28-04-2025
Decisions, Decisions… Next Steps For Practical Applied AI
AI engineering is currently a market hall. Because the automation intelligence market is still growing so fast, it has naturally splintered out in various directions (almost like a food market) with vendors specializing in a number of different disciplines, some offering meat, some fish, some vegetables and with others more focused on providing cutlery, chopping boards or seasoning.
A number of firms are dedicated robotic process automation developers, working to provide us with 'bots' that will automate repetitive tasks - UiPath, Automation Anywhere, Blue Prism and others. In the other corner, we have the process mining, process orchestrion and intelligent document processing specialists - Celonis, IBM, Abbyy and Aris. Then there are the AI-powered supply chain planning and forecasting purists - here we find ERP specialists in particular, such as SAP and Oracle, alongside Blue Yonder and Coupa. There's an entire subset of digital twin specialists spanning IoT AI toolsets, including o9 Solutions, Bentley Systems, Siemens and General Electric. Then there's plain old general-purpose AI, where we have to put IBM Watsonx, OpenAI, Anthropic, Microsoft and Google.
Where many of these vendors will want to go next is a land grab on the more direct use of AI inside business processes. Being able to evidence a more applied approach to AI, so that it starts helping make practical and quantifiable business decisions is the next tier, because AI on its own is just intelligence that sits in the corner.
While discussion surrounding agentic AI that is capable of making more autonomous decisions is rife, this subject is clouded and woolly in some areas; some straightforward decision management functionalities for supply chain, finance and operations might be refreshing.
If any of this backdrop is valid, it may provide some context for why Aera Technology is attempting to put its mark on what the company calls decision intelligence, a set of software processes designed to provide recommended actions inside business processes. With its branded Aera Decision Cloud service, the platform can predict supply shortages or delivery delays, offer shipments rerouting advice and automate decisions relating to inventory replenishment without human intervention.
'Decision intelligence is AI for decisions, it optimizes decision-making across a business,' said Fred Laluyaux, CEO and co-founder of Aera Technology. 'As decisions grow more frequent and complex, organizations need faster, more accurate data-driven decisions. By recommending and automating actions through AI, data and analytics, decision intelligence accelerates the entire decision cycle, helping companies operate efficiently, reduce costs, enhance customer service and respond to change.'
Laluyaux suggests that decision intelligence is profoundly changing the way work gets done and how decisions are made in large enterprises. Significantly, this could reshape the way some parts of business operate and create new roles to architect and guide the use of tech-powered intelligence with finer granularity. As the industry starts to scale the use of tools at this level, it will of course need to weed out bias and be able to capture the business context of every recommendation generated and decision being made. If we then talk about applying these decision tools to truly complex scenarios in real-time, then we may enter a new stage of automation.
'Every day, companies make countless decisions that impact costs, efficiency and customer satisfaction. Most of the time, there's no record or historical record laid down to define why the decision was made the way it was,' said Laluyaux. 'This means that the act of the decision itself provides little (or minimal) insight to help future decisions. Instead, inefficiencies persist. Meanwhile, the pace of decisions continues to increase, heightening complexity.'
Aera talks about the process behind so-called decision intelligence as a technology process that documents each decision and its rationale. It goes beyond creating a single 'data point' or notification that a problem has been detected and needs attention; it works at a more detailed level to aggregate and harmonize data from multiple sources - structured databases, unstructured text, devices across the internet of things, cloud applications and so on - to provide a more comprehensive view. This technology enables enterprises to assemble 'decision flows' (literally a schematic that explains the components and mechanics of any given decision, or string of decisions) and integrate specific AI models while building custom interfaces for specific needs.
Analyst house Gartner predicts that by 2026, 75% of Global 500 companies will apply decision intelligence practices, including logging decisions for analysis. Other figures here suggest that by 2028, 15% of daily work decisions will be made autonomously through agentic AI.
'By not capturing decision rationales, enterprises lose valuable learnings. Traditional 'back testing' can help uncover trends, but not necessarily explain the why behind human choices,' stated Laluyaux. 'For instance, a planner may override an AI-generated forecast because they know a big customer is set to make a late order. The AI would not have known that. If the system doesn't record the override reason, the model fails to learn. That means it won't anticipate the situation the next time.
Or a planner might reject an AI suggestion simply because it feels too risky. They don't want to be blamed if an AI decision is wrong. So, they make other adjustments and inadvertently distort data.'
Aera offers decision intelligence services to enterprises including Unilever, Merck, ExxonMobil, Mars, Kraft Heinz and Dell. But does decision intelligence bring all the traders in the AI market hall together at last and provide new light at the end of the intelligence tunnel? No, yes and somewhat.
There is some light being shed here. Aera's pre-built industry models for common enterprise functions like supply chain management give it some kudos over basic RPA bots. Further, its closed-loop automation capabilities enable the platform to create cycles where decisions and actions happen autonomously based on approved protocols… and that (arguably) makes it more sophisticated than a basic analytics or task execution tool.
But for every specificity and specialism, there is a trade-off. We can reasonably say that the focus on the above-noted enterprise functions narrows the intelligence engine here i.e. IBM's Watsonx is capable of understanding a far wider universe of happenings, from stock market values to the weather to the price of cabbage.
Aera also faces challenges stemming from a confidence in its own core technology proposition i.e. its technology relies on businesses accepting the notion of autonomous decision making, as we allow the software robots to make decisions and take actions inside business systems that have a close relationship with the bottom line. This is 2025, this is not where many enterprises are yet in strategic or tactical operational business terms.
Perhaps most of all, Aera is working in an extremely crowded market that sits up against platform and toolset brand names that are more deeply adopted, more widely implemented and more comprehensively served in terms of partner ecosystem integration services. In a world where technologies like this are complex to install, that's important, no matter how intelligent your intelligence services are.
