HEFMA and CompliMind have hosted their first webinar exploring the fundamentals of artificial intelligence (AI) following the announcement of the partnership between the two organisations earlier this year.
In ‘Getting started with AI’, CompliMind founder Carl-Magnus von Behr focused on large language models (LLMs), how they are trained, how they function and their limitations.
Carl started by providing some startling context about the rapid growth of AI; it took the mobile phone 16 years to reach one hundred million users, but it took ChatGPT just two months. Using AI has significant potential for improved efficiency and productivity across many industries, including healthcare applications, however, this unprecedented rate of adoption is not being matched by AI literacy or an understanding of how to use it effectively. The technology is being incorporated into business functions at a rapid rate, increasing pressure on people to use the tools provided, before teams have been able to agree best practice. What is not being understood fully is that using AI is a skill, and it’s essential to learn how to use it, and just as importantly, when not to use it.
Key takeaways
• AI is trained on many data sources, but this does not necessarily equate to professional evidence, especially in a sector like healthcare when a lot of essential information may be inaccessible to it, for instance, sitting behind firewalls, or relying on the expertise and knowledge of individuals or manual documentation
• AI has no real understanding of language
• AI is trained to predict rather than verify and has no awareness of which parts of its training index are reliable and which aren’t
• AI is good at summarising large amounts of text and drafting reports, saving time for humans, however, extensive review may be necessary for applications that require precision and verification against specific standards
• AI struggles with data and calculation and has no concept of numbers as a sequential scale – that is, the number six and the number seven are just tokens to AI which would not understand that one follows the other
• AI is very good at providing an output that is plausible but not always accurate and not always very good at pushing back against what might be an inaccurate or nonsense question
• Small changes in wording can significantly change the AI response, and such variability must be understood when using AI in professional work.
This webinar demonstrated the extent of the learning which humans must undergo to use AI safely and effectively, something which is not widely understood by many who are happily using the easily accessible tools on a daily basis. In a sector that is as critical to life as healthcare, using AI without understanding how it works and its potential unreliability, as well as its benefits, increases the risk of non-compliance with essential standards and guidance.
The HEFMA and CompliMind partnership aims to support HEFMA members in navigating the growing complexity of compliance, knowledge management and digital transformation. It establishes a member-led approach to the safe and meaningful adoption of AI within NHS estates and facilities services. It will enable the sharing of best practice across Trusts and empower estates and facilities teams to drive the development of AI as well as other digital tools, and their integration into everyday processes, in a way that is controlled, transparent and aligns with professional values.
Future webinars
These webinars are free to attend for HEFMA members. Please log into this site, navigate to Regions and you’ll find the registration links in your diary dates and branch notices.
• Prompt engineering for engineers: the basics of drafting and reviewing with AI
Wednesday, October 7, 2026, 12:30 – 13:30
• Streamlining the paperwork: minutes, action logs, and reports with AI
Wednesday, November 18, 2026, 12:30 – 13:30
• AI in practice: real use cases, benefits, and lessons learned
Wednesday, December 9, 2026, 12:30 – 13:30
