a builder's codex
codex · operators · Claire Vo · ins_progressive-trust-onboarding

Onboard agents the way you onboard an EA: progressive trust, named tiers

By Claire Vo · 3x CPO; founder of ChatPRD · 2026-04-28 · podcast · Claire Vo on running 9 AI agents — Lenny's Podcast

Tier A · TL;DR
Onboard agents the way you onboard an EA: progressive trust, named tiers

Claim

Stage tool access for AI agents in named tiers (read calendar → read mail → draft mail → send mail), the same way you'd ramp a new executive assistant. Don't hand over full account access on day one.

Mechanism

Trust ramps mitigate two failure modes at once: alignment (the agent learns your style on low-stakes work before it does anything irreversible) and security (a compromised agent has bounded blast radius at each stage). Treating provisioning as a ladder forces explicit decisions about what each agent should be able to do, instead of one big "give it admin" moment.

Conditions

Works when:

Fails when:

Evidence

"You don't onboard your EA by giving the password to your email account... they have their own email, they have their own calendar, and you give them access or permission."

"First you get my calendar, then you can read my email, then I guess you could draft some emails, then you can send the emails."

— Claire Vo on Lenny's Podcast, 2026-04-28

Signals

Counter-evidence

No opposing view in current corpus.

Cross-references

Open the interactive view → View original source → Markdown source →