For many supply chain organizations, putting artificial intelligence to work is proving to be a much bigger undertaking than simply adding a new piece of software.
That’s become abundantly clear as companies as diverse as Microsoft, Gallo and Blue Diamond Growers have worked to bring AI and more advanced planning tools into their operations. And although each of these businesses is fundamentally different, their experiences point to many of the same priorities: Clean up the data first, rethink processes before automating them, give employees clear authority over important decisions, and resist the temptation to change everything at once.
Perhaps even more important than any of that, though, is that much of the work happens before AI ever even enters the picture.
“It’s about the business transformation,” says Tushar Bala, chief technology officer at enterprise IT consulting firm CloudPaths. He explains how each of the companies makes the fundamental organizational shift needed for successful AI implementation.
Microsoft faced that challenge as the AI boom dramatically changed demand for the company's cloud infrastructure. But with the right technology, Microsoft Cloud corporate vice president Joanna Kostecka says, that demand can now jump by hundreds of percentage points, rather than following the relatively linear growth planners were accustomed to. At the same time, Microsoft has had to take into account constrained supplies of components and power, as well as longer lead times.
Today, the old planning rhythms can no longer keep pace. Processes that once played out over a month now have to operate weekly, or even daily, forcing Microsoft to plan years ahead with strategic suppliers.
“The normal world of predictability actually does not exist, and you need to reinvent dramatically fast what you're doing and how you're doing it,” Kostecka explains.
As Microsoft expanded its use of AI across that environment, Dhaval Desai, engineering manager at Microsoft Cloud, says the company started by examining the work itself.
“We don't want to automate a bad process using AI as a technology,” Desai says.
The company first looked for waste and unnecessary steps across processes spanning product design, planning, sourcing, manufacturing and delivery. Only after simplifying those workflows did it look at where AI could shorten cycle times or help employees make faster decisions.
It's a distinction that Bala has seen across CloudPath’s own supply chain planning AI implementations. He says that technology also tends to be only one piece of the challenge.
“Implementations are not just about technology,” he says. “It's mainly about people, process and data.”
It’s important to understand that companies can have enormous amounts of information without having data that is clean, connected or useful. They can also implement new planning systems while leaving old organizational silos and workflows largely intact. And even a technically sound implementation can struggle if employees aren't prepared to change how they work.
Blue Diamond Growers encountered that firsthand. The agricultural cooperative, which processes and markets almonds across its branded products and ingredients businesses, had separate systems managing demand and supply. The company, which already uses SAP for ERP, wanted to bring those separate systems together in SAP Integrated Business Planning (IBP), while gaining more flexibility to model disruptions and changes in demand.
But as the project got underway, Blue Diamond discovered that some of the information it needed wasn't available in its existing SAP environment.
“Data first and foremost,” says Steve Birgfeld, vice president of IT at Blue Diamond, when describing the lessons learned from the project.
The company had to regroup, realign its information and, in some cases, create data within IBP while preparing to carry it into its newer SAP platform. Given that, Birgfeld's other takeaway was equally straightforward: “Don't over-engineer out of the gate.”
That incremental approach eventually helped Blue Diamond consolidate planning work that had previously taken place across a number of different systems and spreadsheets. As Birgfeld recalls, “what-if” scenarios that once took roughly six hours could be completed in about 20 minutes. The company also gained a common view across its branded and ingredients businesses, and began tying volume planning more closely to financial planning.
Those gains offer one reason companies may be better served by starting with a defined problem, rather than trying to automate an entire supply chain.
Bala recommends establishing the planning foundation first, then adding intelligence on top of it. From there, organizations can apply AI to selected areas such as demand sensing, forecasting or supply optimization, learn from the results, and broaden their use over time.
The same gradual approach is just as important for a company when figuring out the authority an AI model should have. At veteran winemaker Gallo, where SAP systems already supported areas such as IBP, warehousing and production planning, vice president of supply chain excellence Nitin Murali describes the company's AI work with SAP as a “decision-improvement journey.”
Rather than asking which jobs could simply be handed over to an AI agent, Gallo sought to define what Murali calls a “human decision boundary.” The company then identified where human judgment would be most valuable, which tasks lent themselves to automation and, critically, who remained accountable.
“Even where you automate, it should be the human's decision to automate,” Murali says.
With that in mind, Gallo looked to develop a decision register that separates higher-value decisions where people retain judgment from lower-value tasks that can be automated once employees determine that doing so is appropriate. That could include routine work such as order intake or portions of deployment planning. In the end, the goal is to make employees more effective and engaged, rather than removing them from the process altogether.
Microsoft is considering a similar shift. Kostecka describes the future supply chain planner as an “orchestra conductor,” as opposed to someone who’s responsible for one narrow specialty. AI agents could handle more of the manual analysis and spreadsheet work, while planners take a broader view of the business, weigh trade-offs and make strategic decisions. That leaves humans in a place where they can continue to be responsible for the decisions that matter most.
On top of that, it makes governance inseparable from adoption. At Gallo, Murali says teams work backward from a decision when determining what an AI system should be allowed to use. They first identify the decision, then the signals required to make it, the inputs behind those signals and, finally, the necessary data.
The approach avoids giving AI access to vast quantities of information simply because that information is available. It also provides a clearer basis for explaining why the system made a recommendation.
Murali wants those signals to tell planners what happened, why it happened, why it matters, what happened under similar conditions previously, and how confident the system is in its recommendation. People still make the judgment call, but they can do so with more context than they would have without the AI tools at their disposal.
Bala maintains that across these SAP projects, that may be the most practical AI playbook taking shape, where companies are beginning with the decisions and problems they need to solve, getting their data and planning processes in order, then strategically adding AI where it can make a measurable difference.
From there, the technology can take on a larger role if needed. But the companies moving down that path are also finding that the human side of the transformation doesn't become less important as AI gets more capable. If anything, deciding how people use the technology, and what decisions remain theirs to make, becomes even more important.
“It doesn't just replace the planner, it basically has to make them exponentially better,” says Bala.
Resource Link: https://www.sap.com/index.html
