Why should certificates for AI agent skills expire?

Because the thing being certified keeps changing. The product gains features, agents gain capabilities and new bounds, sources change, and the rules around oversight move. A certificate that never expires describes what someone could do on a version of the workspace that may no longer exist. Expiry forces the credential to say what is true today.

What changes between issue and expiry?

For a person who reviews agents' decisions, a year can bring:

• New capabilities. An agent that only drafted may now execute routine work, so the reviewer sees different held items.
• New bounds. The business may have narrowed limits or added new categories that are always held.
• New sources. A fact the agent used to receive as declared may now arrive sealed from a system of record, or the reverse.
• New failure modes. Teams learn from incidents. A pattern that slipped through once becomes a red line.
• New rules. Supervisors publish expectations; the business's own policy changes.

A reviewer certified before those changes may still be excellent. But the certificate cannot say so. Only a renewal that covers what changed can.

There is a parallel with agents themselves. An agent's certificate lapses after 90 days, and a certification run that clears no depth drops it to depth 1 the same day. The logic is identical: authority should rest on recent evidence. People change more slowly than agents, so their certificates last longer, but the principle holds.

Does expiry punish people who are already good?

Not if renewal is designed well. Expiry does not mean retaking the whole assessment. It means proving you have kept up. A short update module on what has changed, and a brief check that you understood it, is enough for most people. The heavy assessment, such as a practical period of real decisions, is done once and does not need repeating unless something has gone wrong.

The cost of not expiring falls on everyone else. Employers cannot tell a current practitioner from someone certified on a workspace they have not touched in three years. Clients checking a supplier's reviewers learn nothing from a "valid forever" status. And the holder loses the ability to show they are current.

An example: a contractor's certificate

Take a firm hiring contract reviewers for a peak period. One candidate holds a certificate with no expiry, issued long ago. Another holds a certificate due to expire in four months, renewed once already. The firm can see from the second that the holder completed the update on the agent's newer capabilities. From the first, it can see only that the holder passed something at some point.

The firm grants approval rights to the second immediately and asks the first to complete a current assessment. That is the practical value of expiry, and the reason verifying the certificates of contractors who review AI agents starts with the expiry date.

How should expiry be handled in practice?

• Show it publicly. A public register of certificates should show expiry dates and mark lapsed certificates as lapsed, not delete them.
• Remind early. Send reminders well ahead of expiry; the renewal reminder email template for AI agent certificates gives a sequence.
• Pause, do not punish. When a certificate lapses, approval rights that depend on it pause until renewal. The person's past decisions remain valid.
• Stagger renewals so a whole team does not lapse in the same week.

For the holder, a current certificate is a portable asset; whether AI agent certificates are worth it for individuals depends heavily on it being current and checkable. For employers, expiry is part of why AI certificates can be worth it: they can rely on status without re-testing.

Pryme Intelligence Academy sets professional certificate validity at 12 months for Operator and Builder and 24 months for Architect, with renewal as an update module and a short quiz for Operators. The wider structure is in the guide to AI literacy and certification for people who work with agents.