The goal is not AI. The goal is capacity: Lessons from the NHS
What happens when AI moves beyond the pilot and into everyday work across one of the world’s largest health systems?
In Sognos’ recent Redesigning Work Around Care webinar, John McGhie, Head of NHS.net Connect at National Health Service (NHS) England, shared lessons from the NHS Microsoft 365 Copilot journey, which involved more than 30,000 staff across 90 NHS organisations in its large-scale trial and has since expanded to around 500,000 licences across the NHS.
The scale alone makes the NHS experience worth paying attention to. But some of the most useful lessons had less to do with the technology itself and much more to do with people, governance, workflows and what happens once the pilot is over.
Forty-three minutes.
It is an impressive number. During the NHS Microsoft 365 Copilot pilot, users recorded a conservative average saving of around 43 minutes per person, per day across everyday activities such as managing email, summarising information and creating documents.
But during Sognos’ recent Redesigning Work Around Care webinar, John McGhie, Head of NHS.net Connect at NHS England, kept bringing the conversation back to a more important question:
What do you do with those 43 minutes?
If the time saved does not create additional capacity, reduce a wait, remove administration or allow people to work differently, productivity alone is not the outcome healthcare needs.
As John put it, the goal is not AI. The goal is capacity.
AI adoption is a workforce change
One of the earliest lessons from the NHS was not to treat AI like a conventional technology rollout.
AI changes how people work, so the people creating and using those workflows need to be involved from the beginning.
That meant working with business teams, HR, clinical safety, governance and frontline services, not simply deploying technology through IT.
And adoption takes work.
The NHS has invested in training, champions, peer communities, Copilot Cafés and regular opportunities for staff to learn from each other.
John’s message to leaders was equally clear. AI cannot simply be announced and delegated. Successful adoption needs visible leadership, managers who understand the technology and people who are prepared to question whether work could be done differently.
Governance enables adoption
Healthcare organisations cannot separate AI from privacy, security, data governance and clinical safety.
John described the considerable readiness work undertaken by the NHS around information protection, authentication, data loss prevention and clinical governance.
He also raised a challenge already familiar to many organisations: shadow AI.
People may already be using consumer AI tools whether their organisation has a formal strategy or not. For the NHS, that strengthened the case for providing enterprise AI within an environment where appropriate controls could be applied.
At times, the NHS has deliberately slowed implementation when governance or readiness needed to catch up.Responsible adoption means knowing when the foundations are strong enough to move forward. It means knowing when the foundations are strong enough to move forward.
Do not put AI on top of a process that already does not work
One of the strongest themes from John’s presentation was the need to rethink the workflow itself.
AI can make an individual task faster, but that does not necessarily mean the task should continue to exist in the same form.
Where has unnecessary administration accumulated? Which processes are repetitive? Where are multiple teams solving the same problem? What could be done once and reused?
The NHS now has close to 1,500 documented AI use cases, ranging from everyday information work to more specialised healthcare applications.
But the bigger opportunity is not simply doing the same work faster.
It is redesigning the work.
“It’s Copilot, not autopilot”
Healthcare also has to navigate a critical boundary between AI supporting clinical administration and AI influencing clinical decision-making.
John spoke about the work the NHS has undertaken to define that boundary and the importance of clinical safety and individual organisational risk assessment.
His description was simple:
It is Copilot, not autopilot.
The healthcare professional remains the subject matter expert. AI can support administration, information and workflow, but human expertise, judgement and accountability remain central.
From minutes saved to real ROI
The NHS is now moving beyond measuring individual productivity gains.
The next question is whether those gains can translate into capacity, workflow improvements and financial value.
What does a workflow cost today? How many people are involved? How much time does it consume? If AI changes that workflow, what capacity does it release?
Those questions become even more important as organisations move towards AI agents and consumption-based models.
The economics of AI will increasingly depend on understanding the value of the work being changed, not simply the cost of the technology.
Partners matter
John was also candid that NHS England has not done this alone.
Microsoft and a broader ecosystem of specialist partners have supported different stages of the journey, bringing expertise in technology, implementation, governance, workforce adoption and healthcare workflows.
One of those partners is DigPacks, a UK healthcare technology provider whose Microsoft-based solutions have been shaped through work with NHS organisations. DigPacks focuses on practical operational challenges that often sit around core clinical systems, including administration, workforce processes, risk, governance and workflow automation.
That experience is now part of Sognos’ strategic partnership with DigPacks in Australia and New Zealand.
The value for Australian and New Zealand organisations is not in trying to replicate what the NHS has done, because our health, aged care, disability and community care sectors operate within different funding models, regulatory settings and service environments. The real opportunity is to learn from what has already been tested at scale, understand which approaches have delivered value, and then adapt those lessons to the realities of care delivery here.
What does that mean locally?
For an Australian or New Zealand healthcare, aged care, disability or community care organisation, the starting point does not need to be a large AI program.
It can begin with a much more practical question:
If your people saved 43 minutes a day, where would that capacity go, and who would decide?
Would it reduce time spent on administration? Help a team manage more referrals or appointments? Give frontline staff more time with the people they support? Remove steps from a process that has become unnecessarily complex?
And then comes the second question: what would need to change in the workflow to make that benefit real?
That is where John’s NHS experience becomes particularly relevant locally. The opportunity is not simply to introduce AI into existing work. It is to identify the work creating the greatest burden, decide what better could look like, and then determine whether AI has a useful role in getting there.
Watch the webinar on demand
There was much more in John’s discussion than we could cover in one article, including practical NHS examples, shadow AI, agents, clinical safety, ROI, workforce adoption and a broad range of questions from healthcare leaders attending the session.
Not sure where to start with AI?
Sognos and DigPacks are also offering healthcare organisations a personalised two-week AI Solutions Assessment.
We work with your team to understand your current environment, identify high-friction processes and prioritise practical opportunities for AI, Copilot and agents.
The assessment considers productivity, workflow redesign, workforce capacity, governance, security, organisational readiness and high-value use cases.
You will leave with a clearer picture of your current state, prioritised opportunities and practical next steps.
Microsoft funding may be available for eligible organisations, subject to assessment and approval.