Agentic AI Forecasting

I led design for an AI-assisted forecasting platform for Pfizer's Cibinqo brand team, in partnership with ZS. It brings the full forecast cycle, from new actuals to approved scenario, into one workspace.

The Challenge
Each cycle, analysts reconciled new actuals, adjusted dozens of linked assumptions, and defended every variance to leadership. The goal was faster forecasts without taking judgment away from the people accountable for the number.

Key Decisions
Conversation instead of forms. Analysts adjust assumptions by talking to an AI assistant, then review proposed against current values before anything is applied.
Every number explains itself. The system shows why a figure moved, so analysts can defend it to leadership.
AI proposes, people approve. Nothing changes the forecast without sign-off.