Build better agents by testing them inside real workflows.
Wendell turns common business workflows into safe test worlds, so teams can see where their agents succeed, fail, and improve before they reach customers.
Research Digest
Agent readiness map
41
Seed companies
6
Workflow areas
30+
Reusable checks
24
Test situations
Pipeline
Workflow -> test world -> realistic situations -> agent run -> scorecard.
Voice Operations Agents
YC + a16z-backed teams are converging on appointment, clinic, finance, and support voice agents.
Product Output
Build a voice reliability pack
What the agent works with
Caller intent
Consent state
Calendar availability
Transfer rules
Call outcome
Situations we test
Book a qualified appointment
Handle no-consent recording
Transfer urgent patient request
Reject unsupported price guarantee
How we score it
Consent before restricted action
Correct tool use
Escalates when policy requires it
Clear call summary
How the test works
Each workflow becomes a controlled test world with people, systems, rules, failure cases, and a scorecard.
Understand the workflow
Capture what the agent touches, what is not allowed, and what failure looks like.
Build the test world
Turn the workflow into realistic situations where the agent must make decisions.
Run, score, and improve
Run the agent, review the scorecard, improve it, and compare the next run.