Staying in control with Transparent AI
Uncertain supply chain conditions due to change and disruptions in supply and demand mark a new responsibility from the planner to leave the status quo and take the leadership to embrace technological innovations and more comprehensive data to guide their organization through this volatile environment.
This is why Logility, in every step of the development process, keeps in mind that the planner has to be the one in control, not the AI model. We understand that a seamless interaction between the user and the technology is essential to achieving this goal. We call this approach AI-first forecasting with a communication layer as the foundation that allows strengthening the interaction with transparent AI in a bi-direction way.
The planner explains events as he understands them to the system. For example, tell which sales spikes are related to which events and the system in return will also question for info on other spikes that could represent a similar event or could have a new unrelated cause to learn from.
The system tracks the decisions, learns from them, and makes them transparent, thus allowing planners to explain to other stakeholders what the decisions are built on.
In the past few months we have been building on this foundation, providing users with a clear view of the aspects that impact historical demand and the forecast:
- How has the sales baseline evolved over time?
- What is the impact of seasonality?
- What are the effects of promotions and events?
We have made these effects plain to see so planners can derive and act on the causal impact.
World events, competitor actions as well as promotional campaigns are run in different ways, and therefore, to capture intelligence that impacts demand succinctly, planners need to be able to interact. In this latest release, planners can add events at any level, to train the system and respond faster to -and more accurately predict the impact of- future disruptions and volatility.
An expanded performance screen allows you to compare the final forecast against your benchmark and user input.