By Shamila Shalehin — May 30, 2023

Forecasting for Success: Unleashing the Potential of Adoptable AI

Are you wondering how Granularity implements artificial intelligence to enhance demand forecasting?

We explain how to do this to create transparent and adoptable forecasts.

Check out our use case here or below.

Forecasting with artificial intelligence: How to build AI forecasts that are transparent and adoptable

Hi, we're Granularity. Granularity is an AI Demand Forecasting startup focused on bringing the best demand forecasts to planners. We're made of data scientists, engineers, and certified forecasters.

Building AI tools with humans in the loop. Human-Centred MLOps: accelerate model adoption by implementing core MLOps principles, fostering collaboration, and appointing a dedicated champion for transformation. Explainable ML/AI: increase user trust and engagement by enhancing the transparency of AI through interpretable machine learning techniques and comprehensive user education.

Accelerating Model Adoption with MLOps and Human-Centric Integration. With effective change management and MLOps, the ML team worked with the planning team to boost forecasting accuracy by 35% and improve decision-making.

Process improvements include champions to lead engagement, user-centered after model creation, and end users engaged with training and enhancements.

Shared Ownership: sprints are open with business users on backlogs and priorities. Direct Model Feedback: User responses to data and decisions are inputs to the model. Build Experiences: Get the model to users. Enable experimentation and sandboxes. Create the Foundation: A working ML/AI model that meets initial assumptions. Process improvements include building the foundation of working ML/AI solutions, getting user feedback on initial models, and iterating on the model with transparent sprints and backlogs.

Boosting Model Adoption through Enhanced Transparency in Demand Forecasting. With enhanced model explainability, the ML team was able to achieve seamless integration of AI-driven demand forecasting into business processes.

Share metadata and feature importance or partial dependencies to users, allow for user simulations of variables in dashboards and create shared documentation and process information. Explainability improvements include deploying interpretations to all users, simulating the effects of changing variables, and over-documenting and enabling everyone.