Availability: Standard, Pro and Enterprise, and every trial.
Your AI policies are what the organisation has agreed about using AI at work: which tools, what information can go into them, and what needs a person's judgement. In The GAiGE you record each policy with a reason staff can read, publish them together as a numbered version, and see who has confirmed reading it.
Where policies come fromLink to this section
You rarely start from a blank page.
- The starter policies. A small set that is common practice, such as checking anything AI helped produce before it leaves the organisation, and who owns AI-made work. Governance set-up offers them under Agree the organisation-wide policies. Accept or set aside each one.
- Proposed policies. Once The AI State of Play closes, Governance set-up proposes policies from your people's answers, beside the ones the leadership position implies, under Agree the organisation-wide policies. Each comes with a recommendation, its reason, and what it's based on.
- Your own. Record any policy yourself (below).
- The organisation's own document. If you already have a written AI policy, Governance set-up asks for its name and a link. Staff see it linked from Our AI policies as the Related document.
Recording or editing a policyLink to this section
Go to Governance → Policies and choose Record a policy, or open a policy to edit it. The fields:
- Heading: the group it appears under in Our AI policies: Tools, Information, Work you share or Decisions a person makes
- Title: two to four words naming the task or the information, such as Client information
- The policy: one sentence to staff, saying what they can do before what they cannot
- Reason for staff: what the policy protects, such as clients, colleagues, the organisation's name, a promise or a law. Its first sentence also appears under At a glance
- In practice: up to three examples from the work, each marked Yes, Ask first or No
- Types of work, Types of information and Tools: what the policy covers, so staff can filter for what applies to them
- Applies to: Everyone, or Chosen groups (a group includes its sub-groups), with Who it covers for anything the groups don't say
- Reasoning: shown under Decisions, and never to staff
- Date decided and Review by: we remind you when the review date passes
Show the staff view shows the policy as staff will read it.
Publishing a versionLink to this section
Changes don't reach staff one at a time. They wait for the next version, and the Our AI policies card at the top of Policies says how many changes are waiting.
The first version, in Governance set-up. Accepting a policy under Agree the organisation-wide policies puts it in Our AI policies for the next version; to take one out again, use the In Our AI policies switch on Policies. Check and publish Our AI policies then shows the next version as one document, the way your people will read it, with a link from each part to where you change it, and Download the draft PDF, marked as a draft on every page, for anyone checking it with you. Once you've read it, Publish version 1, under the document, publishes it, in the same steps as below.
- Choose Publish version with the next number, for example Publish version 3.
- Check Changes in this version. If anything needs fixing first, it's listed under Before publishing, with a link to put it right.
- Write a Summary: one line staff see for this version.
- For each changed policy, check What changed: one line staff see on the policy for 90 days, without naming anyone.
- Use Preview version 3 if you'd like to see it as staff will, then Publish version 3.
Your people see nothing in Our AI policies until you first brief the team (below): version 1 reaches them with that brief, and each later version as soon as it's published. Either way, they're asked to confirm they have read it. Earlier versions stay listed under Earlier versions.
Two more things on the card: the Contact for problems (who staff tell when information has gone somewhere it shouldn't, and how to reach them), and Download the PDF.
Telling peopleLink to this section
Publishing makes the policies available; telling people is a separate step, Brief the team, in Governance set-up, in three parts:
- Settings shows the version it goes with, and the Contact for problems the message gives. You read Our AI policies as your people will in Check and publish Our AI policies, before anyone is told.
- The message is drafted for you, as The policies we have agreed. Edit its subject, its opening, why the organisation is doing this, and its last paragraph. The policies at a glance and who to tell about a problem are filled in from the version, so they always match what you published. In the first brief after The AI State of Play closes, the message opens by thanking people for their answers.
- Send to everyone emails everyone on People. It goes once for each version, and sending can't be undone.
The message then shows in Activity, and in each person's AI Adoption Portal. When you publish a later version, Brief the team offers a new message for it.
What your people seeLink to this section
In their AI Adoption Portal, and in the browser extension, people find Our AI policies:
- the policies that apply to them, under For you, with All policies one click away
- each policy's reason and examples
- Read the full document and Download the PDF
- I've read these, to confirm. The page tells them that their organisation can see who has confirmed reading each version.
When a new version is published, they're asked to read what changed and confirm again. People confirm the policies that apply to them, so a change to an engineering policy doesn't ask the marketing team to confirm again.
Seeing who has read themLink to this section
- On the Dashboard, the Policies read figure under How it's going, for example 63% and "38 of 60 people". If fewer than 80% have confirmed 14 days after the version was published or the team was briefed, a row in Needs you says so.
- In Reports, Policies and questions shows the current version, how many have read it, and Who has confirmed: each person with the date they confirmed, or Not yet.
- Send a reminder, on the report or on that row in Needs you, emails everyone with an account who still needs to confirm the policies that apply to them, at most once a day. You see how many people it reached.
The same report lists the Questions asked about what's allowed, and which policy or tool decision answered each one. If nothing answers a question, consider writing a policy for it.
Keeping them currentLink to this section
Each policy has a Review by date, and the Review lists decisions as they fall due. When you change a policy, it waits for the next version, so staff always see a numbered, dated set.
For the decision on each tool, which staff see beside the policies, see Recording a decision on an AI tool.
Something missing or out of date here? Tell us.