News and Insights
November 4, 2026
Imagine a system that successfully improves every performance metric it tracks, yet still falls short of achieving its ultimate goal.
This is the challenge captured in Goodhart’s Law, often summarised as: “When a measure becomes a target, it ceases to be a good measure.” It highlights a common problem in complex systems: organisations can become so focused on improving a metric that they lose sight of the broader outcome it represents. As debate grows around the future of neurodevelopmental services, and following Professor Peter Fonagy’s recently published independent review into prevalence and support for mental health conditions, ADHD and autism (October 2026), it feels like an important moment to ask whether we are measuring the things that matter most.
For years, much of our effort has focused on improving access to ADHD and autism assessments. Given the huge increases in demand and waiting times, this makes sense. But families are rarely seeking a diagnosis as an end in itself. They want answers and practical ways to help their children. Schools want to understand and respond to children’s strengths and challenges. Young people want support that improves their wellbeing and everyday lives. Yet in many areas, diagnosis remains a key gateway to support, so many spend months, or more often years, waiting for an assessment even when their needs are already clear or are urgent. And when a diagnosis does arrive, the support families have been waiting for does not always follow.
So, diagnosis can be incredibly valuable in helping individuals understand their experiences and access appropriate interventions, but it is only one step. The real goal is ensuring children and young people receive the support they need, when they need it. If diagnosis becomes our measure of success, we risk mistaking a means for an end.
This aligns closely with Fonagy’s prevalence review, which calls for a shift from a diagnosis-dependent system towards a “needs-led plus diagnosis” approach. In this model, need is identified through a broader assessment of functioning, participation, impairment, strengths, risks and personal circumstances. Crucially, systems can then respond to these needs without always waiting for a diagnosis. Diagnosis will, of course, continue to play an important role. The point is not to move away from it, but to recognise that identifying a clinical condition and assessing and responding to needs are different—and both matter.
So how do we make this shift? Digital innovation and AI could play an important part. Rather than simply speeding up diagnosis, these tools could identify support needs sooner, help practitioners build a shared understanding of a child’s strengths and challenges across different systems, or give families tailored support before a formal assessment. For example, we’ve been supporting Assembly, which is developing a digital assistant to help parents better understand their children’s holistic needs and how to meet them. At UCLPartners, we are keen to hear from clinicians, commissioners, or innovators who have more ideas about how digital innovation can help with this shift.
Of course, Goodhart’s Law applies here too. AI tools are built to optimise something, and that can quickly become the new target. If we judge a triage tool mainly by how many assessments it processes or how quickly it moves people through the system, we risk repeating the same mistake with more sophisticated technology. The metric may improve while the questions that matter (what does the young person need, and are they getting the right support?) go unanswered. How we evaluate these tools matters as much as the tools themselves. Success should reflect whether children’s lives are improving, not just whether the system is moving faster.
For me, that is the key lesson from both Goodhart’s Law and the prevalence review. As we explore digital innovation and AI, we should ask not only how technology can make services more efficient, but how it can help children and young people get the right support, at the right time, in a way that reflects their individual needs. Used well, technology could help us move from a system organised around proving need to one organised around responding to it. If we keep that goal in sight, we have an opportunity to build a system that is not only faster, but better.
