Free management resource

Workplace AI Risk Checklist

A practical 20-point check for managers and professionals before introducing, approving or expanding an AI use case at work.

AI risk depends on more than the tool. It also depends on the task, the information involved, who may be affected and what happens if the system is wrong. Use this checklist to identify gaps before an AI use case becomes routine.

How to use it: Check each statement you can confidently answer “yes” to. Every unchecked item is a control gap to investigate or resolve. A high score does not replace legal, privacy, information-security, HR, compliance, safety or other specialist review.

1. Purpose and potential impact

Define the use case and understand what could happen if it fails.

The business purpose, intended users and expected benefit are clearly defined.

The decision or activity supported by AI—and the parts that remain human-led—are explicit.

The consequences of incorrect, incomplete, biased or unavailable output have been assessed.

A lower-risk method has been considered where the potential harm is significant.

2. Data, privacy and security

Control what information enters the system and where it may go.

Only necessary, authorised information will be entered, uploaded or connected.

Personal, confidential, commercially sensitive and restricted data are handled appropriately.

Data retention, model-training, access and third-party sharing terms have been reviewed.

Required security, procurement, privacy and technology approvals are in place.

3. Accuracy and transparency

Make verification possible and avoid treating fluent output as proof.

Important outputs can be independently checked against reliable evidence or source material.

Known limitations, uncertainty and situations in which the tool should not be used are documented.

Users will know when they are interacting with AI or receiving AI-assisted material where appropriate.

Records are sufficient to explain the input, output, review and final decision when needed.

4. Fairness and people impact

Consider who may benefit, who may be disadvantaged and how concerns will be handled.

Potential bias, exclusion and unequal impact across relevant groups have been assessed.

The use case does not rely on AI alone for consequential employment, customer or access decisions.

Affected people have a practical way to question, correct or appeal an important outcome.

The effect on roles, workload, skills, trust and accountability has been considered.

5. Governance and ongoing oversight

Assign ownership and monitor the use case after launch.

A named owner is accountable for approval, operation, monitoring and corrective action.

Human review is meaningful, properly trained and proportionate to the possible consequences.

Performance, errors, complaints, incidents and unexpected impacts will be monitored.

There is a clear process to pause, escalate, modify or stop the use case when necessary.

Your readiness score: 0/20 Pause and assess the fundamentals.

17–20: Ready for formal reviewControls are broadly defined. Resolve remaining gaps and obtain required approvals.

12–16: Strengthen controlsImportant safeguards remain incomplete. Address them before proceeding.

0–11: PauseThe use case is not ready. Rework the fundamentals before approval or launch.

Important: This checklist is a practical management aid, not legal, regulatory, privacy, security, HR or safety advice. The appropriate review depends on the use case, jurisdiction, industry and consequences of error.

Frequently asked questions

Does a high score mean the AI use case is safe?

No. A high score indicates that common risk controls have been considered. It does not prove safety, compliance, accuracy or fitness for purpose.

Should every AI use case receive the same level of review?

No. Review should be proportionate to the information involved, the people affected and the consequences if the system performs poorly.

When should specialist review be required?

Seek qualified review whenever the use case involves sensitive data, regulated activity, employment decisions, safety, financial or legal consequences, vulnerable people or other material risk.

Build safer, more useful workplace AI practices

The Practical AI Toolkit for Managers & Professionals includes 500 AI prompts, workflows and decision tools for real work, together with worked examples, review tools, change leadership resources and practical frameworks.

Explore the Practical AI Toolkit

Prefer to start free? Get 50 Executive AI Prompts Every Manager Should Know.