Capacity Forecasting in Agentic Delivery: Your Estimates Broke Before Your Pipeline Did
Agents did not make delivery unpredictable. They made your forecasting method visible. Point estimates and velocity assume a stable mean, and agentic work has a fat right tail. Here is why variance now dominates the plan, and the throughput, percentile, and Monte Carlo model that replaces story points.
Your roadmap is not slipping because the team got slower. It is slipping because you are forecasting with a statistic that assumes the one thing agentic delivery took away: a stable, narrow distribution of how long work takes.
Every delivery organization I talk to has the same pair of charts. Throughput is up. Predictability is down. Leaders read that as a discipline problem and respond with more refinement, tighter commitments, and a request for better estimates. That response makes the problem worse, because the estimates were never the broken part. The model underneath them was.
Members only
Keep reading for $2 / month.
The rest of this essay — along with every other members-only post, playbook, and working note — is behind a small paywall. Cancel anytime.
Payment handled by Stripe. Card details never touch this site.