Trust, Not Technology, Will Decide Whether AI Succeeds in Healthcare

One of the privileges of my career has been seeing healthcare transformation from multiple perspectives. I've been the clinician completing documentation after the patient interaction. I've been the operational manager trying to create capacity with limited resources. I've been the CIO and CCIO balancing innovation with clinical safety, risk and public accountability. Today, through my work at CDW UK and my wider involvement across health and care, I spend a great deal of time speaking with NHS leaders about artificial intelligence. 

What strikes me is that very few of those conversations begin with the technology. They begin with uncertainty. 

I think these are exactly the right questions. In healthcare, caution is not a sign that people are resistant to change. Caution is part of how we protect patients. I would be far more concerned if clinicians accepted every new technology without question. 

The future of AI in healthcare will not be determined by capability alone. It will be determined by whether the people expected to use it trust it enough to make it part of everyday care. 

AI is already being used in documentation, diagnostics, scheduling, triage, operational planning and patient-facing services. But access to capable technology does not automatically create adoption. 

During my time as an NHS digital leader, I saw excellent technology struggle because people had not been involved early enough. I also saw less mature technology succeed because teams understood the purpose, could see the benefit and felt supported through the change. 

If a clinician does not trust an AI-generated note, the clinician will rewrite it. If a recommendation cannot be understood or challenged, it may be ignored. If a board cannot see how accountability works, scale will stall. The most sophisticated system creates little value if it is not used consistently and responsibly. 

Confidence grows when people are involved, when limitations are discussed openly and when leaders are visible about where responsibility remains. Trust is not a communications activity at the end of deployment. It is designed into the programme from the beginning. 

Great Ormond Street Hospital-led London study 

The NHS England-sponsored evaluation of ambient voice technology covered nine London NHS sites and more than 17,000 patient encounters. 

• The study reported 23.5% more direct patient interaction time and an 8.2% reduction in appointment length.
• A&E sites reported a 13.4% increase in patients seen per shift.
• Patients and families responded positively, with 92% consenting to use of the AI scribe.
• Clinicians remained responsible for checking and editing the generated note before it was saved to the electronic patient record.
• GOSH states that detailed governance and safety checks were completed before the technology moved into routine use. 

For me, the most important lesson is not simply that the tool produced measurable benefits. It is that consent, clinical review, safety checks and clear information for patients were treated as part of the deployment, not as an afterthought. 

This is what trusted adoption looks like. The technology supports the consultation, while professional judgement and accountability remain visible. Patients understand what is happening. Clinicians retain control. The organisation can explain how safety, information governance and data use have been addressed. 

One of the misconceptions I encounter most often is that governance slows innovation down. My experience tells me the opposite. Poorly designed governance creates delay, duplication and late surprises. Good governance gives people the confidence to move. 

When clinical safety, information governance, cyber security, data protection, procurement and operational leaders are engaged early, organisations can identify risks before those risks become blockers. When the board understands the risk appetite and the workforce understands the guardrails, adoption becomes easier. 

AI governance cannot sit solely with the technology team. It needs multidisciplinary ownership. The governance model should be clear about how AI is selected, validated, deployed, monitored, changed and, if necessary, withdrawn. It should also make clear who is accountable for the output, how concerns are raised and how patients and staff are involved. 

For NHS organisations, established clinical safety and information governance requirements still matter. Depending on the use case, this may include Clinical Safety Officer oversight, DCB0129 and DCB0160 considerations, Data Protection Impact Assessments, supplier assurance, cyber controls and post-deployment monitoring. 

The board conversation has matured. The question is no longer simply, ‘What can AI do?’ It is, ‘How do we adopt it responsibly and demonstrate that responsibility?’ 

These are not product questions. They are leadership questions. Trust cannot be delegated to a vendor or a project team. It starts with the organisation being clear about its purpose, responsibilities and appetite for risk. 

At CDW UK, our most useful AI conversations rarely begin with a product. They begin with the organisation's ambition, the problem being addressed and the confidence needed to move forward. 

The support can bring together the different disciplines that need to be part of responsible AI adoption. Examples include: 

1. AI readiness and use-case discovery 

Helping leadership teams move from broad AI ambition to a prioritised set of use cases grounded in patient, workforce, operational and financial outcomes. 

• Clarifying the problem and intended benefit before technology selection.
• Assessing digital foundations, data quality, interoperability, identity, cyber resilience and workflow readiness.
• Identifying dependencies and risks that could prevent a pilot from scaling. 

This keeps the conversation focused on outcomes, not novelty.