Grocery retail runs on thin margins and perishable stock, making demand forecasting one of the most demanding problems in the supply chain. Amar Pali has spent more than two decades on that exact problem. With over 21 years of industry experience across manufacturing, distribution, and retail, he develops supply chain forecasting models that pair deep operational knowledge with Agentic AI, helping grocery businesses move from prediction to action even as conditions change.
His focus sits at a difficult intersection. Fresh food spoils, shopper habits move quickly, and a single missed signal can mean a stockout on a Friday afternoon or a bin of unsold produce on Sunday. Pali builds forecasting systems meant to keep pace with that volatility rather than lag behind it.
Why grocery forecasting demands a different approach
A grocery shelf tells a complicated story. Demand shifts with the weather, the season, a local festival, or a sudden change in what shoppers want this week. Products expire. A forecast that looks right on Monday can be wrong by Thursday. Traditional statistical methods, built for steadier categories, tend to miss these fast, overlapping signals.
The grocery category also faces pressures most retail sectors never do. Perishability sets a hard clock on every decision. Promotions can double the demand for an item overnight, then leave a hangover of excess stock once the offer ends. Substitution muddies the picture further, since a shopper who cannot find one brand of yogurt will often reach for another, quietly reshaping demand across a whole shelf. Supplier lead times add their own constraints, so a plan made too late cannot always be corrected in time.
Pali designs his models around that reality. Rather than treating a forecast as a fixed monthly output, he approaches it as something that should update as new information arrives. That distinction matters most in fresh and short shelf-life categories, where a small error in either direction turns quickly into waste or empty shelves. A model that recognizes an early demand shift and acts on it protects both margin and availability.
How Agentic AI changes the forecast
Agentic AI describes systems that do more than produce a number. They interpret changing inputs, weigh competing signals, and recommend or trigger a response without waiting for a planner at every stage. Applied to grocery demand, a model can read shifts in weather, seasonality, local trends, and consumer behavior, then adjust its view of demand in near real time.
Older forecasting tools tend to work in a single direction. Historical data goes in, a projection comes out, and a human decides what to do with it days later. An agent-driven system closes that loop. It senses a change, evaluates what it means for a specific store and product, and moves toward a decision, checking its output against fresh data as the day unfolds. Planners stay in control of strategy and exceptions, while routine adjustments happen faster than any manual process could manage.
The practical shift is from prediction to action. A conventional forecast might flag that demand is likely to rise. An agent-driven approach can take the next step, adapting the plan as the signal strengthens or fades. For a category measured in days of freshness, that responsiveness is the difference between a useful forecast and one that arrives too late to help.
Turning forecasts into replenishment plans
A forecast only earns its keep when it drives the next order. This is where material requirements planning, the discipline of translating expected demand into purchase and production plans, becomes part of the story. A sharper forecast that feeds a clumsy replenishment process still leaves gaps on the shelf.
Pali’s work connects the two. His research on grocery forecasting pairs demand prediction with a more streamlined approach to material requirements planning, so the numbers a model produces flow directly into what a store or distribution center actually orders. In a peer-reviewed paper on advanced forecasting and streamlined material requirements planning for grocery, he examines how tailored methods can address the retail grocery sector’s specific consumption patterns. The emphasis throughout is on models that reflect how a business actually operates, not generic templates dropped on top of it.
Forecasting models built for real-world complexity
Pali’s background spans the full supply chain, from global manufacturing through distribution and into retail. That range shapes how he builds. His forecasting models are designed for flexibility and speed, so organizations can revise their plans as market conditions, consumer demand, and outside factors move.
Cross-industry experience tends to surface patterns that a single vantage point can miss. A manufacturer thinks in terms of production runs and raw materials. A distributor thinks in terms of lead times and network flow. A grocery retailer thinks in terms of shelf life and daily footfall. Pali’s approach draws on all three, helping his models account for how a decision in one part of the chain ripples through the rest. That perspective matters in grocery, where the distance between a supplier and a shelf can be measured in hours.
Speed of adaptation is central to the design. Markets rarely give clean warning before they move, so a model that can only be retrained on a slow cycle falls behind. Building for continuous adjustment keeps the plan aligned with conditions as they actually are, not as they looked when the last report was run.
What the shift means for grocery businesses
The aim of this approach is straightforward. Better signal reading is meant to help businesses anticipate demand more precisely, keep inventory aligned with real consumption, and cut waste from overstocking perishable goods. Stronger availability of the right products, at the right store, at the right time, follows from the same foundation.
There are wider benefits as well. Reducing overstock in fresh categories trims the volume of food that ends up discarded, which carries both a cost and an environmental weight for grocers operating at scale. A forecast that stays close to real demand supports leaner operations and less waste at once, without forcing a trade-off between the two.
For an industry where conditions rarely hold still, Pali’s view is that forecasting should be intelligent, adaptive, and tied directly to action. His broader goal is to make that kind of forecasting practical for businesses operating in a fast-moving global market, so the plan on the screen keeps pace with what is happening on the shelf.
