What is risk literacy?

Risk literacy is the ability to understand and weigh risk in a decision: how likely a harm is, how severe it would be, and how easily it could be reversed, and to act in proportion. A risk-literate person does not treat every decision as dangerous or every decision as safe. They know where to slow down.

Why does risk literacy matter for agents?

Agents handle a great deal of routine work and some consequential work. The people overseeing them have limited attention and must spend it where it matters. A reviewer who gives a change of email address and a change of bank account the same few seconds has misjudged the risk, even if both drafts look tidy.

Reversibility is the dimension most often missed. A wrong reply to a customer can be corrected. A payment sent to a fraudster may be gone. Depth reflects this: depth 4 covers moving money and actions that are hard to undo, and those carry a higher bar for both the agent and the person approving.

How does it show up in practice?

A risk-literate reviewer asks three questions of each held item:

1. What could go wrong if I approve this and I am wrong?
2. How likely is that, given the evidence and its provenance?
3. Can it be undone, and at what cost?

The answers set how much time the item gets. A low-value, reversible action with sealed evidence may take seconds. A large, irreversible one with external evidence deserves a call to the customer on a known number and perhaps a second approver. Risk literacy also means accepting refusals: when an action is withheld at the platform floor, such as reversing a settled payment on the platform's own rail, a literate person understands why it is never left to judgement in the moment.

What is an example?

An insurer's agent drafts claim payments. Two land in a reviewer's queue together: a small payment to a policyholder's long-standing account, and a larger payment to a repairer whose bank details were added last week from an emailed invoice. A reviewer without risk literacy works through them in order at the same pace. A risk-literate one approves the first quickly, then spends real time on the second: checks the repairer against signed records, calls on a known number, and holds it when the details cannot be confirmed. Insurance teams face this often; see Article 4 literacy for agents in insurance.

How is risk literacy different from data literacy?

Data literacy is about evidence: where it came from and how far it can be trusted. Risk literacy is about consequences: what happens if a decision based on that evidence is wrong. The two combine in every good review. Weak evidence for a reversible action may be fine; strong evidence for an irreversible one still deserves care. Both sit within the broader AI literacy a role needs.

Which terms are related?

• Reversibility: how easily an action can be undone.
• Depth: how far an agent may act, which tracks risk.
• Literacy policy and records: set and evidence the skills each role needs; see the AI literacy policy template and training record template.

For the wider topic, read AI literacy and certification for people who work with agents.