Register and manage AI uses

Use the AI Register when you need to find an existing AI use, continue a draft or check whether a use has already been recorded.

Search by the business use rather than only by product name. The same product can support several different AI uses in the organisation.

Register an AI use when the organisation is considering, testing or using AI for a particular business purpose. Record the use itself—not just the product—so reviewers can understand who uses it, who may be affected and how it fits into the business.

Registration gives later assessment, risk, control and approval work a reliable starting point. You can save a draft when information is incomplete and return when the right person has confirmed it.

A useful registration explains the real business activity in language that a reviewer outside the team can understand.

Describe what the AI does, who uses its output, who may be affected and where a person remains involved. Avoid supplier marketing language or broad statements that do not explain the organisation’s use.

Recording the intended benefit makes it possible to judge later whether the AI use is helping and whether the result justifies its cost and risk.

Choose the expected improvements that genuinely apply and add a practical measure of success where one is known. This is business planning information, not proof that the benefit has been achieved.

Initial triage identifies facts that may require closer assessment or specialist input. It is an early screening step, not a risk rating or approval.

Answer from what is known today. Use I’m not sure when another person is better placed to confirm the answer rather than guessing.

Product and supplier information helps reviewers understand the technology behind the AI use and reuse relevant assurance information.

Choose a recognised product when it is an exact match. If it is not listed, internally built or not yet known, use the option that reflects the facts rather than choosing a convenient substitute.

The review step is your opportunity to read the registration as a complete business record before it becomes the starting point for governance work.

Correct anything that is misleading or incomplete. Saving a record does not mean the AI use is assessed, safe or approved; it means the information is ready for the next required process.

Drafts let you pause registration or editing without changing the published AI use.

Return to the draft when the missing information is available. Check the name and last-updated date so you continue the correct piece of work.

Use the ordinary edit process for factual corrections and routine updates that do not change the governance assumptions behind the AI use.

Use a material change instead when the purpose, technology, data, users, level of automation or another important fact changes enough to require review, reassessment or reapproval.

The governance lifecycle brings registration, assessment, controls, approval, monitoring and review into one sequence.

Each stage summarises the work recorded in Swell. Open a stage when you need to understand an outstanding requirement; do not assume that a later-looking stage means all earlier work is complete.

The operating model summarises how the AI use works in practice: who uses it, who receives the output, what decisions it influences and how people remain involved.

Review this information when assessing risk or when the real process changes. Correct the source information rather than working around an inaccurate summary.

Strategic alignment connects an AI use to the initiatives and priorities it genuinely supports.

Use these links to explain the delivery relationship, not merely to improve a report. One AI use can support several initiatives when that reflects the real work.

History provides an attributable record of important changes to the AI use and its governance work.

Use it to understand when something changed and who recorded it. Correct current information through the relevant workflow; history itself remains read-only.