Agentic AI Requires Deterministic Engineering Boundaries
At QCon AI NYC 2025, Aaron Erickson presented agentic AI as an engineering problem, arguing reliability requires combining probabilistic models with deterministic boundaries and operational telemetry. He highlighted practical mitigations — constrained schemas, deterministic templates, specialized agents, tool catalogs, and runbooks — and said these patterns improve safety and repeatability when deploying agent layers over real systems; full talk video releases January 15.
Key Points
- 1Frames agentic AI as engineering layer combining probabilistic models with deterministic execution and telemetry
- 2Warns that unbounded query generation and tool abundance increase errors and selection failures
- 3Recommends constrained schemas, deterministic templates, specialized agents, and runbooks for reliable deployments
Scoring Rationale
Practical, actionable engineering guidance with clear mitigations; limited novelty and based primarily on a single conference talk.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
