.jpg&w=1536&q=75)
Free Lesson
Decision Models: Cut the LLM Calls You Don't Need
Part of The 2027 Shortlist
45 min
Nov 11, 2026 10:00 AM
What you'll learn
Know which of your LLM calls should be a decision model
Work through the LLM call spans in a live agent: intent routing, tool selection, guardrails and more.
See the decisions worth handing off, with real numbers
Routing, triage, tool selection, guardrails replacement, with the cost and latency delta for each.
Run a model three ways: hosted, self-hosted, fine-tuned
Call a hosted decision API, self-host open weights locally, then adapt one to your traffic with confidence thresholds.
Why this topic matters
Decision models went from nothing to a benchmarked category in two months, and most of your stack's LLM calls only return a label or a score. Those calls run 10x slower and several times more expensive than they need to, and they give you no confidence number to act on. This lesson swaps a live agent's routing call for a decision model then sets a threshold and build in a fallback path.
You'll learn from
.jpg&w=384&q=75)
Aurimas Griciūnas
Founder @ SwirlAI • Ex CPO @ neptune.ai (Acquired by OpenAI)
