Efficacy depends on patient adherence
Pharma professionals have known for decades that efficacy is only part of the story. A treatment can succeed in every test and trial but still fail in application because of human behaviour. Whether the patient forgets, decides the treatment isn’t working and stops taking it, or a busy schedule gets in the way, patient adherence is often a significant challenge.
And so it is with artificial intelligence.
We know that, if adopted and used in the right way, AI can save time, make us more effective, and empower us to achieve things previously beyond our capabilities.
In short, the effective use of AI is a people issue.
And yet in too many organisations AI is run as a technology project with maybe a bit of training bolted onto it. Then, after the initial excitement of implementation, leaders begin to wonder why they’re not seeing the results they expected.
After three decades working in marketing, I’ve spent the past few years training pharma marketing teams to use AI effectively in their roles and beyond.
Often, I find they’ve been overwhelmed by internal launches of multimillion-pound AI investments and processes that promise to transform the business. Then they return to their desks, look at Copilot, and wonder what on earth they’re supposed to do with it.
Even after just an hour of training, people can get some very useful outputs out of their AI tools and quickly recognise the value. Yet very soon the difficult questions arise: What can I upload? What content can I use? Who checks this? If I use this too effectively, will I put myself out of a job? These are the sort of questions that are rarely addressed in organisational launch decks.
At Kanga we believe that people need to be at the heart of AI implementation and adoption. We treat every rollout as a change programme, using the perhaps overworn but always applicable mantra of keeping the human in the loop.
It’s all very well to give people the licences and even initial training, but when you’re changing workflows, agreeing who’s accountable, and helping managers deal with difficult questions, somebody’s got to take responsibility for making it happen.
Deployment is the easy bit
Rollout can traditionally be divided into four stages:
- Deployment, which is making the system available
- Adoption, which means people are using it repeatedly and in the right way
- Embedding, so making it a day-to-day part of workflows and management routines
- And value realisation, when the change in work improves quality, speed, capacity or risk.
Most programmes appear to focus on measuring the first stage, declare victory and then report on the fourth. Yet no one in pharma would accept prescription volume as proof of patient outcomes. Similarly, licences issued must be the most expensive vanity metric in AI.
Gallup reports that 47% of US employees say their organisation has integrated AI. Only 30% use it weekly, and 14% strongly agree it has transformed their work.
Someone in the organisation needs to define what real adoption looks like for a brand manager, a medical writer, or an MSL and then measure that.
Focus on workflow
McKinsey researchers found that while workflow redesign had the strongest relationship with bottom line value, only 21% of respondents had redesigned even some of their workflows. BCG found that providing better tools improved AI impact by just 5 percentage points, whereas putting a clear strategy in place improved the same metric by 25 percentage points.
It shouldn’t surprise anyone in pharma that clear direction and overall strategy beat better software. AI tools can produce promotional content extremely quickly. But if that just produces more drafts that join the queue for medical, legal, and regulatory review, with all the same handoffs and workflows, the cycle time isn’t going to change. We’re just going to end up with more stressed MLR colleagues. You get a longer queue, faster.
The same is true of any process. If the process is broken or if the inputs are flawed, then AI will produce the wrong answer or result remarkably quickly. So again, we need to ask difficult questions: How does the workflow need to change? Where does medical judgement stay essential? How is quality checked? Can the MLR process be streamlined? And I’m pleased to say that these are questions that many of our clients are beginning to consider very seriously.
A second equally important question is: what happens with the time that is released when AI takes on tasks? BCG found that 42% of frontline AI users save a working day or more each week and yet two-thirds had little or no guidance on what to do with that time. The time saved only becomes valuable if it’s put to productive use, such as better content, more contact with customers, and lower agency spend. But of course, users fear it means fewer people are needed.
The essential role of the line manager
Like pretty much everything else in the workplace, employees experience a company’s AI strategy through the lens of their line manager. Gallup found that employees with a manager who supports AI adoption are 2.1 times as likely to use it frequently, 6.5 times as likely to find it useful, and 8.8 times as likely to say it gives them opportunities to do their best work. Gallup labels these associations but whatever they’re called, the message is clear: line managers are a critical communication channel when it comes to AI adoption.
We need to help them. Managers need worked examples from their own teams’ work; clear guidelines on who can upload what and where, what is secure, and what is and is not allowed in the workplace; and a named person to call when something goes wrong.
Without a company policy and guidelines on AI use, each manager ends up creating their own private AI policy, shaped by their own enthusiasm or otherwise. And that’s not an approach that lands well in highly regulated industries like pharma.
AI training needs to go beyond teaching people which button to press, although frankly that’s important too. It needs to focus on behaviour, and pharma already does this well.Field forces run on call planning, one-to-ones, shadowing, and territory reviews. Surely it’s not beyond us to have similar support for people using AI in the office.
Resistance is a symptom, not the disease
Organisations that raise concerns about AI resistance are probably looking in the wrong direction. We know in medicine that non-adherence often tells us more about the treatment than it does about the patient, and the same applies here. So, when adoption is poor, look to the design and the implementation before you look at the people.
In a classic example of unintended consequences, in an experiment with 1,026 engineers, work labelled as AI-assisted was rated around 9% lower for competence compared with identical output labelled as human-generated.
No wonder people keep quiet about using it.
Yet hidden use is probably worse than low use because it’s hard to govern what’s done behind the scenes and may be done at home on a platform that isn’t subject to a company’s higher security standards.
So if a company wants to see people use AI openly, it needs to stop penalising them for using it.
Governance too has to work. In the EU, AI literacy is now an organisational compliance matter, and European Commission guidance expects it to reflect people’s roles and the way they use AI.
A policy document in a shared folder falls well short of that. People need answers to practical questions: what data can go in, what must be verified, and when to disclose.
Change in action
Across our work with global pharmaceutical companies, we see a consistent pattern: introducing an AI tool is relatively straightforward; changing how people plan, create, review and reuse content is much harder.
At Kanga, we therefore approach AI adoption as part of wider business transformation, not simply as the introduction of a technical tool. That means redesigning workflows, embedding AI into brand planning, providing role-specific training, and ensuring managers and partner agencies understand what needs to change. It also means preparing teams for a content environment increasingly shaped by answer engine optimisation (AEO), where the way information is structured and made discoverable matters as much as how efficiently it is produced.
This work needs clear ownership from the outset. Business leaders may define the opportunity and IT may oversee the platforms and associated risks, but someone must connect the technology with workflows, governance, training, user support and measurement. Without that responsibility, AI may be available to employees without becoming meaningfully embedded in their everyday work.
Pharma already understands this principle through its efforts to improve patient adherence: lasting behaviour change depends on creating conditions that make the desired action clear, practical and easy to sustain. AI adoption is no different. If nobody owns how people learn, plan and work with AI, all you have is software, not transformation.
Do you need support with change management or AI adoption?
At Kanga Health, we help global pharmaceutical companies approach AI adoption as part of wider business transformation, not simply as the introduction of a technical tool. We bring technology, people and processes together through change strategy, workflow design, role-specific training, brand planning, governance and measurement. We also help teams prepare their content for a digital environment increasingly shaped by answer engine optimisation (AEO).
Get in touch with our team to explore how Kanga can help you embed AI into everyday working practices and deliver meaningful, measurable change
Written by Rupert Watts, a Senior Strategist at Kanga Health
