Being wrong, wary and slightly worried: will Cinderella get to the Ball?

Last week, I ended my blog saying that I was off to see my GP on the Monday for a possible shingles vaccination. Given I’m into Day 28 of a shingles episode and still feeling pretty horrible, I didn’t expect to be given the vaccination. I was wrong. My GP advised that I should have it, so I did, and she gets the associated £12.06 fee. I felt I really did get a lot out of my 7-minute consultation. I’m a fan of primary health care, and it was a good experience.

Later in the day on that Monday, I found myself standing on a train, bound for Nottingham and the Health Partnership Network (HPN) Mental Health conference. I stood all the way from Manchester to Sheffield, before the hundreds of students got off and seats became available. It was a very hot and unpleasant journey. Arriving at my hotel, I rewarded myself with a delicious glass of Malbec (which actually cost £1.30 more than the shingles vaccination fee). But it was worth it.

The conference was different to any other I had been to, but more about that later. That said, the entire conference arrangements reflected how mental health care is still a Cinderella service, when compared to acute services. There was an acute conference running in the hotel concurrently with ours. Theirs was held in a large, air-conditioned room, with many white-clothed tables, loaded down with bottled water and sponsor goodies. You could see the screen from any table in the room.

In comparison, our venue was in a cramped room, with portable air conditioners that barely troubled the hot air. We sat in tightly-packed rows of seats. There was no water, and certainly not any sponsors’ gifts. Viewing the screen depended on how large the person in front of you was. We were made to feel very much like the poor relation. I’m not sure why they didn’t combine the two groups as the conference’s foci appeared very similar – the emergence of neighbourhood health care; the use of new technologies, particularly AI; what a mess the changes to NHS England, ICBs and the Department of Health and Care were; population health and more AI. 

Indeed, the NHS 10-year health plan mentions AI 100 times across its 168 pages. Now I’m not a great fan of is AI (artificial intelligence). Many of my colleagues are, and now routinely use AI as part of their everyday work. I don’t really want to engage with AI, either on a personal or professional level. It’s probably the philistine in me.

What made the conference different was that the sponsors had arranged 15-minute sessions with individuals, rather like speed dating, I guess. When booking a place on the conference, you were asked to describe the challenges facing your organisation; something I didn’t get around to doing. So, someone did it for me, and it transpired that our challenges were: AI, reducing talking therapies waiting lists, and how a crack team of psychiatrists could (for a fee) come and sort out any problems we might be experiencing with our consultant psychiatrists. Strangely, (not sure that is the right word) only one of these ‘challenges’ was relevant.

Now don’t get me wrong, some of the presentations and conversations were very interesting. I might even go as far as saying ‘inspiring’. There was one company that took us through the system they were showcasing, via a case study. Two things struck me as interesting. The first was a telephone number that allowed for a single point of access to services, and which in true AI style could hold a conversation with the caller. It wasn’t a chatbot, and through audience participation (asking random questions), it wasn’t a pre-recorded response either.

The programme could speak in 45 different languages and used everyday words to both seek further information, or in responding. It was very clever. The second part of the system both impressed and worried me. It used agentic AI. Agentic AI systems don’t simply answer a prompt (how can I stay cool in this hot weather?) but take many ‘actions’ to reach a conclusion or answer: reading records, conversing with other systems, accessing data sets and information repositories.

What we weren’t told was the cost of such services. They are not run using familiar licences (Microsoft, ChatGPT and so on) which can be used as much as you like for the set licence fee. Apparently, agentic AI utilises a costing model that charges for every action undertaken. So, some AI tasks might actually cost more to an organisation than the salaries of staff such programmes were meant to replace. The assumption, that a digital worker is cheaper than a human one, will really need challenging!  Whilst AI will have a place in future health care, in mental health care, as with other health and care services, people will always be important, and even Cinderella got to go to the ball.