Artificial intelligence comes up in almost every conversation about the future of healthcare and rehabilitation right now. New tools appear regularly, often presented as a way to save time, reduce administration and improve clinical effectiveness.
The wider discussion about AI in the media often focuses on what the technology might replace. That concern seems well founded. Some tasks can be done faster and more cheaply by a machine than by a person, and the early effects appear to be falling on junior roles, where much of what used to be somebody’s first job seems to be the sort of thing these systems handle reasonably well. In health and social care, where clinical judgement and navigating complex systems are fundamental, the picture so far looks somewhat different. Recent work on AI and the NHS workforce suggests these technologies may change the shape of the work more than they reduce it, bringing new responsibilities around checking, validation and professional oversight.
This may be particularly true in case management, which doesn’t break down neatly into tasks. It draws on judgement built up over years: knowing when to build momentum and when to step back, reading the dynamics within a family or support team, and holding a difficult conversation and getting the tone right. So much of the skill sits in the relationships and in the decisions.
There are places where these systems do something other than a quicker version of a human task. Applied to large clinical datasets, they can identify associations across more variables than any of us could hold in mind, and there is work in brain injury looking at functional outcomes and clinical pathways in this way. It is early, and a model built on one population may perform poorly against another, so it is something to follow with interest for now.
AI tools are emerging that are specifically targeted to rehabilitation and case management organisations, but these are only one part of the picture. Many of us probably use AI several times a day without thinking about it, because it’s built into software we were already using. It sits in search results, in email, in the document we’re drafting, in dictation and transcription, and increasingly in case management systems themselves. For many organisations, it may have arrived through a software update, before anyone had much chance to form a view about it.
Work is needed on what all of this means for confidentiality, data protection, and clinical governance, and that work is under way. BABICM has been developing its position on AI and is running a webinar series for members this autumn which takes those questions seriously. We’re not going to try to cover that ground here. It deserves more than a few paragraphs, and we’ve been thinking about another set of questions that tends to get less attention.
Many new AI products, and traditional software products that have integrated AI, are presented in terms of time and efficiency. Hours saved, administration reduced, capacity freed up. Time is easy to describe and easy to picture, so that’s how these things generally get talked about. Some technology saves time in ways nobody notices or values. Other things can make a real difference to somebody’s working week without ever appearing in a measure of output.
Health and social care has seen this before, and many of us recognise the pattern from the move to electronic records. Those systems were introduced partly for efficiency, and the experience of using them has been widely reported as adding to documentation work instead of reducing it. Analysis of the wider evidence on health technology and staff time suggests this is not unusual, with a substantial proportion of studies finding no time saving at all, and the difference often coming down to how something was implemented. Efforts to improve matters have produced a familiar pattern: people report feeling better about the work, while objective measures remain unclear.
What we’ve found is that the same tool can often have different impacts depending on the person, their needs, and their preferences. Two case managers with similar caseloads, similar levels of experience, and the same job description could get very different value from the same piece of software, and the reason may have more to do with how each of them thinks and works than with the tool’s intrinsic benefits.
Each of us has our own strengths and challenges in the role. For example:
There is great variety in the aspects of the role that each of us prefers and those which challenge us, and it follows that any tool could be experienced differently. This could be true of anybody, though emerging evidence suggests it may be particularly the case for people who are dyslexic or autistic or have ADHD, and that the difference may not be small. If somebody has spent years finding ways around parts of a job that don’t sit well with how they think, then removing one of those obstacles might matter more to them than to a colleague who never had it.
There is a further point here which may be easy to miss. Someone who has worked out their own way of doing things has often arrived at something that suits them, and it may be doing more than it appears to. Administrative work is a reasonable example, since for some people it is how they build a sense of a case, or a break from work where the answers are rarely straightforward. A tool others consider more efficient could disrupt an arrangement that was holding more together than anyone realised, and the person may not know until it has gone.
A substantial body of work in health and social care explains why technology programmes succeed in one organisation and are abandoned in another. The people expected to use something matter as much as the thing itself, along with the organisation around them and how the technology arrived. Where these programmes have gone well, the people using them could generally see the value for themselves. Where they have gone badly, the difficulty has usually been found somewhere other than in the technology.
Anyone who has worked within brain injury rehabilitation will recognise the pattern from client work. A memory strategy can be hardest to keep going for the person who would gain most from it, since remembering to check the diary is the same skill that made the diary necessary. What seems to determine whether these things work often has more to do with the assessment, the fit, whether the person wanted it, and whether somebody provided the ongoing support needed to help them get used to it.
We assess every client as an individual. We build support around what that particular person wants and what they’re working towards, and most of us have seen what happens when a generic solution is applied to somebody it doesn’t suit. It seems fair to ask whether we apply the same thinking to our own teams. Whether a particular piece of software is adopted at all, how it is introduced, and who it is meant to help are all questions we could approach the same way.
Nobody has, or could have, worked all of this out. Technology is moving faster than most of us can properly evaluate it, much of what we think we know is borrowed from neighbouring fields, and the governance questions we set aside earlier remain open. What we have settled on is a starting point: the people doing the work and what might make their week better.
The approach we’ve taken at AKA and Social Return is to begin with conversations about what people find difficult, what they’d rather keep hold of, and where a little support might change the shape of somebody’s week. Asking first is slower, a little more awkward, and considerably harder to write a policy around, though it may be the version that leaves case managers doing more of the work that brought them here in the first place.
AI and employment effects
Brynjolfsson, E., Chandar, B. and Chen, R. (2025) Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford Digital Economy Lab. Available at: https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/
techUK (2026) What’s actually happening with entry-level and graduate jobs? Available at: https://www.techuk.org/resource/what-s-actually-happening-with-entry-level-and-graduate-jobs.html
AI in everyday software and search
Ofcom (2025) Online Nation 2025, published 10 December. Available at:
UCLPartners, Health Innovation Network South London and Imperial College Health Partners (2026) Beyond productivity: AI and NHS workforce implications. Commissioned by NHS England London Region. Available at: https://uclpartners.com/new-report-calls-for-workforce-centred-ai-adoption/
Professional guidance in case management
British Association of Brain Injury and Complex Case Management (2026) Intelligent Case Management: The Practical Application of AI, webinar series commencing 20 August. Available at: https://www.babicm.org/events/
Technology, efficiency and staff time
The Health Foundation (2025) Tech to save time: how the NHS can realise the benefits. Available at: https://www.health.org.uk/reports-and-analysis/analysis/tech-to-save-time-how-the-nhs-can-realise-the-benefits
Department for Business and Trade (2025) The Evaluation of the M365 Copilot Pilot in the Department for Business and Trade, August. Available at: https://assets.publishing.service.gov.uk/media/68adbe409e1cebdd2c96a19d/dbt-microsoft-365-copilot-evaluation.pdf
Electronic records and documentation burden
Moy, A.J., Schwartz, J.M., Chen, R., Sadri, S., Lucas, E., Cato, K.D. and Rossetti, S.C. (2021) ‘Measurement of clinical documentation burden among physicians and nurses using electronic health records: a scoping review’, Journal of the American Medical Informatics Association, 28(5), pp. 998–1008. Available at: https://academic.oup.com/jamia/article/28/5/998/6090156 (free full text: https://pmc.ncbi.nlm.nih.gov/articles/PMC8068426/)
Why technology programmes succeed or fail
Shin, H.D., Hamovitch, E., Gatov, E., MacKinnon, M., Samawi, L., Boateng, R., Thorpe, K.E. and Barwick, M. (2025) ‘The NASSS (Non-Adoption, Abandonment, Scale-Up, Spread and Sustainability) framework use over time: a scoping review’, PLOS Digital Health, 4(3), e0000418. Available at: https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000418
Assistive technology and fit
Jamieson, M., McClelland, H., Goudie, N., McFarlane, J., Cullen, B., Lennon, M., Brewster, S., Stanley, B., McConnachie, A. and Evans, J. (2024) ‘AppReminders: a pilot feasibility randomized controlled trial of a memory aid app for people with acquired brain injury’, Neuropsychological Rehabilitation, 34(4), pp. 535–571. Available at: https://www.tandfonline.com/doi/full/10.1080/09602011.2023.2220969 (free full text: https://pmc.ncbi.nlm.nih.gov/articles/PMC11166046/)
Ramirez-Hernandez, D., Stolwyk, R.J., Chapman, J. and Wong, D. (2022) ‘The experience and acceptability of smartphone reminder app training for people with acquired brain injury: a mixed methods study’, Neuropsychological Rehabilitation, 32(7), pp. 1263–1290. Available at: https://pubmed.ncbi.nlm.nih.gov/33563100/
Machine learning and clinical prediction in brain injury
Bark, D., Boman, M., Depreitere, B., Wright, D.W., Lewén, A., Enblad, P., Hånell, A. and Rostami, E. (2024) ‘Refining outcome prediction after traumatic brain injury with machine learning algorithms’, Scientific Reports, 14, 8036. Available at: https://www.nature.com/articles/s41598-024-58527-4