How to apply decision intelligence to automate decision-making

Decision intelligence is one of those terms that sound vaguely familiar, even if you’ve never come across it before. Like many category-defining terms, it can mean different things to different people. This is a feature category-defining terms either have by design, or acquire through extensive use.

Gartner defines decision intelligence as “a practical domain framing a wide range of decision-making techniques bringing multiple traditional and advanced disciplines together to design, model, align, execute, monitor and tune decision models and processes. Those disciplines include decision management (including advanced nondeterministic techniques such as agent-based systems) and decision support as well as techniques such as descriptive, diagnostics and predictive analytics”.

Erick Brethenoux, a distinguished VP analyst on artificial intelligence (AI) data science and decision intelligence (DI) at Gartner, frames DI as, “a practical discipline used to improve decision-making by explicitly understanding and engineering how decisions are made, outcomes evaluated, managed and improved by feedback”. 

For the first time since Brethenoux has been involved with AI techniques (for more than 35 years now) there is a word on which everybody (IT, data specialists, process experts, AI engineers, business people, subject-matter experts and even executives) agrees and has an almost similar definition, he notes. That word is decision.

In October 2021, Gartner identified DI as a 2022 Top Trend. Several vendors have identified with that category: Busigence, Domo, Diwo, Peak, Quantellia, Sisu Data, Tellius, Urbint and Xylem, plus Google, IBM and Oracle, to name but a few. Aera Technology is also among them, claiming to have been doing DI before it was called DI.

Today, Aera is announcing new capabilities for its Aera Decision Cloud at the Gartner Supply Chain Symposium/Xpo™ 2022. Aera founder and CTO Shariq Mansoor weighed in on decision intelligence, Aera’s DI offering and how it is relevant for supply chains and beyond.

Decision intelligence is an amalgamation of techniques

What everyone seems to agree on is that DI is an amalgamation of many techniques. DI encompasses elements of data engineering and data science, analytics and machine learning, as well as business intelligence and rule-based systems.

Mansoor believes this list is on point, but he also believes there are a few more items in the mix. Prior to launching Aera in 2007, Mansoor had a vision for what he called “the self-driving enterprise”. He had been working on building the data foundation for the Aera Decision Cloud, when he met Frederic Laluyaux, Aera President and CEO, in 2017.

Mansoor and Laluyaux shared the vision. They joined forces and after four years of hard work and $50 million raised, Aera’s Decision Cloud was released. As Mansoor elaborated, the platform is multi-layered, with many touchpoints aimed at different roles in organizations. At the heart of the platform is what Aera calls the Cognitive Operating System, which combines elements of data, intelligence, automation and engagement.

The data layer includes crawlers that automatically discover and extract data from a multitude of enterprise systems, a workbench to transform and process data into a unified decision data model and a digital twin designed for decision-making.

The intelligence layer delivers capabilities in analytics, AI/ML and what-if analysis and planning. The automation layer features a process builder, automation rules and a mechanism to execute decisions on external systems.

Finally, the engagement layer features a decision board that shows how organizations are executing decisions; an augmented digital assistant that interfaces with users via search, mobile and voice; and a personalized inbox for decision recommendations.

Mansoor identified all of the above as being part of DI. He acknowledged that none of that is new or revolutionary in its own right, however he emphasized that bringing all these together in an integrated platform that is easy to access and engage with is where the value lies.

Brethenoux on his part noted that DI allows everyone to focus on the outcome and consider the technology as an enabler. But to make that happen, he added, the first capability needed is the ability to visually map those decisions:

“There are many more capabilities that are coming together to form what we have called decision intelligence platforms (DIP),” he said. “Those capabilities are coming from various software clusters that are contributing to the emergence of those platforms, like composite AI platforms, process-focused platforms, proactive intelligence systems, etc.”

Automating decisions with trust and accountability

Both Mansoor and Brethenoux seem to point to the same direction of technology convergence and enablement. Mansoor referred to the decision cloud as “an engine to digitize human decision logic.”

He noted that Aera combines machine learning probabilistic engines and rules-based deterministic logic to automate the entire decision process, all the way from generating recommendations to enabling people to review and accept or reject them and then executing them and learning from the outcomes.

Brethenoux believes that DIPs are the future of business AI systems. “Conversations I have with clients confirm the need for DIPs,” he said. “When those conversations shift to the “decision angle” it becomes much more natural to consider how AI techniques can contribute to the solution of the problems exposed by clients.”

The reasoning behind the notion of automating decision-making, as Mansoor presented it, comes down to avoiding decision fatigue and making the best possible decisions. However, there are some key issues to address here: trust and accountability.

Trust, that is, in the sense that human decision makers need to trust the data and the reasoning process based on which a recommendation for a decision is made. Accountability, in the sense of assigning credit, or blame, where they are due. Should a decision maker get credit for a recommended decision that turned out to be good, or blame for one that turned out to be not so good?

Aera deals with the issue of trust by having what Mansoor referred to as “a permanent memory”: storing the entire context of recommendations and decisions, along with the point in time and the data used to generate the recommendation. Aera also provides confidence scores along with recommendations and uses a feedback loop mechanism that connects with external systems to execute decisions and record outcomes.

Aera also offers different modes of operation regarding provided recommendations. The first mode is called decision support and provides contextual analytics for humans to make decisions. The second mode is called decision augmentation and provides recommendations with a high confidence score based on historical data and users are asked to accept the recommendation. The third mode is called decision automation and takes the human out of the loop.

As for accountability, Mansoor likened this to the discussion around autonomous vehicles. “The question is, is the developer responsible, is it the business owner, or the process? Decisions are made today which were not handled in the past,” he said. “That’s the other thing which we are seeing: we go to customers and they are not even making decisions, because they don’t even know that a problem exists.”

As humans, he continued, “we cannot handle hundreds and thousands of decision points in real time. As a platform vendor, what we do is we provide tools and guardrails for our customers to design robust skills. We provide a way to add logic to course-correct in real time. People can make mistakes, things can happen, even the machines can make mistakes. But if you can catch it and correct it in real time then the impact is much lower.” What they usually see, however, is business taking ownership of the decisions made, he went on to add.

Supply chains and external events

Aera’s Decision Cloud is used by global enterprises across consumer packaged goods, healthcare and life sciences, food and beverage, manufacturing and more, including the likes of Merck and Unilever. Supply chains are of particular interest for them.

As Brethenoux noted, mapping out supply chain decisions allow organizations to extend the thinking beyond only automation and make explicit the various sub-decisions that have to be part of a bigger supply chain decision flow.

“Making those decisions explicit allows organizations to adapt much faster when disruptions strike, knowing what to change, adjust and re-direct.  Supply chains are also part of much wider decision networks, allowing them to understand the causality and dependency aspects of their efficiency”, said Brethenoux.

Supply chains are also a good example of how unanticipated events can cause disruption, rendering recommendations obsolete. Mansoor said that Aera already incorporates external data, such as weather forecasts or competition analysis. The next step, he added, is to be able to incorporate that in the decision-making process.

“Vendors are really excited because this is one of the biggest challenges they have,” Mansoor said. “If there’s a disruption like a fire in this location, or a hurricane, what will the impact be on the customer? Because they don’t have the inside data.” 

If there is a three-day delay, he added, perhaps it is perfectly okay for the customer because the customer is holding seven days of inventory. “But for some customers, it may have a big impact there,” he said. “So this is a next step, which we are working on now with some who are providing this kind of information out there.” 

Aera skills and updates

Aera’s platform works with what Mansoor referred to as skills, alluding to some similarities with Alexa skills; i.e., domain-specific applications. There are many out-of-the-box skills, such as demand forecasting and planning and connectors to systems such as major ERPs. That, and the fact that it’s a SaaS platform, speeds things up considerably. End-to-end implementation can happen between four and six weeks, or between seven and ten weeks if new skill development needs to take place, according to Mansoor.

Skills can also be customized and this is where Aera’s product update comes in. The new release provides new skills, as well as enhancements for data engineers, data scientists and developers that optimize their experience in Aera Developer™, the platform’s integrated development environment (IDE) for creating and deploying skills, or modifying any existing ones.

Updates include integration with Jupyter data science notebooks to easily integrate data science projects and other custom code using Python and R, AutoML options to enable development of ML decision models without requiring expertise in model creation, tuning and deployment and enhancements automating monitoring, deployment and versioning of machine learning models to make it much easier to operationalize ML models.

The confidence score framework, which learns from past, similar recommendations and outcomes to help determine the decisions that can be automated, is also updated. Finally, there is a new Graph Explorer, offering advanced graph capabilities to enable data relationships to be used effectively in the decision-making process.

This visual interface is already attracting lots of attention, due to its ability to visualize complex networks and dependencies, as Mansoor noted. But there is more than meets the eye here: Aera leverages complex graph analytics, combining infrastructure from graph vendors with proprietary implementation. One of the areas in which Aera has filed for a patent is what Mansoor called a depth-first execution graph, which takes complex human decision logic and digitizes it.

Mansoor noted that today’s comprehensive release for both business end user and data science teams underscores how Aera is continually advancing and evolving its platform to enable digital decisions at scale in an increasingly complex business environment.

Originally appeared on: TheSpuzz