Is your organisation ready to adopt an AI tool?

Introducing an AI system into healthcare is not simply a question of whether the technology works. The same system may be manageable in one organisation and difficult to use safely in another because local governance, infrastructure, workforce capacity and clinical oversight differ.

The NHS AI Adoption Assessment Tool and the AI Centre for Value Based Healthcare’s AI Deployment Toolkit provide a structured way for multidisciplinary teams to explore that organisational fit. Both resources are publicly available. The assessment tool is described by its developers as a prototype that is still being improved, rather than official policy or an authoritative decision-maker.

What does the assessment examine?

The online tool pairs 12 dimensions of AI complexity with 12 corresponding dimensions of organisational readiness. Complexity includes features of the technology, its data and its integration with existing systems. Readiness includes organisational capabilities such as governance, data infrastructure and capacity for clinical oversight.

The aim is to make gaps visible where the demands of a particular tool may exceed the organisation’s current capability. The assessment produces one of four broad outputs: a possible quick win, deploy with monitoring, build readiness first, or avoid at present.

These outputs should be treated as prompts for discussion, not as approval to procure or deploy a product. The developers state that the tool is diagnostic rather than prescriptive and is most useful early in horizon scanning, business-case development, procurement preparation or pilot planning.

Why multidisciplinary input matters

No single person is likely to hold all the information needed to assess a clinical AI system. The tool is intended for collaborative use by people involved in clinical safety, digital transformation, information governance, procurement, technical delivery and clinical practice.

For surgical teams, that collaborative approach is particularly relevant. A proposed system may affect clinical workflow, consent, training, escalation, documentation, cybersecurity and post-deployment monitoring. Different members of the team may identify different hazards or practical constraints.

An unanswered question should not be converted into a confident score. It is better treated as a prompt to obtain information from the appropriate local lead, supplier or specialist.

What does the deployment toolkit add?

The AI Centre’s wider toolkit links the assessment with editable implementation materials. The published list includes an AI governance lifecycle procedure and clinical safety materials such as a clinical risk management system, clinical risk management plan, hazard log and clinical safety case report.

This moves the conversation beyond selecting a product. It encourages teams to consider implementation, monitoring, benefits realisation and transition into routine use. However, the availability of a template does not establish that every necessary requirement has been met. Documents need to be adapted, reviewed and approved through the organisation’s own processes.

What the assessment does not replace

The assessment website explicitly states that it does not replace formal clinical safety work under DCB0129 and DCB0160, a Data Protection Impact Assessment, the Digital Technology Assessment Criteria, or any regulatory approval required for software that may be a medical device.

It also does not establish clinical effectiveness, cost-effectiveness, interoperability or suitability for a particular patient group. Those questions require appropriate evidence, local scrutiny and, where relevant, specialist regulatory, information-governance and procurement input.

The tool can search public sources for contextual information about a product, but its developers tell users to verify that information independently. Automated searches can miss evidence, misidentify a product or retrieve material that is incomplete or out of date.

Points for surgical teams to discuss

Before progressing an AI proposal, a multidisciplinary group could use the resources to structure questions such as:

• What clinical or operational problem is being addressed?
• What decisions or actions could the system influence?
• What happens when the system is wrong, unavailable or used outside its intended setting?
• Is the evidence relevant to the local population, pathway and workforce?
• Who will oversee outputs and respond to alerts or failures?
• Can the organisation support integration, training, monitoring and incident review over time?
• Which formal safety, data-protection, regulatory and procurement processes apply?
• How will patients and staff be involved, and how will benefits and unintended effects be reviewed?

The assessment should be repeated for each product under consideration. Organisational capacity is also cumulative: several deployments may place more pressure on governance, technical teams and clinical oversight than any single assessment suggests.

A useful starting point, with clear limits

These resources may help NHS teams turn a broad discussion about AI into a more transparent and auditable review. Their value lies in surfacing assumptions, gaps and areas requiring further scrutiny.

They should not be presented as an NHS-wide approval mechanism, a regulatory standard or proof that a technology is safe or effective. At the time of writing, the online assessment describes itself as a prototype and does not present independent validation or outcome evidence on its public page.

ASTI is sharing these resources for information and does not endorse the tools, their developers or any technology assessed through them. Decisions remain with the responsible healthcare organisations and professionals, using current official requirements and appropriately qualified input.