From Diets to Learning Systems
AI and the Future of Weight Care

This is the first in a series of articles ‘AI in…..’ Where I explore my ideas to use AI to innovate industries. I’m starting with a sector close to my heart: weight care.
For decades, weight care has been built on the assumption that people need better instructions.
Eat this. Avoid that. Exercise more.
That anyone who’s ever put on weight just needs information about exercising more and eating less. And yet obesity continues to rise globally.
Millions of people understand nutrition. Many have tried multiple diets. Many can even recite calorie counts and macro ratios of the food they eat. Indeed, 96% people I surveyed who wanted to lose excess weight said they fully understood healthy eating. The problem has never been a lack of information.
It has been a lack of understanding… of biology and behaviour. And understanding of how those two forces interact inside each individual. And this is where AI could be deployed to reshape weight care.
Understanding Complexity
Modern medicine is still largely built around population averages; clinical trials tell us what works on average, and then guidelines translate those findings into recommendations at the population level.
But obesity is personal. What causes it for one person may not cause it for another. What cures it for one person may not work at all for someone else.
Some people gain weight because of genetic predisposition. Some because of metabolic differences. Some because of medication. Some because of stress, trauma, or habit patterns built over decades. Two people can follow the same diet and see completely different outcomes. And scientists like Professor Giles Yeo have long argued that your body weight is in your genes.
AI has the potential to accelerate our understanding of those biological differences.
By analysing large-scale datasets, genetic information, metabolic markers, and treatment outcomes, AI systems can begin identifying patterns that traditional research methods struggle to detect.
In time, we may understand much more clearly why certain people gain weight more easily, and why different treatments work for different individuals.
Understanding Medication
The rise of GLP-1 medications has been one of the most significant developments in obesity treatment ever. For many patients, these drugs dramatically reduce appetite and support meaningful weight loss. But even here, there is variability.
Some patients respond exceptionally well. Others see more modest results. Some experience side effects. And while plenty do regain weight upon stopping treatment, not all do. And we don’t know why 3 million people are using GLP-1 weight loss medications (e.g. Wegovy or Mounjaro) in the UK, mostly privately. And because most of these people are accessing medication privately, there’s no central data source that can track efficacy or outcomes.
But imagine if this data did exist. AI systems could analyse real-world treatment data and begin to identify patterns across these millions of patients.
- Which individuals respond best to which medications?
- What dosage patterns produce the best results?
- What behavioural support improves long-term outcomes?
- How exactly are the drugs affecting nutrition?
- What could the knock-on effects be of weight loss?
Medication becomes just one component of a much more personalised system of care, while AI could also help identify and mitigate risks we haven’t even thought about yet.
Understanding Behaviour
Food is not just fuel. It is reward, comfort, stress relief, habit, celebration, distraction, social inclusion, even. We eat multiple times a day, shaped by context, emotion, attention and environment. Capturing that complexity has previously been impossible.
But AI-enabled digital systems could begin closing that gap. Over time, behavioural patterns like eating speed, satiety recognition, emotional triggers, habitual responses may become easily measurable. The team at Eatiful started this with the concept of mealtime check-ins. With technologies like computer vision, it will be possible to automate aspects of behavioural observation, something companies like emteq labs are exploring.
Let’s think systems, not point solutions
Now imagine a different kind of weight care. A personalised care plan that understands your metabolism, genetics, physiology and lifestyle. A system that knows what has and hasn’t worked for you in the past. One that understands your emotional relationship with food, your habits, your triggers. Medication tailored to your biology. Psychology-based tools delivered at the moments when behaviour starts to shift. Motivational coaching arriving when support is actually needed. Nutrition and exercise advice adapted to you as an individual based on your tastes, lifestyle and even location - rather than based on generic guidelines.
And crucially: a system that monitors its own impact. Tracking outcomes. Learning from behaviour. Refining recommendations over time. In other words, weight care becomes a learning system.
Learning from other industries
Industries like logistics, advertising and finance already operate this way. Algorithms test, observe, learn and adjust constantly.
Healthcare has moved more slowly in part because the stakes are higher, and rightly so. But as digital health systems begin capturing behavioural data, treatment responses and long-term outcomes, new possibilities emerge. Instead of waiting years for new clinical guidance, systems could begin learning continuously from real-world evidence.
This wouldn’t replace clinical research. But it could accelerate understanding.
So what are the barriers?
Some AI is already entering healthcare clinical documentation tools are gaining popularity..
But most innovation remains cautious and incremental.
Healthcare isn’t shaped only by technology. It is shaped by institutions, processes and evidence frameworks that have evolved over centuries.
Those systems exist to protect patients. But they can also slow progress.
Take diets. Yes, there is evidence showing that a theoretically perfect calorie-controlled diet could lead to weight loss. But there is also overwhelming evidence that most people cannot sustain such diets long-term. 63% of adults in the UK are currently living with overweight or obesity.
Clinical knowledge evolves slowly; through accumulating data, testing hypotheses, peer review and eventual adoption. That process can take years.
If AI systems are forced only to reproduce existing clinical thinking, we risk missing their real potential.
AI challenges the status quo
Many digital health products today simply replicate old models digitally. Generic dietary advice delivered through an app is still generic dietary advice. It still ignores individual physiology, lifestyle, psychology, environment, preferences and behaviour.
It ignores what someone just ate, what they enjoy eating, the weather, their energy levels, their stress levels or how they feel that day. Seen through that lens, it becomes obvious why diets so often fail.
Yet even today, government-funded weight management programmes frequently offer the same generic advice, just delivered digitally. Old thinking. New interface. Zero innovation.
AI systems learn
AI systems generate knowledge as they operate. Instead of simply implementing clinical ideas, they can produce new evidence that challenges them. That prospect is uncomfortable.
Medicine is built around carefully validated evidence cycles that can take five to ten years. But AI systems can analyse patterns across millions of interactions in weeks or sooner.
Imagine a system that learns Alfie needs around 1692 calories on an average day, but today he walked for hours, slept poorly and is likely to feel hungrier. It suggests a warming, satisfying meal and offers a small snack while he cooks to prevent overeating. Consider Elsa, who prefers minimal guidance and simply wants to optimise her Mounjaro dose while managing occasional low blood sugar. The system respects her independence but monitors signals that might require intervention.
In these cases, the system adapts to the patient, rather than forcing the patient to adapt to the system.
AI systems create new knowledge
The system I described above is entirely possible to build today. Large data sets exist. Machine learning can analyse complex biological and behavioural signals at scale. Generative AI can provide conversational interfaces that make these systems usable for patients.
AI could surface insights for clinicians, showing the real-world effectiveness of medications, psychology-based interventions, nutrition guidance and behavioural tools. It could monitor outcomes continuously and refine care plans in real time.
In other words, weight care could finally become adaptive.
But achieving that future requires something more difficult than technology. It requires a willingness to rethink how knowledge in healthcare is created.
For centuries, medicine has generated evidence slowly: through controlled trials, peer review, and gradual clinical adoption. That process has protected patients and built trust in modern medicine.
But AI introduces a new possibility. Systems that learn continuously from real-world outcomes. Systems that can surface patterns across millions of patients. That prospect is uncomfortable because it shifts knowledge away from static, trusted guidelines and toward dynamic systems that evolve.
Yet obesity is already a crisis affecting the majority of adults in many countries. Continuing to apply the same generic advice and expecting different outcomes is not evidence-based medicine. It is inertia.
The real opportunity of AI in weight care is not simply better apps or smarter coaching tools. It is the possibility of finally understanding obesity as the complex, individual condition it actually is.
And if we allow systems to learn from real outcomes, rather than forcing them to replicate outdated assumptions, AI could help us discover what weight care has struggled to answer for decades:
What actually works. For each person.
Katie Lips is CEO and Founder of Rebel Futures. She helps organisations grasp the opportunities at technology inflection points, from the early web and pre-iPhone mobile to cloud-based SaaS and now AI. She has held senior roles at EY and Aimia, and was Head of Digital Policy at Which?. Katie holds a Harvard Kennedy School qualification in Leading in Artificial Intelligence and is a Founder Institute graduate.
Related: explore our work on AI Transformation.