AI Agent for Shock Resilient Demand Planning

Do you have a demand forecasting model whose accuracy keeps deteriorating with every major change in an influential external signal?

Enter DeepVu's shock-resilient demand-planning AI agents solution for Dynamics365!

DeepVu's Resilient Demand Planner solution is based on multiple AI-decisioning agents that are trained on top of multi-scenario digital twins that simulate normal and shock scenarios. Therefore, you get a normal-scenario trained agent, an agent trained on a specific shock, and additional agents training on other shock scenarios. You get to choose which agent and which recommended demand plan to deploy depending on the context.

Our system also takes into account an extensive set of external signals that represent world and industry context from our supply chain Knowledge Graph (VuGraph). There is still room for forecasting models that are part of this decisioning system to offer the baseline default forecast, but they don’t do the heavy lifting in terms of adapting to the external shocks and revising the demand projection accordingly; the AI decisioning agents do. These AI agents then recommend decisions and predict the impact on business key performance indicators (KPIs). Human planners will choose which agent is most applicable and choose the recommended decision from the selected AI agent to deploy as is or override, fully informed by its KPI impact.

The entire system is always on, continuously learning and self-tuning based on usage and feedback. It’s important to note that the external signals have different velocities: Commodity prices are available on every trading day, and government data (interest rates, unemployment rate per state, CPIs, PPIs etc) follow a monthly cadence, for example.

Our app comes with some shocks pre-built, such as consumer spend (PCE Personal Consumption Expenditures) and a few commodity categories, but it's a continuous exercise to keep adding shocks based on customer requests as we scale this solution.