous enterprise. For a while, it seemed like business process modeling might allow enterprises to construct a better model of how their businesses worked. However, this required a lot of manual effort from expensive process efficiency consultants. And let's face it, how many might be inclined to fudge the facts for these consultants when our jobs are at stake?
The recent transformation of process mining and process capture tools into process intelligence tools speaks to the need for a more automated and simpler approach to understanding complex decision-making chains. Decision Intelligence could be on the same trajectory if vendors like Aera can improve the speed and value of decision-making processes at scale.
