Responsible Use of AI Toolkit

Before you start

1. 01 Principle 1: Human oversight of AI If this goes wrong, can it be undone, and who would it affect?
2. 02 Principle 2: A single source of truth for AI Where does this fact live, and who keeps it current?
3. 03 Principle 3: Confidentiality of information shared with AI Is this mine to share, in this tool?
4. 04 Principle 4: Limited permissions for AI agents What’s the worst this agent could do with the access it has?

While you work

1. 05 Principle 5: Independent judgment and skills What do I think?
2. 06 Principle 6: Challenge and bias awareness What’s the strongest case against this?

Before you share

1. 07 Principle 7: Verification proportionate to the stakes What would it cost if this were wrong?
2. 08 Principle 8: Quality of AI-assisted work Would I want to receive this?
3. 09 Principle 9: Transparency about the use of AI Would they feel misled if they knew how this was made?
4. 10 Principle 10: Accountability for AI-assisted results Am I ready to put my name to it?

By kind of work

• Writing4
• Meetings2
• Research4
• Data and numbers2
• Contracts and policies3
• Customers3
• Sales and pricing2
• Marketing and images2
• Email and agents1
• Decisions and plans3
• Learning1
• Operations and projects2
• People and hiring1

One example for each moment

• Principle 1 · Customers Refunds in the support queue Before: “Answer the complaints in the support queue and refund anyone who qualifies.”
• Principle 5 · Decisions and plans Setting a price Before: “What price should we set for the new product?”
• Principle 7 · Research Writing with statistics Before: “Write a paragraph on the growth of the European AI market, with statistics.”
• All prompt examples

• Principles 1, 4, 7 Your agent is ready to send 40 emails An agent has drafted personalized follow-ups to a webinar list and asks whether to send them all.
• Principles 2, 7 The assistant quoted last year’s price A client needs a price fast. The assistant answers in seconds, from a presentation you suspect is out of date.
• Principle 3 You want AI to review a client contract A long contract, a deadline and a free AI tool one click away.
• Principles 4, 7 An email asks your agent to change a supplier’s bank details Your inbox agent wants to update a payment record because an email told it to.
• Principle 7 The finance director asks for an urgent transfer on a video call A familiar face, a familiar voice and an urgent, confidential request.
• Principles 7, 10 The report cites a study you can’t find A polished AI-assisted report, a deadline and one reference that doesn’t seem to exist.
• Principle 6 An assistant that agrees with every idea Three weeks, three versions of a plan and a lot of praise. Something feels too easy.
• Principle 8 A six-page AI summary of a 30-minute meeting A colleague means well. The summary is longer than the meeting was.
• Principle 1 You’re asked to rank job applicants with AI Two hundred applications, one week and a suggestion from above.
• Principle 9 Your manager asks whether you wrote the report yourself An honest question about a report you drafted with AI.
• Principle 10 You spot a wrong figure in a report you sent last week Page 4 says 21%. The source says 12%.
• Principle 5 A junior colleague’s analysis looks perfect, but they can’t explain it Clean charts, confident conclusions and no clear idea of how they got there.
• Principles 3, 9 An AI note-taker joins your client call Someone’s assistant is recording, and nobody asked.

• Tribunal decision · February 2024 Air Canada: liability for a chatbot’s answers A tribunal ruled that an airline was responsible for what its website chatbot told a grieving customer.
• Refund reported by the press · October 2025 Deloitte Australia: a report with invented references A consulting firm repaid part of its fee after a government report was found to contain invented references and a made-up court quote.
• Court decisions · 2023 to 2026 Invented case law in court filings Lawyers were fined for citing cases an AI chatbot invented. A public database now lists more than 2,000 similar court and tribunal decisions.
• Police case and company confirmation · January 2024 Arup: deepfake fraud on a video call An employee transferred about HK$200 million after a video call with fake versions of senior colleagues.
• Research · 2025 to 2026 METR: perceived and measured productivity with AI Experienced developers felt 20% faster with AI, but their tasks took 19% longer. A 2026 follow-up pointed the other way, and METR calls its results unreliable.
• Survey research · September 2025 Workslop: the cost of unedited AI output In a survey of 1,150 US employees, 40% had received AI-generated work that looked polished but didn’t do the job.
• Government evaluation · 2025 UK government: human review of AI-sorted consultation responses An AI tool sorted public consultation responses into themes. People checked every one of its decisions, and the results were published.
• Government evaluations · 2025 UK government: evaluating an AI assistant across departments A large trial found time saved and people who wanted to keep going. A smaller evaluation found that results varied by task.
• Our own practice · 2026 Founderz: one source of truth for people and AI Every document has an owner, a version and a date. AI assistants read the same knowledge people do, and drafts wait for a person.

• Session plan Adoption workshop A one-hour session plan for adopting the standard as a team: the definition, the ten key questions, the team charter and three prompt examples to try in the following week.
• Template Team AI charter One page that records how your team applies the ten principles: approved tools, data classification, checkpoints, sources, agents and who to contact when something goes wrong.
• Template Human checkpoint map For each AI task your team repeats: the level of oversight, who approves what, what they see and how to stop and undo.
• Template Single source of truth starter kit Give your team’s key facts one home, one owner and a date, so people and AI find the same right answer.
• Template Agent permission card One card per shared agent: what it’s for, what it can reach and do, what needs approval, its limits and how to stop it.
• Template Verification levels Agree which outputs get a glance, which get spot checks and which get full verification, so everyone checks in proportion.

• Glossary 32 terms, from AI agent to workslop, each with a plain definition.
• Sources 48 sources, each with what it establishes and its limits.
• Where the law fits The standard describes good practice, and the law sets the floor. Here’s where the law most often meets everyday AI use at work.
• About the RUAI Standard Who publishes the standard, how it’s made and maintained, how to cite it and how to help.
• The standard in Markdown The full text in one file, for AI assistants, search tools and anyone who wants to reuse it.