Nelson Advisors: The Future of the AI Enabled Metabolic Healthcare Model
Why the next decade of metabolic care will be organised around data, not drugs and what that means for investors, founders and acquirers
For most of the last three years, the story of metabolic health has been told as a pharmaceutical story. Semaglutide and tirzepatide rewrote the economics of obesity, turned Novo Nordisk and Eli Lilly into two of the most valuable companies in the world, and pulled a once marginal clinical category into the centre of health policy. In the first nine months of 2025 alone, Lilly's Mounjaro and Zepbound generated $39.5 billion in revenue, overtaking Keytruda as the world's best-selling medicine. Oral GLP-1s are now on the market on both sides of the Atlantic, with Novo's Wegovy pill approved in December 2025 and Lilly's orforglipron (Foundayo) following in April 2026 at a US cash price of up to $299 a month.
The UK obesity drug market passed £2 billion by the end of last year, and NHS England is working through a phased primary-care rollout of tirzepatide that is designed to reach 220,000 people over three years and 3.4 million over twelve.
Yet if you spend time with the people actually building metabolic health businesses in 2026, the drug is rarely the interesting part of the conversation. The drug is becoming a commodity input: increasingly available, increasingly cheap, increasingly oral, and increasingly generic as semaglutide exclusivity lapses across markets that together contain roughly a third of the world's adults living with obesity. What is not a commodity, and what is quietly becoming the real product, is the layer that sits around the drug: the continuous data, the personalised titration, the behavioural scaffolding, the tapering protocol, the maintenance plan and the payer contract that turns all of that into money.
That layer is where artificial intelligence lives. This article sets out how we think the AI enabled metabolic healthcare model develops from here, why it matters for capital allocation and where the value is likely to accrue over the next five to ten years.
From episodic weight loss to continuous metabolic management
The first wave of GLP-1 businesses were, for the most part, prescription funnels. A consumer arrived through a paid advert, completed an asynchronous questionnaire, was prescribed a drug by a clinician they never met, and received it by post. The model was profitable while demand outstripped supply and while a monthly subscription could be marked up against pharmacy cost, but it was never a healthcare model. It was a fulfilment model with a clinician in the loop for regulatory reasons.
Three forces are now breaking that model apart.
The first is price compression: with two oral products competing at under $300 a month in the US, generic semaglutide arriving in Canada, India, Brazil and China, and UK private tirzepatide list prices jumping 170 percent in a single move last September, the margin available on drug resale is shrinking fast.
The second is persistence. Real-world data has been sobering. A Cleveland Clinic analysis published in March 2026 followed nearly 8,000 patients who had stopped semaglutide or tirzepatide within three to twelve months of starting, and a systematic review in eClinicalMedicine this year confirmed that weight regain after cessation is real and, in randomised settings, rapid. The Cleveland data are more encouraging than the trials, with average regain of just 0.5 percent a year after stopping, but the reason is instructive: many patients restarted, switched to another drug, or moved into lifestyle support. In other words, the patients who did well were the ones who stayed inside a care system. The drug alone did not carry them.
The third force is the payer. Employers, insurers and national health systems have absorbed the initial shock of GLP-1 spend and are now demanding a different bargain. They want lower total cost, defined duration, evidence of maintained outcomes, and a partner who can be held accountable for all three. Twin Health's "GLP-1 stewardship" model, launched in May 2026 on the back of a randomised trial published in NEJM Catalyst, is a clear signal of where this goes. In that trial 85 percent of participants came off GLP-1s while maintaining their weight loss, and the company now offers employers four distinct coverage pathways, from programme-gated access to defined-contribution caps, each of which uses the AI platform to control the drug rather than the drug to sell the platform.
Put these three forces together and you get a structural shift. Metabolic care stops being an episode of weight loss and becomes a continuous, longitudinal management problem, closer to how we think about hypertension or chronic kidney disease than to a diet programme. And continuous management of a heterogeneous population, with hundreds of data points per patient per day, is exactly the class of problem that AI is good at and that human clinical teams are not.
The anatomy of the AI enabled model
It helps to be precise about what "AI-enabled" actually means in this context, because the phrase is now so widely used that Rock Health's H1 2026 funding report concluded it no longer even makes sense to break AI companies out as a separate category. The interesting question, as Rock Health put it, is no longer "who has AI?" but "who has something AI alone can't provide?"
Looking at the companies that are genuinely building the next model rather than bolting a chatbot onto a prescription funnel, the architecture has five distinct layers.
The first layer is continuous sensing. Over-the-counter continuous glucose monitors changed the substrate of metabolic care more than any software did. Dexcom's Stelo and Abbott's Lingo took a device that was a diabetes-only prescription product and turned it into a consumer wearable at a price point of roughly a pound a day. Signos, which raised $20 million in May 2026 in a round led by Dexcom, Blue Cross Blue Shield of Alabama and Google Ventures, holds the first FDA clearance for a CGM system indicated for weight management rather than diabetes. Add smart scales, wrist-worn activity and sleep data, quarterly lab panels and, increasingly, at-home biomarker tests, and the average enrolled patient generates a data stream that is orders of magnitude richer than anything a GP sees in an annual review.
The second layer is the individual metabolic model. This is where the phrase "digital twin" has migrated from marketing into real clinical use. Twin Health's platform builds a per-patient model of glucose, weight and metabolic response and uses it to recommend food, activity and medication adjustments in near real time. Its ADA-presented data and a real-world study in Scientific Reports showed type 2 diabetes remission rates well above standard care. The important point for investors is that the model, not the coaching, is the asset: it improves with every patient enrolled, it is expensive to replicate, and it is what makes outcomes-based contracting possible.
The third layer is the AI care agent. Nourish, which raised a $100 million Series C led by Menlo Ventures this year and has now raised $215 million in total, pairs every patient with a registered dietitian but embeds an AI health agent in the app to handle the daily work of behaviour change: meal logging, nudges, education, symptom triage and escalation. On the clinician side the same system runs a copilot that surfaces patient insights and strips out documentation. Nourish reports average weight loss of 8 percent and annual savings of more than $2,000 a patient across a payer footprint of over 200 million covered lives, and it has tripled year on year. The economic logic is simple: the AI agent lets one dietitian safely manage a panel several times larger than would otherwise be possible, which is the only way the unit economics of high-touch metabolic care can work at population scale.
The fourth layer is pharmacological stewardship. This is the layer that did not exist two years ago and is now the fastest-growing part of the stack. It covers AI-guided dose titration, early identification of non-responders, prediction of discontinuation risk, structured tapering, and the choice between injectable, oral and combination therapies as the pipeline of more than 190 assets in development reaches market. Omada Health, which grew Q1 2026 revenue 42 percent to $78 million, reached a million members and turned adjusted-EBITDA positive, built its GLP-1 Care Track for precisely this purpose; the company reports members on its programme losing roughly 1.8 times the total weight and more than twice the body-fat percentage of controls, with better muscle preservation. Omada has also joined Lilly's Employer Connect programme as an independent administrator, a reminder that the pharmaceutical companies themselves are now distributing through the software layer.
The fifth layer is the contract. None of the above matters commercially unless it can be sold to a payer in a form that shifts risk. The most sophisticated companies now sell a defined outcome, whether that is percentage weight loss maintained at twelve months, HbA1c reduction, drug discontinuation rate or total cost of care, and price against it. This is the layer that separates a technology from a healthcare model, and it is the layer that is hardest for a consumer-first company to build retroactively.
Why the UK and Europe are the interesting test case
The United States is where most of the capital is: Rock Health counted $7.4 billion of digital health venture funding in the first half of 2026, with 19 mega-rounds accounting for 45 percent of the total and weight management a top-three clinical indication. But the UK and Europe are arguably where the AI-enabled model faces its most instructive test, for the simple reason that here the payer is a national health system with a fixed budget and a formal appraisal process.
NHS England's interim commissioning guidance for tirzepatide is, in effect, a specification for the model we have just described. Every patient started on the drug in primary care must have access to nutritional and dietetic advice and behavioural change support for at least nine months. The rollout is deliberately phased by BMI and comorbidity cohort over three years, with a standardised GP IT template and SNOMED coding so that the whole pathway can be reported nationally. General practices are being offered financial incentives to participate. What the guidance does not specify is how nine months of wraparound support for hundreds of thousands of people is to be delivered by a primary care workforce that was already at capacity.
The honest answer is that it cannot be delivered by humans alone, and the NHS knows it.
NICE has already laid the groundwork. Its assessment of digital weight management services recommended four platforms, Liva, Oviva, Roczen and Second Nature, for use in specialist weight management pathways, with prescribing capability built in, subject to a four-year evidence-generation period and NHS Digital Technology Assessment Criteria approval. NICE's own modelling projected around 48,000 additional people gaining access and roughly 145,000 clinician hours saved. That is the template: conditional adoption, a defined evidence window, and a clear expectation that the digital provider carries responsibility for outcomes.
For founders, the implication is that the UK is not a market you enter with a consumer app and a prescribing partner. It is a market you enter with a clinical governance framework, DTAC compliance, integration with EMIS and SystmOne, the ability to report SNOMED-coded outcomes, and a health-economic case that survives a NICE committee.
That is a higher bar than the US employer market, but it is also a deeper moat once cleared, and it produces evidence that travels well into Germany's DiGA framework, France's PECAN pathway and the Nordic systems, all of which are grappling with the same GLP-1 budget question. Italy's decision to recognise obesity formally as a chronic, recurring disease is a sign of where European policy is heading, even if, as IQVIA noted, no European country has yet announced broad public reimbursement of the drugs themselves.
The business models that survive
If the drug is commoditising and the data layer is where value sits, which business models make it through the next cycle? We see four, with very different risk profiles.
The first is the payer-contracted metabolic care platform. This is the Omada, Twin Health, Nourish and Virta model: sell to employers, insurers or health systems, take on some form of outcomes or cost-of-care risk, and use AI to make high-touch care affordable at scale. It is capital-intensive to build, slow to sell, and has the best long-term economics of the four because it owns the contract and the longitudinal data. Omada's path to profitability at a million members is the proof point that the model works at scale, and its raised 2026 guidance of $322 to $330 million in revenue suggests the market believes it.
The second is the sensor-anchored consumer platform. Signos, Levels and the direct-to-consumer arms of Dexcom and Abbott sit here. The strength of this model is engagement and data density; the weakness is that Rock Health's data show 64 percent of weight management and mental health startups are direct-to-consumer, against 29 percent for digital health as a whole, which means the category is crowded, acquisition costs are high and churn is brutal. The survivors will be the ones that, like Signos, use the consumer business as a wedge into employer and payer contracts rather than as an end in itself.
The third is the pharma adjacent services model. Lilly's Employer Connect and Novo's equivalent programmes are turning digital health companies into distribution and support infrastructure for the drug makers. This is attractive revenue, but it is revenue that depends on the strategic priorities of two companies, and as orals and generics make the drug easier to obtain, the pharma companies' need for a differentiated support layer will grow while their willingness to pay a premium for it may not. Founders should treat pharma partnerships as an accelerant rather than a foundation.
The fourth is the clinical AI infrastructure play: companies that do not deliver care at all but sell the metabolic model, the titration engine, the tapering protocol or the risk-stratification layer to those who do. This is the smallest category today and probably the most interesting for the next fund cycle, because it is where the NHS, European health systems and the large US integrated delivery networks will look when they decide to run the model themselves rather than outsource it.
What the next five years look like
Some predictions, offered with the humility that anyone who forecast the GLP-1 market three years ago deserves.
By 2028, we expect the default metabolic care pathway in most developed health systems to be a continuous, AI-managed programme in which the drug is one input among several and the decision to start, adjust, pause or stop it is made by a clinician supported by a per-patient model rather than by a protocol.
The current fixation on weight loss as the primary endpoint will give way to a broader cardiometabolic frame, with HbA1c, blood pressure, lipids, liver fat and sleep apnoea outcomes bundled into a single contract.
The word "obesity" will feature less in company names and more in the small print.
Sensing will become almost invisible. CGMs will be cheaper, smaller and longer-lasting; wearables will contribute continuous cardiometabolic signals that are today only available in a clinic; and the most valuable sensor of all will be the combination of a pharmacy record, a wearable and a lab panel, reconciled by a model that knows what it is looking at. Companies that own that reconciliation layer, and the consent framework around it, will be the ones that health systems cannot easily replace.
The tapering and maintenance phase will become the commercial centre of gravity. As the cohort of patients who started GLP-1s in 2023 and 2024 moves off the drug, whether by choice, by cost or by policy, the market for keeping them well without it will be larger than the market for starting them was. The evidence base for AI-guided discontinuation is thin but growing quickly, and the first company to demonstrate durable maintenance at scale in a randomised setting will command a valuation premium that looks irrational until it does not.
Regulation will catch up with the model rather than blocking it. The MHRA's AI Airlock, the EU AI Act's high-risk provisions for medical AI and the FDA's evolving stance on adaptive algorithms all point the same way: towards a world in which an AI that adjusts a patient's medication is a regulated medical device with a defined change-control process. Companies that build for that now will find it a competitive advantage; companies that treat regulation as a later problem will find that a NICE committee or a DiGA assessor asks about it on the first day.
And consolidation will accelerate. Rock Health counted 115 digital health acquisitions in the first half of 2026, with the second quarter the busiest since 2021. Metabolic health is a natural consolidation category because the stack is modular: a sensor company needs a care model, a care model needs a payer contract, a payer-contracted platform needs a clinical AI engine, and a pharma company needs all of it. We expect the buyers to include the CGM manufacturers, the large virtual-care platforms, the diversified health insurers, the pharmacy chains with clinical ambitions and, in Europe, the private hospital groups and the larger DiGA and NHS-contracted providers looking to add pharmacological stewardship to a lifestyle offering.
Implications for investors and founders
For investors, the practical question is how to tell a durable metabolic care business from a well-marketed prescription funnel. Our view is that the answer lies in five things. Does the company own longitudinal outcome data on a large cohort, and does that data improve its model? Does it hold a payer or health-system contract in which it carries some form of outcome or cost risk? Can it demonstrate, ideally in a peer-reviewed or randomised setting, what happens to its patients after they stop the drug? Is the AI component a regulated, documented clinical decision support system or a marketing feature? And is the gross margin driven by care delivery efficiency rather than by pharmacy markup?
A company that can answer yes to four of those five is building a healthcare model. A company that cannot answer yes to any of them is renting a moment in the drug cycle, and that moment is closing.
For founders, the strategic advice is uncomfortable but clear. Build for the payer from the start, even if the first revenue is consumer. Invest in evidence early, because in the UK and Europe it is the currency of adoption and in the US it is the currency of contract renewal. Treat pharma partnerships as channels rather than foundations. Design the clinical AI as a regulated product on day one. And spend less time on the drug than your competitors do, because the drug is becoming the cheapest and least defensible part of the business.
The next decade of metabolic health will not be won by whoever has the best access to semaglutide. It will be won by whoever builds the best system for knowing, continuously and for each individual patient, what to do next. That is a data and intelligence problem before it is a pharmaceutical one, and the companies that understand this are the ones worth backing, building and, in due course, buying.
Nelson Advisors > European HealthTech, MedTech, Digital Health Investment Banking
Nelson Advisors specialise in Mergers and Acquisitions, Partnerships and Investments for Digital Health, HealthTech, MedTech, Health IT, Consumer HealthTech, Healthcare Cybersecurity, Healthcare AI companies.www.nelsonadvisors.co.uk
Nelson Advisors regularly publish Thought Leadership articles covering market insights, industry trends, deal commentary, market analysis & predictions @ https://www.healthcare.digital
Nelson Advisors publish Europe's Leading Healthcare Technology Investment Banking Newsletter every week, join 5000+ HealthTech and MedTech subscribers today! https://lnkd.in/e5hTp_xb
Nelson Advisors pride ourselves on our DNA as ‘Founders advising Founders.’ We partner with entrepreneurs, boards, corporates, venture capital and private investors to maximise shareholder value and investment returns.www.nelsonadvisors.co.uk
#NelsonAdvisors #HealthTech#MedTech#DigitalHealth #HealthIT #Cybersecurity #HealthcareAI #FemTech#ConsumerHealth #Mergers #Acquisitions #Partnerships #Growth #Strategy #NHS #UK #Europe #USA#Canada#Commonwealth#CorporateDivestitures #VentureCapital #PrivateEquity #Founders #SeriesA #SeriesB #Founders #SellSide #TechAssets #Fundraising #BuildBuyPartner #GoToMarket #PharmaTech #BioTech #Genomics
Nelson Advisors LLP
Hale House, 76-78 Portland Place, Marylebone, London, W1B 1NT
Meet Nelson Advisors @ 2026 Events
Digital Health Rewired > March 2026 > Birmingham, UK
NHS ConfedExpo > June 2026 > Manchester, UK
HLTH Europe > June 2026, Amsterdam, Netherlands
HIMSS AI in Healthcare > July 2026, New York, USA
Bits & Pretzels > September 2026, Munich, Germany
World Health Summit 2026 > October 2026, Berlin, Germany
HealthInvestor Healthcare Summit > October 2026, London, UK
HLTH USA 2026 > October 2026, USA
Barclays Health Elevate > October 2026, London, UK
Web Summit 2026 > November 2026, Lisbon, Portugal
MEDICA 2026 > November 2026, Düsseldorf, Germany
Venture Capital World Summit > December 2026 Toronto, Canada
Comments