Webinar Recap: Engineering Decision Intelligence into your Operating Model

Summary

In our Future.Now webinar, “Engineering Decision Intelligence into your Operating Model,” co-hosted with Deloitte, we explored what it takes to turn a promising decision intelligence capability into lasting, organization-wide value. Supply chains once rewarded the best modelers and the sharpest hunt for efficiency. That has changed. Volatility and complexity now define the work, and the practitioners who thrive are the ones who can absorb that unpredictability rather than plan it away. Technology plays a powerful role in advancing how decisions get made, yet it delivers its fullest value only when paired with the right operating model. Without that context, too many programs stall in a kind of proof-of-concept purgatory, where pilots show promise yet never reach real scale.

This second episode of a three-part series focused on the operating model. Think of it as the connective tissue that lets technology and volatility work together through governed decision-making. We examined the questions that clients ask most often. Where does decision intelligence fit in the existing landscape? How does it change the way people work? How do you govern the decisions it produces, measure whether it lifts performance, and organize the teams that build and scale it? A live look at the Aera platform, along with a customer example, showed how each piece comes together in practice.

Key Takeaways

• Decision intelligence sits above existing systems as a layer of intelligence. Rather than replacing an advanced planning system, a WMS, or a TMS, the platform draws on them as inputs, coordinates decisions across the wider supply chain, and writes actions back for those systems to execute. In many cases this strengthens the underlying tools and unlocks their fullest potential. The platform becomes the system of intelligence, the place where decisions are made, governed, and remembered, while your existing systems stay the record for data.
• When to implement depends on where value and readiness meet. Some organizations start before an ERP migration to capture high-value, non-regret opportunities, since waiting carries a real opportunity cost. Others move in parallel with the migration to define the mandate of each layer cleanly. Still others wait until afterward, when the goal is to accelerate scaling across markets. The right timing hinges on resource availability, urgency, and a few minimum data conditions, not on perfect data.
• Decision intelligence reshapes the way people work. A short-term planner once spent hours locating imbalances across scattered systems. Then came days of weighing transfers, expedites, and production changes before securing sign-off. With a supply-demand balancing skill, that same planner works from a single inbox of recommendations. Each one shows its drivers, its trade-offs, and the cost of doing nothing. The role shifts from analyzing data to monitoring outcomes and handling exceptions, so more decisions get made with greater accuracy and speed.
• Governance is a set of deliberate choices, not an afterthought. Before decisions run at scale, organizations decide how far to automate and where to place the human: in the loop, on the loop, or out of it. Those choices can be tied to the value of a decision or the importance of a customer, and they can expand as trust grows. Confidence scores and thresholds decide what runs on its own. A control room then tracks every recommendation, outcome, and rejection, giving teams the visibility to govern with care.
• Value can be measured at three levels, and it compounds. The impact of a single recommendation shows up in revenue, cost, and metrics such as OTIF. A full skill carries a cumulative effect as decisions improve and automation grows. The real gains arrive when skills combine: pairing an inventory-balancing skill with a truck-builder skill, for instance, produced more than 40 percent in additional savings. Layered skills create network effects that move an organization toward enterprise-wide optimization.
• A center of excellence sustains the capability as it scales. Deployment tends to organize around three domains. A digital studio designs and builds new skills with the business. A digital factory keeps existing skills running and speeds up their scaling. A value office coordinates governance, architecture, and performance. Together they mark a shift, moving from process-led teams that execute transactions to a hybrid model that builds digital products to support decision-making.
• Everything comes back to value. A holistic operating model has to match the market’s need for agility. That means the physical network and the organization must be able to absorb decisions made in seconds. One customer brought three skills live and generated $2.7 million in value within the first eleven weeks. That same program identified shortages three times faster and pushed out excess purchase orders fourteenfold. Choosing the right use cases and protecting time-to-value remains as critical here as it was in the first episode.

Speakers

	Suraj Ramalingam, Senior Solution Engineer, Aera Technology Suraj brings extensive experience in global supply chain management, having held senior roles at Salesforce and Procter & Gamble. At Aera, Suraj focuses on enabling customer success through technology-driven transformation, helping organizations operationalize AI-driven decision-making at scale.
	Kirsten Delnooz, Partner, Supply Chain & Network Operations Practice Leader, Deloitte Kirsten is a partner in Deloitte Belgium’s supply chain practice, specializing in operating model transformation and network design. Kirsten helps organizations adapt their workforce and operating model so decision intelligence capabilities can be adopted, embedded, and evolved over time.
	Kevin Overdulve, Partner, Supply Chain & Network Operations Practice Leader, Deloitte Kevin leads the supply chain practice at Deloitte Belgium and drives the firm’s global decision intelligence agenda. He also serves as the Global Aera Technology Alliance lead, helping organizations transform supply chain decision-making through AI.

Full Recording

Access the full webinar recording here.

Q&A

During the presentation, attendees submitted questions. Below are several of the key questions and answers.

Q: We already run Kinaxis and o9. Honestly, what does this do that our APS doesn’t already?

A: An APS is built to plan. It produces the optimized plan, and it does that well. What it does not do is own the cross-functional decision that follows when the plan breaks, the transfer, expedite, or manufacture call that cuts across planning, logistics, and procurement at once. That is where Aera comes in. It sits above the APS, orchestrates that decision across your systems, and writes the resulting action back into them. So this is not a replacement for Kinaxis or o9. It runs on top of them and takes over where their job ends.

Q: Our data is a mess, and half of it is validated and regulated. How is this not another two-year data project before we see anything?

A: You do not have to get your data perfect before you begin. The Decision Data Model aggregates, certifies, and harmonizes decision data from across your systems, and it does that with low system impact through open, bi-directional integration rather than a rip-and-replace migration. Because you deploy one skill at a time, the first can go live in weeks instead of waiting on a multi-year enterprise data program. Regulated data is handled to the standards your compliance teams already enforce: access is role-based and row-level, enforced at the moment of execution, with a full audit trail on every decision, which is often the first thing validation and compliance teams ask to see.

Q: Who is actually accountable when the model makes a bad call? I can’t tell my board “the AI decided.”

A: Accountability stays with the person who authorized the decision, and the whole system is designed around that principle. You decide, for each type of decision, whether a human sits in the loop, on the loop, or out of it. Thresholds then determine what the platform can auto-accept and what gets routed to an approver, and every decision carries the logic and the data behind it. When something does go wrong, you can trace it back to the cause, whether that is the logic or the underlying data, and correct it. It is fully auditable, not a black box you have to defend on faith.

Q: We piloted AI and got some results, but we were not satisfied. Why is this different?

A: That is a fair concern, and a common one. A pilot can look promising and still stall, usually because it proved a model works without ever touching a real decision, so the value never had a chance to compound. What makes this different is the starting point: one use case, running on your own data, wired into the systems your teams already use, with value tracked from the very first day. Each decision also feeds the next. That is why the program keeps improving in production instead of plateauing once the proof of concept is over.