How do you ensure AI agents' work is trustworthy?

Michael Jenkins

C.T.O. · Threatlocker

Peer Review

Michael Jenkins says he uses peer review, arguing AI-written code should be reviewed by somebody else just as a developer's code is before production.

Ajay Krishna Amudan

Cofounder, CTO · Maximor Ai

Prove Correctness

Ajay Krishna Amudan says proving your work is correct is a big part of doing the work, and that not treating that as a first-class citizen can cause issues six months to a year down the line.

Rhett Alden

CTO, Elsevier Health Markets · Elsevier

Robust Validation Framework

Rhett Alden says his team has a robust validation framework for CKAI and publishes results on errors and hallucination risk to provide grounding for clinicians.

The answer starts at 3:59.

Angie Jones

VP Engineering, AI Tools & Enablement · Block

Recursive Adversarial Review

Angie Jones says she runs a recursive adversarial review after anything she builds, having a fresh sub agent review the diff without context, because an agent checking its own work is biased and approves it.

Daniel Herrington

Chief AI Officer · The Zebra

Human Review

Daniel Herrington says that for anything using customer data, his team still has a human reviewer take a second look, even with an automated reviewer, to confirm it's not sending data out and is still on track.