Opportunities for Regulators in the Age of AI

The rapid expansion in Artificial Intelligence (AI) capabilities and products has invited individuals and organisations to consider how such capabilities could be harnessed in their professional spheres.
 

Professional regulatory bodies are engaging with the conversation about the use of AI by their registered members and publishing guidance for their use – for example, recent guidance by the Financial Reporting Council, the General Medical Council and the General Osteopathic Council. 

However, what role could AI play in the day-to-day aspects of the regulator’s operational activities?

First, the necessary caveat.  “AI” is a tool which can be deployed with varying degrees of competency or skill.  If used without proper training or care, the output may be limited in its use, or it may be actively harmful. Of course, it is not correct to refer to AI as “a” tool, as there is also significant variation in the type of AI used, from public generative AI to narrower, specifically-designed bespoke models such as the diagnostic and decision support systems being used in healthcare settings; different systems will bring different benefits and generate different types of risk.  A separate blog in this series has addressed the risks of AI to professional regulators. This current blog therefore works from the assumption of (a) properly trained and effective use; (b) robust scrutiny of output; and (c) a secure, confidential, ethical model which is properly targeted to the task(s) asked of it.

From that starting point, here are some of the potential use cases for AI to be used within the operational activities of professional regulators.

1. Education and resources
 

Professional regulators will usually have two primary functions – their fitness to practise or professional discipline function, and the provision of resources, education and training, and guidance to those registered with them (“registrants”). 

AI could be used to prepare guidance documents, refine, refresh and “liven up” training materials, and summarise research papers. Chatbots could be used for interactive questioning and training, to solicit suggestions from registrants as to their training and support needs, or to help registrants find guidance or key sections within regulator rules or good practice guides.  Effectively used, AI can act as a friendly critic and suggest improvements to drafted materials or identify missing information from existing policy or guidance.

2. Research and data management
 

Professional regulators hold a significant amount of data in relation to their registrants, through their registration and application process, and as a result of many years of fitness to practise investigations.

Regulators are also usually resource-constrained (both financially and in staffing) and answerable to their registrants for how their budget is spent.

AI could be used to analyse this data at a speed that is simply not possible if carried out by humans.  This could help in understanding the demographics of registrants, trends in applicant numbers or de-registration, and themes in complaints made about registrants; analyse budgets and financial expenditure; consider decisions of fitness to practise committees for quality assurance or to better understand why allegations were not proved; or consider the effectiveness of conditions of practice orders from the decisions of sanction review hearings.

AI could also help in early assessment of risk, either in triaging complaints that are received about registrants (of which, more below) or in determining if an interim order may be required while an investigation is ongoing – although any such decision-making is more controversial and the regulator would need to ensure that this was transparent and could be properly checked (the “Black Box” concern).

3. Administrative support
 

Relatedly, there is a significant administrative burden on regulators in managing their registration and renewal functions, arranging stakeholder engagement, not to mention progressing what may be many hundreds of active fitness to practise investigations at any one time (see the data for the GMC, Health and Care Professions Council, or Teaching Regulation Agency.) 

Scheduling a hearing requires coordinating the availability of Panel members, legal advisers, registrants and their representatives, and multiple witnesses. Currently, this is a process which is carried out “manually”, through correspondence with each party and then an officer referring to a calendar. This is a clear area where AI could assist with both receiving the information (from chatbots or coordinated calendar systems) and then determining possible hearing dates, thereby reducing the demands on staff.

4. Interacting with the public
 

Professional regulation is, fundamentally, aimed at protecting the public – taking action against those are may not be safe to practise unrestricted, and sending a message to the profession and the public that wrongdoing will be robustly investigated and action taken, where needed.  As such, each regulator will deal both directly with the public (in complaints received or as witnesses in investigations) and indirectly (in the publication of information about hearings and decisions, or in consultation documents).

A common use for AI is as an editor or sense-checker of a draft; regulators could use an AI tool to suggest ways to make public-facing communications or written communications clearer to a lay person, or to a vulnerable individual. It can suggest potential wording for difficult correspondence, or provide a translation function. AI could also be used as an initial point of contact for someone wishing to make a complaint about a registrant, providing guidance as to what is or is not a regulatory matter – perhaps in the form of a chatbot.

5. Fitness to practise investigations
 

As mentioned above, regulators are resource-constrained and many face greater caseloads than can be quickly progressed. Increasingly, they may also receive AI-generated material, which as my colleague Sarah Atkinson points out, will further increase the demands on their time.

Properly used, AI has a wide range of potential time-saving benefits within fitness to practise investigations, from triaging and summarising the first complaint through to preparing the transcript or summary of a fitness to practise hearing.  AI offers the opportunity to prepare summaries of documents or evidence, prepare chronologies, identify missing information from a complaint, log material received and prepare schedules for disclosure, identify what parts of a code or guidance may have been breached, analyse witness accounts for discrepancies, and so on.

There may come a time when AI plays a role in the decision-making of regulators in their fitness to practise sphere – for example in assessing the risk posed by an individual - although the possibility of AI making a final decision in a fitness to practise hearing seems far off at this stage.  

Conclusion
 

With a technology as broad and rapidly developing as AI, there will be any number of potential uses and benefits, particularly to drive efficiencies.  This is particularly the case for professional regulators, given the limited resources and significant demands on the time of their staff and an increasing number of complaints.  However, adopting these technologies must be done with proper regard to, and mitigation of, the risks – an area in which KN is very well placed to provide advice and guidance.

About the author

Laura was called to the Bar in 2012 and joined the Regulatory Department in May 2017. Laura is responsible for leading investigations across the full range of fitness to practise cases brought by the Teaching Regulation Agency and the Education Workforce Council of Wales, covering matters of fraud and examination malpractice, safeguarding concerns, sexual misconduct and inappropriate behaviour, bullying and intimidation, and criminal convictions.