AI is now part of daily work in life sciences. It can draft a trial protocol in minutes, flag a safety signal early and prepare regulatory documents overnight.
And yet, according to McKinsey, nearly eight in ten companies use generative AI, and 80% of them report no tangible bottom-line benefit.
The contrast is hard to ignore. We trust algorithms to predict how a molecule behaves in the human body, but not to draft a routine contract. We let AI scan terabytes of trial data, then sit in a leadership meeting where nobody dares to decide what to do with the time it frees.
The barrier is not the technology. It is how organisations decide, adapt and absorb change.
The cost of waiting is concrete. Every month of hesitation means slower trials, unused capacity and patients waiting longer for therapies that could already reach them.
In 2025, McKinsey’s Life Sciences Practice and QuantumBlack, AI by McKinsey, analysed more than 270 workflows and 1,200 tasks across pharma and medtech. Their findings:
Over three to five years, McKinsey estimates the potential at 5 to 13 percentage points of additional growth and 3.4 to 5.4 points of EBITDA in pharma, and 3 to 7 points of growth and 2.2 to 4.7 points of EBITDA in medtech.
On paper, the value is there. In practice, most of it is not being captured.
The reason is rarely technical. Most AI projects stay in pilot mode. Innovation teams test new use cases while frontline teams keep working the old way. Leaders wait for perfect data before deciding. Risk aversion slowly eats the momentum.
A pilot is usually designed to prove that a tool works. It is rarely designed to prove that the organisation can change around it. Those are two different experiments, and most companies only run the first.
Meanwhile, the pressures on the industry keep building:
AI works. But unless leadership and culture change with it, the paradox holds: we can simulate molecules in silico, yet struggle to redesign a single meeting agenda.
The main reason AI stalls in life sciences is neither data quality nor regulation. It is hesitation at the top.
McKinsey is direct on this point: without a clear mandate from leadership, AI stays stuck in pilots. Yet many leadership teams play it safe. They wait for perfect data, create another committee or add another strategy deck.
Meanwhile, trials drag on, margins shrink and patients wait.
Companies will spend millions on AI platforms, then hesitate to spend one hour deciding how the freed-up capacity should be used. Should it go to innovation, to patient-facing work, to reducing burnout? Leaving that question open is a decision too, and an expensive one.
In our work with leadership teams, we see four traits make the difference:
One healthcare group we know illustrates this. A decade ago, it was built around acute-care hospitals. Its leaders saw the ground shifting: rising costs, new reimbursement rules, different patient expectations. They could protect the old model or reinvent it.
They chose to reinvent. They expanded into ambulatory services, invested in digital tools and reshaped their governance so that leaders could balance short-term results with long-term bets.
It was not easy. The board wanted results. Clinicians worried about workload. The transformation took years. What kept it moving was a leadership team that had decided, together, what it was trying to build.
The real test is not the next AI pilot. It is whether the leadership team is willing to choose. Without that, AI remains one more slide in one more strategy deck.
AI rarely fails because the models are weak. It fails because people do not use it.
McKinsey estimates that up to 95% of life science roles could have AI “teammates”: agents that help with data, documents or workflows. Almost every scientist, clinician or manager could be working alongside AI tomorrow.
In practice, many are not:
The barrier is not the code. It is confidence. When teams do not trust a system, they work around it, and adoption dies without anyone announcing it.
You can buy the best models in the world. Without trust and ownership, they will sit unused.
What builds that trust is rarely a big programme. It is small, repeated practice:
These habits turn hesitation into confidence. They make AI part of daily work rather than a black box people fear. This is cultural transformation in the most practical sense: not a values statement, but a change in what people do on a Tuesday afternoon.
Middle managers carry most of that change. They decide whether a team is allowed to try the tool on real work, whether a mistake is treated as learning or as a fault, and whether the old way is quietly kept as a safety net. When they are left out of the design, adoption stops at their door. When they help shape the pilot, they become its most credible advocates.
The pattern is not limited to pharma and medtech. Hospitals and clinics face it too.
Electronic patient records, ambient documentation and agentic scheduling all promise time back for clinicians. Yet in many hospitals, the new tool ends up sitting next to the old form, and staff fill in both.
The OECD describes the same issue in its work on the health workforce and the digital revolution: digital infrastructure provides the tools, but it takes health workers to put them to productive use, and the rules of work were designed before the digital era.
Digital healthcare transformation succeeds or stalls in the middle layer. Unit heads, nurse managers and care coordinators decide whether the old process is retired, who checks the AI output and what happens when the tool gets it wrong. If nobody gives them that mandate, they protect their teams the only way they can: by keeping both systems running.
For hospital leaders, the lesson is the same as for pharma executives. Before buying the next tool, decide three things with the people who will use it:
Answering those questions takes hours, not months. Skipping them is what turns a promising pilot into two parallel systems.
Read also: Why the Smartest Hospitals Are Rebuilding from the Admin Up
Adding AI to a broken workflow makes the problem faster, not smaller.
McKinsey is clear on this: the value of AI does not come from adding it to old systems, but from redesigning entire workflows.
Consider two situations:
The useful question is different: what should this work look like if AI is a teammate from the start?
For clinical operations teams, that means mapping the full process, from final protocol to first patient in, before choosing where agents fit. Process improvement comes first, automation second.
Some answers are already visible:
Life sciences companies can design therapies that reprogram cells, yet many still route invoices by hand for approval. AI will not fix that until the workflow itself is redesigned.
Freed-up capacity does not create impact on its own. If leaders do not decide where it goes, it is quietly absorbed by more meetings and more reporting.
Most AI dashboards track the wrong thing. They count licences, connected users and pilots launched. None of these says whether the work has changed.
What matters is whether people work differently, and what that changes for patients, timelines and costs. More useful indicators include:
A low adoption rate rarely points to a training gap. More often, it shows that the workflow around the tool, or the trust in it, has not changed. That is a leadership question before it is a technical one.
AI in life sciences is not a future opportunity. It is already here. The question is whether leaders will turn it into value or let it stall in pilots.
Five moves make the difference:
These moves sound simple. They are what separates the companies that capture AI’s value from those still discussing it at next year’s board meeting.
AI is no longer the question. It drafts trial protocols, analyses terabytes of data and automates compliance tasks in ways that were out of reach a few years ago.
The question is whether leadership and culture will make use of it. Technology moves in quarters. Organisations move in years, unless someone decides otherwise.
The stakes are not only financial. Every delay means slower trials, wasted capacity and patients waiting for therapies that could already change their lives.
At Bee’z, we have seen the turning point: it comes when leaders rework decisions, culture and workflows at the pace of the technology. That is when AI stops being a pilot and starts to change outcomes for patients, for teams and for the business.
The algorithms are ready. Which workflow in your organisation will you redesign first, and who will decide what happens to the time it frees?
Let’s talk: https://www.beez-consulting.com/contact
Most AI projects stall because organisations change the tools without changing the work. Pilots prove that a model works, but workflows, decision rights and habits stay the same. McKinsey reports that 80% of companies using generative AI see no tangible bottom-line benefit. The gap usually comes from leadership hesitation, low trust among teams and AI added on top of old processes.
Agentic AI refers to AI agents that can carry out multi-step tasks rather than answer one question at a time. In life sciences, they can draft regulatory documents, prepare site contracts, track supply chains or analyse trial data. McKinsey estimates that 75% to 85% of pharma workflows contain tasks that such agents could automate or augment, freeing a significant share of capacity.
Digital transformation in healthcare is the redesign of how care, research and administration work using digital tools such as electronic records, AI agents or remote monitoring. The tools are only part of it. A transformation is real when routines change: which steps disappear, who checks the output and how decisions are made. Without that, organisations add software on top of old workflows.
A pilot usually proves that a tool works in one team. Scaling it requires changing routines across many units, and that depends on middle managers. If unit heads and managers have no mandate to retire the old process or decide who checks the AI output, they keep both systems running to protect their teams. The new tool then adds work instead of removing it.
Teams decide every day whether to use an AI tool or work around it. If they do not trust the output, they double-check everything or ignore it, and the expected gains disappear. Trust grows through small, repeated practices: testing on one workflow, short debriefs after each use and visible early wins. Without that culture, even the best model sits unused.
Leaders set the mandate, choose where AI should create value and decide how freed-up capacity will be used. McKinsey notes that without a clear mandate from the top, AI tends to stay stuck in pilots. Leaders also shape the climate: when they treat early mistakes as learning, teams are far more willing to test new tools on real work.
Yes. Adding AI to a broken process makes the same waste faster. The more useful approach is to map the full workflow first, remove unnecessary steps and then decide where AI agents fit. In clinical operations, for example, that means reviewing the process from final protocol to first patient in before automating any part of it.
Count changes in the work, not licences. Useful indicators include the share of a workflow actually handled with the AI agent, the time freed per role and how it is reused, the number of manual double-checks still performed, and cycle times on key outcomes such as site activation or regulatory submissions. These show whether AI has changed how people work.
Freed capacity needs a deliberate decision, otherwise it is absorbed by more meetings and reporting. Leaders can redirect it to higher-value work such as faster R&D, more time with patients or reducing burnout. McKinsey estimates that agentic AI could free 25% to 40% of capacity in pharma, so the choice of where that time goes has real strategic weight.
A pilot tests whether a tool works in a limited setting, usually with motivated volunteers and extra support. AI at scale means the tool is part of everyday work across teams, with updated processes, clear responsibilities and indicators that track outcomes. Moving from one to the other is mostly an organisational challenge: decisions, trust, management and workflow redesign.
.png)
Change management in healthcare: the 2025-2026 reports list every lever. None says who translates the strategy in the units, or how to prepare them.
Agile in healthcare works when teams, not tools, change. What agile really means for hospitals, why implementation stalls, and 6 practical steps.
Development isn't decided once a year. What a development conversation contains, why it so rarely happens, and what actually builds capability at work.
