Skip to content

Commit 7d736ba

Browse files
ralph-hawkinsralph-hawkins
authored
Merge pull request #249 from NHSDigital/ai-health-coach-first-post
AI health coach first post
2 parents fc29879 + 9fa981a commit 7d736ba

File tree

4 files changed

+137
-0
lines changed

4 files changed

+137
-0
lines changed

‎app/ai-health-coach.md‎

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,11 @@
1+
---
2+
layout: collection
3+
title: AI health coach
4+
description: Exploring how AI and digital health coaching could support healthy behaviour change
5+
area: personalised-prevention
6+
pagination:
7+
data: collections.ai-health-coach
8+
reverse: true
9+
size: 50
10+
permalink: "ai-health-coach/{% if pagination.pageNumber > 0 %}page/{{ pagination.pageNumber + 1 }}{% endif %}/"
11+
---

‎app/ai-health-coach/2025/09/index.md‎

Lines changed: 119 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,119 @@
1+
---
2+
title: AI Health Coach discovery summary
3+
description: An overview of the research, analysis, and conclusions from discovery phase
4+
date: 2025-09-26
5+
tags:
6+
- discovery
7+
- AI
8+
---
9+
10+
This discovery explored the challenges faced by users of prevention and behaviour change services. It also examined whether these challenges could be addressed through the application of Artificial Intelligence, with a particular focus on AI’s potential to provide health coaching.
11+
12+
## Our problem statement
13+
14+
> Users of prevention services often experience support that feels disconnected or incomplete, and struggle with tools that only help with one aspect of their health – rather than supporting their broader, interconnected needs. While AI has the potential to improve personalisation and connection across services, it’s currently unclear how best to use AI to support across all aspects of someone’s prevention journey.
15+
16+
## Our users
17+
18+
Our users are those participating in (or could benefit from) healthy behaviour change or prevention services. Our review of existing research focused on users at various stages of that journey:
19+
20+
- Better Health users and NHS App users
21+
- adults in the UK aged between 18 and 60, who were mostly from deprived areas (socioeconomic groups C2, D, E)
22+
- people with at least one raised health risk factor (being overweight, being inactive, smoking, heavier drinking)
23+
24+
Other teams across NHS England and Department for Health and Social Care had already carried out extensive research to understand the pain points and barriers across this journey, so we drew upon these rather than conducting new primary research.
25+
26+
## Research and analysis
27+
28+
We identified pain points and user needs, and assessed how these are being met by the existing landscape. We explored:
29+
30+
### The existing prevention and behaviour change journey
31+
32+
We mapped the user journey based on what we knew about prevention and behaviour change services from our experience across NHS England and Better Health. The journey was split into 9 sections, and 4 high‑level stages:
33+
34+
1. become aware and engage
35+
- awareness for change
36+
- seek help
37+
2. find and select a service
38+
- find a prevention service
39+
- select a service
40+
- wait for service to start
41+
3. enrol and engage in a service
42+
- first contact and triage
43+
- ongoing service engagement
44+
- completion of service engagement
45+
4. feedback and next steps
46+
- feedback and next steps
47+
48+
Most of the behaviour change activity we wanted to focus on happened in step 3 – because our research showed that people find it hard to sustain lifestyle change. We thought this was where we could help people the most – supporting them after they'd already started using prevention services.
49+
50+
### User experience of this journey and pain points
51+
52+
From previous research conducted by NHSE and DHSC, we identified 60 pain points for users across the journey. We mapped these to 9 sections of the journey and grouped into 6 themes:
53+
54+
- mental and emotional barriers
55+
- lack of personalisation
56+
- difficulty navigating and poor join-up between services
57+
- too much complexity
58+
- access barriers
59+
- poor follow‑through and support
60+
61+
### Attitudes to AI
62+
63+
We reviewed research on attitudes to AI and digital health coaches. This showed opportunities to reduce health inequalities:
64+
65+
- AI can improve access for those less likely to seek traditional healthcare
66+
- users with lower education levels may find chatbots more useful
67+
68+
It was also noted that digital literacy can impact acceptability, and that AI models can underperform for underserved groups. We observed some principles to consider when designing the service – focus on making it accurate, easy to understand, trustworthy, and empathetic.
69+
70+
### Whether user needs are being met by the current market
71+
72+
We looked at existing health coaching apps in the UK and abroad, from big tech, to public sector, to start-up collaborations with local councils. We uncovered some recurring capabilities and identified opportunities for us to meet user needs in ways which the current offerings do not. These didn’t meet all of the user needs and left an opportunity for us to:
73+
74+
- improve equity of access
75+
- use NHS data and services that private sector apps can’t access
76+
- improve trust, transparency, and ethics of AI use
77+
- support across a wider range of behaviours or risk factors
78+
- make it locally‑relevant
79+
80+
## What we concluded from our research and analysis
81+
82+
We were left with clear areas of unmet user need and concluded that we had found significant evidence to support both parts of our problem statement, so we decided to progress into exploring possible solution options and their feasibility.
83+
84+
## Ideation and future vision
85+
86+
We used the journey map, the pain points, and user needs to generate a series of ‘How might we?’ questions which were then used to develop capability, feature, and solution ideas for further exploration.
87+
88+
### The vision for health coach
89+
90+
We identified some of the capabilities or features we’d like to test and which may form a part of a health coach offering (and the pain points they addressed):
91+
92+
- an intelligent, adaptive chatbot which provides guidance and support in real-time – asking questions, gathering insights, and building up a profile in a friendly and supportive tone.
93+
(Mental and emotional barriers; Lack of personalisation; Access barriers)
94+
- smart nudging through prompts or reminders to keep users from dropping out or disengaging, grounded in personalised evidence and aware of user context.
95+
(Difficulty navigating and poor join-up; Poor follow through and support)
96+
- a progress dashboard that tracks all your health behaviours in one place
97+
(Too much complexity; Poor follow-through and support)
98+
- learning from patterns in how people use the service to make better recommendations.
99+
(Mental and emotional barriers; Lack of personalisation)
100+
101+
We also proposed that Health Coach would, in the future, likely bring together data from numerous sources (and organisations) to create Healthy Lifestyle data record. We also resolved to explore the possibility of Health Coach appearing in the Better Health apps or NHS app.
102+
103+
### Feasibility and viability of delivering this
104+
105+
We explored the technical design. This could use several AI models working together to provide the best answer in line with agreed behaviour change models and clinical safety parameters. We began to look at ways in which Health Coach could deal with escalations from lifestyle topics into clinical – for example by integrating with the 111 Online service. This is something that would continue to be tested in alpha.
106+
107+
We also considered:
108+
109+
- cost and benefits – the societal benefits of delivering Health Coach are likely to significantly outweigh the costs of delivering it, but not in direct cash-savings for the NHS.
110+
- information governance and legal basis – Health Coach would likely use a consent‑based model for processing data, but there are further dependencies to explore (including where it sits and which data it uses).
111+
- strategy and policy – Health Coach fits with health policies, with a specific mention in the 10‑Year Health Plan.
112+
113+
## Conclusion and plan for alpha
114+
115+
We found significant evidence to support both elements of our problem statement, and that a solution is feasible and viable. We recommended that the project progress into alpha to test our riskiest assumptions:
116+
117+
1. can we support people to improve their health through building habits and creating sustained behaviour change?
118+
2. can we do this in a way that reduces or does not worsen health inequalities?
119+
3. can we manage clinical risk safely?

‎app/ai-health-coach/ai-health-coach.json‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,6 @@
1+
{
2+
"layout": "post",
3+
"eleventyNavigation": {
4+
"parent": "AI health coach"
5+
}
6+
}

‎eleventy.config.js‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -98,6 +98,7 @@ export default function (eleventyConfig) {
9898
'select-people-for-invitation',
9999
'vaccinations-in-the-app',
100100
// Personalised prevention service collections
101+
'ai-health-coach',
101102
'personalised-prevention',
102103
'lung-health-check',
103104
'nhs-health-check-online',

0 commit comments

Comments
 (0)