.png?width=275&height=77&name=HETT%20insights%20logo%20RGB-01%20(2).png "HETT insights logo RGB-01 (2)")
-Sep-22-2026-09-24-54-8459-AM.png)
Healthcare organisations already understand the importance of managed services. Networks are monitored. Security operations centres protect critical systems. Infrastructure is maintained and optimised. Identity and access controls are continuously reviewed.
As healthcare and NHS trusts do more with AI, it too increasingly needs the same (if not more) level of operational discipline.The NHS England rollout of Microsoft 365 Copilot to more than 500,000 staff is a major milestone. The headline saving of 43 minutes per person per day is impressive, particularly in an environment where time is one of the most constrained resources. But that is only part of the story.
The wider opportunity lies in what happens next, when AI starts to reshape the workflows that sit around those individuals. This is where personal productivity develops into operational AI.
Copilot can help someone summarise a meeting, draft a document, analyse information or find an answer more quickly. That is valuable, and it matters. But agentic AI takes the conversation further.
AI agents can support tasks, trigger workflows, coordinate processes, retrieve information, interact with systems and operate with increasing levels of autonomy. In the right context, and with the right controls, they can support the work that happens between people, teams and systems.
That is where things start to get really interesting. Much of the pressure across healthcare is not caused by one person being inefficient. It is caused by complex processes, fragmented systems, aging processes, every growing demand, duplicated administration and too many people spending too much time trying to connect the dots.
This is why AI needs a "managed services" mindset. Not because AI is “just another technology platform”, but because once AI becomes part of day-to-day operations, it needs to be monitored, governed, secured, measured and continuously improved.
As healthcare organisations look to build and deploy agents they will need visibility into performance, permissions, usage, security, cost, risk and value. They will need to understand what AI is doing, what data it is using, who owns it, how it is behaving and whether it is delivering the intended outcome.
In simple terms, once AI becomes part of the workforce, somebody needs to manage it.
From Personal Productivity to Operational AI
The first wave of AI adoption in healthcare has understandably focused on productivity. If clinicians, nurses, administrators, operational teams and support staff can save time on documentation, email, meetings, reporting and information retrieval, that time can be redirected into higher-value work.
But the bigger opportunity is what happens when AI starts to connect those individual gains into end-to-end workflows. This is where agentic AI starts to matter.
In finance, agents could support month-end reporting, budget analysis, variance checking, invoice processing and routine financial queries. In HR, agents could help employees navigate policies, onboarding processes, recruitment workflows, training requirements and case management. In procurement, agents could support supplier checks, contract reviews, renewal tracking and request management.
In IT and service management, agents could triage user requests, retrieve knowledge articles, resolve routine issues and escalate the right information to the right support team. In governance-heavy areas such as complaints, FOI requests and policy management, agents could help gather context, summarise information, check consistency and support faster response preparation.
None of this removes the need for people. It removes friction from the processes that consume people’s time. The goal should not be to automate for the sake of automation, but to help staff spend less time chasing information, repeating manual tasks or navigating fragmented workflows, and more time applying judgement, expertise and care.
Where This Starts to Show Up in the NHS
There are already signs of this shift happening across the NHS and wider healthcare ecosystem. The NHS App is a good example. AI-powered triage capabilities are being explored to help direct patients to the most appropriate service, whether that is a GP appointment, pharmacy, urgent treatment centre, community service or self-care guidance.
This is a very different use case from summarising a meeting or drafting an email. Here, AI starts to influence patient access, service navigation and demand management. That does not mean AI replaces clinical judgement. It means AI can help guide people more effectively through complex service pathways, provided the right safeguards, escalation routes and human oversight are in place.
Another example is the use of AI to identify patients at risk of becoming high-intensity users of emergency services. By analysing healthcare data, NHS teams can identify patterns earlier and offer more personalised preventative support before pressure lands in A&E.
In pathology, AI-assisted analysis is helping specialists review growing volumes of diagnostic information more efficiently, while still keeping clinical experts accountable for interpretation and decisions. These examples show the direction of travel: AI is moving beyond personal productivity and into the operational fabric of healthcare.
That is exciting, but it also raises the stakes. The more AI becomes embedded in real healthcare workflows, the more important it becomes to govern it properly, operate it safely and maintain trust.
The Trust Gap: Governance Alone Is Not Enough
Governance is important. But governance on its own is not enough. Healthcare is built on trust. Patients trust clinicians. Staff trust systems. Organisations trust processes. Regulators trust that appropriate controls are in place.
AI has to strengthen that trust, not weaken it. That means we need to move the conversation beyond whether an AI tool has been approved or whether a policy exists. The bigger question is accountability.
Every AI agent needs an owner. Not just a technical owner, but a business, operational or clinical owner who understands what the agent is there to do, what data it can access, what actions it can take, what risks it introduces and when a human needs to step in.
If an AI agent recommends an action, who reviews it? If it performs an action, who approved it? If it gives the wrong answer, how is that detected? If something goes wrong, who is accountable?
The answer cannot be “the AI made the decision”. Accountability always remains with people. That is especially important in healthcare, where decisions can impact access, prioritisation, care quality, patient communication and trust.
As AI becomes more capable, organisations will need stronger operating models around ownership, monitoring, audit, exception handling and escalation. In other words, AI governance needs to become operational governance.
Keeping Humans in the Loop
The phrase “human in the loop” is used a lot in AI conversations. In healthcare, it matters more than most. A human in the loop is not a token approval step. It is a safety net, a judgement point and often the trust point.
AI can summarise information, identify patterns, prioritise queues, suggest next actions and help prepare communications. But people remain accountable for judgement, empathy, clinical decisions, operational decisions and final outcomes.
A clinician remains accountable for clinical decisions. A manager remains accountable for operational decisions. An HR professional remains accountable for employee decisions. A governance lead remains accountable for policy and compliance decisions. AI can support those people, but it should not silently replace their accountability.
That means organisations need to be very clear about where AI can act independently, where it needs approval, where it can only recommend, and where it should not be used at all.
For example, AI may help draft a discharge summary, but a clinician should approve it. AI may help prioritise a queue, but a human should remain accountable for the decision. AI may help a patient find the right service, but escalation to a person should be easy when the patient needs it. AI may help summarise an HR case, complaint or FOI request, but a person should remain responsible for empathy, judgement and final action.
This is not about slowing AI down. It is about making sure AI is used in the right way, in the right place, with the right level of oversight.
Data Governance, Confidentiality and Patient Choice
Trust also depends on data. Healthcare organisations are custodians of some of the most sensitive information that exists. That means AI adoption cannot be separated from information governance, confidentiality, identity, access management, audit and data protection.
Organisations need confidence that AI systems only access the information they are allowed to access. They need to know permissions are properly managed, activity is logged, sensitive information is protected, outputs are grounded in appropriate sources, and data is not being used in ways that breach policy, regulation or patient expectations.
But there is another point that often gets missed: patient choice. Some patients may be perfectly comfortable interacting with an AI-powered service. Others may not be. Both positions are valid.
If a patient is interacting with AI, or if AI is materially shaping the service they receive, there needs to be transparency. Patients should understand when AI is being used, what it is being used for, what it can and cannot do, and where appropriate, they should have a clear route to a human.
Trust is not created by telling people AI is safe. Trust is created by showing how AI is governed, monitored, controlled and used responsibly. It is created by being transparent, respecting patient expectations, and ensuring people still feel cared for, not processed.
Building the Foundations for Sustainable AI
The healthcare organisations that appear to be getting the most value from AI are not simply the ones deploying the most tools. They are the ones investing in the foundations that make AI sustainable.
Those foundations include:
- Trusted and governed data
- Strong security and identity controls
- Clear ownership and accountability
- Human oversight and escalation processes
- Role-based access and least-privilege permissions
- Reliable infrastructure and networks
- Monitoring, audit and operational reporting
- Clinical safety and assurance where required
- Adoption, education and change management
- Patient transparency and trust
That means understanding the data landscape, the security posture, the governance model, the infrastructure, the adoption approach and the operational support model. Whether the goal is improving back-office efficiency, reducing administrative burden, supporting patient access or enabling more intelligent service delivery, the same principle applies.
AI must be designed, governed and operated as part of the wider organisation. Deploying it is only the first step. Managing it effectively is what delivers long-term value. This starts at the core of the organisation.
The Next Healthcare AI Conversation
The healthcare sector has made significant progress in understanding what AI can do. The next stage is understanding how AI should be governed, supported and operated once it becomes part of critical services.
The future conversation will not simply be about how many copilots or agents an organisation has deployed. It will be about whether those systems are trusted, secure and accountable; whether patients understand how they are being used; whether staff know when to rely on them and when to challenge them; whether the data foundations are strong enough; whether the human oversight is clear enough; and whether the organisation has the operational capability to manage AI safely at scale.
The organisations that succeed will not necessarily be those that adopt AI first. They will be the organisations that learn how to operate it responsibly. Because in healthcare, deploying AI is only the beginning.
Governance, management and accountability it is what ultimately determines its trust, value and success.
-Sep-22-2026-09-24-54-8459-AM.png)
Cisilion
%20(1).png?width=275&height=77&name=HETT%20insights%20logo%20RGB-04%20(1)%20(1).png "HETT insights logo RGB-04 (1) (1)")