Historically the role of AI in healthcare has been somewhat department-limited. Its proven role in accelerating drug development has certainly given the industry a taste of AI’s capabilities, but there remains a landscape ripe with opportunity along the health and care supply chain.
The industry remains below average in its adoption of AI, and it’s perhaps no surprise. Heavily regulated and notoriously complex to navigate, the healthcare and pharmaceutical sectors rightly prioritise concerns around reliability, safety and trust.
In contrast, the public is already using AI in everyday life. According to the UK Government’s AI Skills for Life and Work survey, 73% of people had used AI in the past month. 35% had used generative AI chatbots and 15% had used AI to support health and wellbeing.
Healthcare and pharmaceutical organisations cannot move at the same pace as the general public. Nor should they. But expectations are still changing, and as patients and service users become more familiar with AI-enabled tools, the industry has to answer a harder question: how can it deliver faster, better, more personalised care without weakening the trust it depends on?
Much of the conversation focuses on highly visible use cases, but early adoption doesn’t necessarily mean the best use of AI. In fact, there are more practical versions of trust that get far less attention.
Did the medicine arrive when it was supposed to, was the right person sent to the right patient, and did the service tell people the truth about what was possible?
The missing middle of healthcare AI
Healthcare organisations are networks of constrained decisions. Every day, they make choices about allocation, scheduling, routing, capacity, prioritisation and trade-offs.
Some of those decisions are clinical or scientific. Many are operational, and those that are have a direct effect on experience, cost, reliability and trust.
This is the part of AI that often gets overlooked in healthcare. AI is usually discussed as something that sits at the front of the organisation: a search tool, a content assistant, a patient portal, a digital front door. All useful. But the front door is only as good as what sits behind it. If the operation can’t deliver on the promise, a better interface just gets people to the disappointment faster.
In healthcare, those promises are held to a higher standard. A faster response, a same-day delivery, a more personalised service or a better triage experience cannot simply sound good. It has to be safe, accountable and deliverable. The interface can set the expectation. The service has to meet it.
But what happens in the middle? The busy layer where organisations decide who goes where, when, with what, under which constraints, at what cost, and with what level of acceptable risk.
That middle layer isn’t glamorous. It’s scheduling, routing, allocation, sequencing and replanning. It’s also where much of the measurable value sits.
Why logistics is really an experience problem
Logistics doesn’t sound like the future of AI in health and care, but it’s the operational backbone that determines whether critical goods and services reach people reliably.
Central to this is the ‘middle mile’: the movement of goods, equipment or people within a service, or to enable it. The middle mile is critical to efficient operations, but often underappreciated.
A pharmaceutical wholesaler is not just delivering products. It is navigating the movement of medicines and medical supplies through depots, routes, customer windows, cut-off times, cold-chain rules, driver regulations and service expectations. A home-care provider does the same when scheduling a visit. It has to match people, skills, travel time, equipment, medicines and patient needs. Not just fill a slot in a diary.
These are healthcare experience problems as much as operational ones. When they work, nobody notices. When they fail, trust disappears quickly. A late delivery becomes a pharmacy problem. A missed visit becomes a patient problem. A badly timed message ends up with the call centre. An unrealistic service promise becomes a reputational one.
The organisation might call it utilisation, routing, workforce planning or cost-to-serve. The person relying on the service experiences it more simply: something was promised, and it didn’t happen.
In healthcare, optimisation is a trust problem before it’s a cost problem.
The role optimisation should play
Optimisation is not the shiny bit of AI. That might be exactly why healthcare should pay attention. It is older, quieter and unlikely to produce a flashy demo. But it is very good at solving a kind of problem healthcare is full of: making better decisions under constraints.
AI and optimisation aren’t competitors; they do different jobs. Generative AI is powerful for language, synthesis, drafting and interaction. Machine learning finds patterns in data. Simulation helps leaders explore scenarios.
Optimisation helps an organisation choose the best available action while respecting rules, trade-offs and constraints that support service users, workers and the planet. Delivery windows, loading times, geography, driver rules, service priorities, emissions, cost, cold-chain needs, skills and availability all affect what is possible. These complex operations are where optimisation thrives.
Unlike generative AI, optimisation algorithms work from defined inputs and mathematical models rather than probabilities. They calculate the best available outcome based on the information provided, rather than making predictions or assumptions, meaning they cannot ‘hallucinate’ in the way large language models can. In operational environments where reliability is critical, that distinction matters. That is how trust becomes more than a claim. It becomes something the system can prove.
None of this removes the need for judgement. Good outcomes still depend on good data, clear objectives and the messy rules people carry in their heads. That’s precisely why optimisation is useful: it makes assumptions visible, trade-offs testable and outcomes measurable.
Proof in practice
This isn’t theoretical. Satalia has already applied this kind of optimisation in live, operationally demanding environments.
For Tesco, Satalia built a smart deterministic engine that creates optimised delivery schedules in under 90 minutes, accounting for driver rules, vehicle types, depot constraints and live traffic conditions. The same approach applied to the supermarket’s last-mile operations has enabled savings of more than 11 million miles a year and 8,000 tonnes of CO2.
We used this grounding when first working in health and care – the same class of capability went into NHS home-visit scheduling: matching skilled practitioners, medicines and equipment to patient needs while accounting for routing, timing and operational constraints. Healthcare isn’t the same as grocery logistics, but both depend on trusted decisions made under real-world constraints.
In healthcare, the results were also clear. Utilisation increased by around a third, while travel times and distances reduced by 13%. By making sure the right people with the right skills attend to patients at the right time, healthcare organisations can improve both operational efficiency and patient outcomes. The same principles apply across the movement of medicines, equipment and services, helping make the unpredictable more predictable.
From efficiency to service strategy
The most valuable operational AI opportunities extend far beyond shaving a few percentage points off cost. They can change what an organisation is able to offer.
Take a healthcare logistics network with multiple depots, thousands of daily deliveries, fixed routes, tight customer windows, cut-off times and limited visibility into actual utilisation. On the surface, this looks like a transport-planning problem. Underneath, it’s a service-strategy question. Which routes become viable? What has to change in picking, labelling, loading, driver planning and customer communication? Where is the business case strong enough to justify changing the proposition?
If routes are fixed and planning cycles are slow, the organisation is constrained in what it can offer. It may suspect that a different delivery model could cut cost or improve utilisation. But without dynamic planning and a credible simulation of the trade-offs, the decision is hard to make. This is the under-discussed value of AI in operations: it both automates the current process and tests whether it’s the only option.
Healthcare organisations often separate operations from experience. Ops teams manage capacity and constraints; marketing, digital and service teams manage the promise. Service users and healthcare professionals experience the two as one, so when the promise isn’t connected to operational truth, trust erodes.
A digital front door that gives guidance without understanding capacity is only half useful. A patient support journey that nudges people toward a service that can’t respond quickly enough creates frustration, and a pharmacy delivery promise that ignores the true shape of the network creates avoidable inbound queries.
Personalisation should not just tailor the message. It should tell the truth about the service behind it.
What this means for leaders
For pharma, this widens the AI agenda beyond the obvious areas. AI can accelerate discovery, support medical writing and improve content operations – this we know. But it can also improve the decision systems around launch planning, evidence workflows, patient services, field-force effectiveness, supply chain and commercial execution.
For health systems and providers, digital transformation can’t stop at the interface. Access, triage, scheduling, workforce planning and patient communication all depend on the operational layer underneath. In practice, the highest-value starting point is rarely a new platform. It is optimising the plan against the rules and constraints already in place and returning a better schedule into the rota and workforce systems teams already use.
For consumer health, AI-mediated journeys will create new expectations around advice, availability, fulfilment and support. The organisations that win trust will be the ones whose AI-powered experience stays disciplined about what’s true, safe and deliverable.
Making AI operationally trustworthy in healthcare
The next phase of AI in healthcare will be defined less by who has the most impressive demos and more by who can make AI operationally trustworthy. That involves moving from isolated tools into trusted, measurable workflows.
Sometimes this means rolling out proven optimisation capabilities quickly. Other times it means co-building around a hard, specific problem. Often it means combining optimisation, machine learning, generative AI, simulation and human expertise in a single, governed workflow.
The starting point matters more than the technology choice, and it rarely needs to be a platform decision. Most organisations already know where trust is breaking down: a missed visit, a late delivery, a promise the front line can’t keep. That single workflow, not the whole estate, is usually where the case for change is strongest and the risk of starting is lowest, because it means testing a better plan against the rules, the rota and the systems already in daily use, rather than replacing them.
Worth asking before any of that begins: what needs to happen better, faster or more reliably, what constraints make that difficult today, what would need to be true for people to trust the output, and how would the organisation know if the outcome had actually improved. If those answers aren’t clear yet, that’s the conversation worth having before the technology one.
In healthcare, people need to trust what AI says. Just as importantly, they need to trust what the organisation does next.
For healthcare organisations, the opportunity extends far further than simply improving efficiency. Saving time, reducing unnecessary travel and making better operational decisions allow more resources to be directed towards patient care, innovation and research. AI can continue to accelerate drug discovery and support clinical decision-making, but there is just as much opportunity to transform the operational backbone that sits underneath it.
Applied at scale, this is about more than shaving a few points off cost. It decides whether the medicine, the visit or the service arrives the way it was promised, which is the same test this piece opened with. Grocery and retail have already shown what optimisation can do for complex, high-volume supply chains, and Satalia has already carried that capability into health and care, in NHS home-visit scheduling among other settings, where reliability, safety and trust can’t be treated as separate problems from cost. If there’s a single workflow in your organisation where that promise keeps breaking, that’s usually where the conversation with us starts.
