AI in healthcare: exciting possibilities, responsible
progress
Reflections from the AI Ambassadors Network Conference, 11
November 2026
Maria Hughes- Knowledge and Evidence Specialist
After a year or so of joining events online, I was excited
to attend the first in person AI Ambassadors Network Conference in London. It
was a chance to meet colleagues from across the NHS, exchange ideas and bring
some of that learning back to MTW.
I came away with plenty of enthusiasm, but also a clear reminder: making AI useful in healthcare is about much more than introducing new technology. It is about the problems we want to solve, the people using it and the patients it should benefit.
Start with the problem, not the technology
One message from the presentation by Rob Thompson, Chief
Digital, Data and Technology Officer for the Department of Health and Social
Care and NHS England, really stayed with me: try an idea, see whether it helps,
and learn from the result. If it saves time or resources, or improves a
process, that is worth exploring. If it does not add value, we should be
willing to stop.
This is something I often remind colleagues when they are
trying Copilot. We do not need to use AI just because we can. Testing, learning
and iterating matters far more than finding a use for a tool at any cost.
Rob Thompson recognised the scale of the digital and AI
opportunity in the NHS, while suggesting that fully realising its potential
could take seven to ten years. That does not mean waiting to act. There are
small, practical steps we can take now, including helping staff develop the
skills and confidence to work with AI.
Better data, fairer care
The presentation on women’s health by Kate Lancaster, Chief
Executive of the Royal College of Obstetricians and Gynaecologists, was
particularly moving. She described how gaps in research and data can shape the
care women receive, and why digital innovation needs to address those gaps
rather than reproduce them.
Her message was clear: the data used to develop AI needs to
represent the people it will serve. That includes women and ethnic minority
groups whose experiences may be under-represented. For me, this brought the
conversation back to evidence. An exciting tool is only part of the picture; we
also need to ask whose experiences are reflected in the data, and whose are
missing.
One example Kate Lancaster shared was a woman having to
record her ethnicity differently on three separate forms because the options
were not standardised. It was a simple illustration of how inconsistent data
collection can misrepresent people and make information harder to use effectively.
Kate Lancaster also echoed the earlier message about
starting with a problem, not a technology. The question should be what needs to
improve, and whether AI can genuinely help.
She highlighted the need for strong infrastructure and
governance that can keep pace with innovation, and time for staff to learn.
Just as importantly, AI should support human connection, not replace it. Giving
clinicians and operational staff more time is valuable if it helps them focus
on the people who need their care.
Promising screening innovations—with important safeguards
The presentations on breast cancer and diabetic eye
screening offered encouraging examples of AI in practice. Speakers described
the potential to save time and money while maintaining the quality of care.
The evidence presented on diabetic eye screening looked
promising, with the AI model described as performing comparably to human
reviewers in identifying eye disease on scans. But the panel’s caution was just
as important as the results: showing that a model works is not enough. It must
also be implemented fairly, accurately and safely, with quality assurance
continuing throughout its lifecycle.
Breakout sessions: upskilling and supporting staff
In the afternoon, I chose the “Upskilling and Supporting Staff” breakout session, which
felt particularly relevant to my role. Chaired by Elizabeth Nassuna, it
featured presentations from Laura Boyd, Marina Khan
and Dr Jeff Hogg.
Although the presentations were interesting, I felt they did
not always connect as closely to the session’s title as I had hoped. Marina
Khan’s discussion of building an AI Community of Practice for pharmacy and
medicine was a relevant example, highlighting an approach to bringing
colleagues together around AI.
What stood out most to me, however, was the concern among
attendees about the lack of a standardised approach to Copilot training. For
me, this was the clearest link to the session’s theme: how do we move beyond introducing
tools to making sure staff have consistent support to use them confidently and
responsibly?
Staff need time and support to learn
One of the main themes I picked up on throughout the day was
concern about the lack of training around Copilot. There is plenty of interest
in what AI might offer, but staff need support to feel confident using it. For
me, that reinforces why training should be part of implementation, rather than
an afterthought.
At MTW, our Library and Knowledge Services offer Getting
started with Copilot. The practical session covers safe and responsible
use, effective prompting and demonstrations of everyday work scenarios. It is
designed to be accessible across roles, including for staff with no previous experience.
Want to get started? Search for Getting started
with Copilot on MTW Learning to book a session. If you would prefer
training for your team or directorate, contact me, Maria Hughes, at maria.hughes8@nhs.net
to discuss arranging a session.
Keeping the conversation going at MTW
If you are interested in AI in healthcare, I would love you
to join the MTW AI Community of Practice. Hosted by the Library and
Knowledge Service, it brings together colleagues interested in clinical and
non-clinical uses of AI, with updates on policy, research and developments,
alongside meetings, case studies and shared learning. You do not need any prior
knowledge—curiosity is enough.
To join, email maria.hughes8@nhs.net to register your
interest. It is an opportunity to ask questions, hear from colleagues and share
what we are learning.
My takeaway: enthusiasm with care
I left the conference feeling excited about the innovation
happening across healthcare. From screening pathways to tools that could ease
everyday workloads, there is much to explore. But enthusiasm should not mean
rushing ahead without asking difficult questions.
For me, the day’s messages came together around a few
priorities: start with a genuine need, look carefully at the evidence, listen
to staff and patients, and make sure the data represents the people who will be
affected. Training, governance and ongoing quality assurance belong alongside
innovation—not behind it.
There is a lot to be optimistic about. Our challenge is to
turn that promise into useful change, cautiously and thoughtfully, so that AI
is implemented safely and fairly while keeping people at the heart of
healthcare.

