Agentic AI in Pharma: Are We Ready for AI That Acts?

In this episode of Pharma Talks, Nataliya Andreychuk sits down with Michal Dojlidko, a specialist who has watched agentic AI in life sciences move from concept to deployment inside pharma, with a particular focus on how to make AI run fast, efficiently, and safely at scale.

The conversation covers the shift from generative to agentic AI, why data foundations decide who wins, how pharma should think about trust and upskilling, and what role humans play once the agents start taking action on their own.

From Answering Questions to Taking Action

Michal frames the last few years as a series of breakthroughs arriving every few weeks. Generative AI started as a chat-style experience: teams used it to draft content, generate images, and get answers to questions. Agentic AI is a different category altogether.

We moved from answering questions to AI being autonomous, taking care of processes and orchestrating workflows behind the scenes, from answering questions to taking action.

That shift changes what pharma teams need to manage. It is no longer just about output quality; it is about the workflows an agent owns and the decisions it’s allowed to make on its own.

Treat AI Like a Member of the Workforce

Nataliya asks whether this means pharma now needs to think about agentic AI in terms of KPIs, almost as if it were a role rather than a tool. Michal agrees, and pushes the analogy further:

“We could, and should, treat an AI agent like a regular workforce. Just like we have a development plan and objectives for employees, we need the same thing for AI: clear goals and boundaries. AI performs well when it has clear guardrails and objectives.”

His bigger point is about ambition. The temptation is to use agentic AI to fine-tune the way things already get done. Michal argues that’s the wrong frame — agentic AI is a chance to reimagine the workflow itself, not just speed up the old one.

Trust, Traceability, and the Data Foundation Problem

Nataliya raises a tension many of Viseven’s partners still face: AI is often treated as a “magic wand,” which works against the trust that needs to exist between people and the systems they’re deploying. At the company level, that trust question scales fast — she points to Takeda, which has reportedly built more than 600 agents internally, as an example of how quickly this is becoming an enterprise-wide reality rather than a pilot project.

For Michal, trust starts with two non-negotiables:

“When we set up AI for an agentic landscape, we have to make sure it has full traceability and auditability, and is compliant with all the regulations. That’s a guardrail we cannot breach.”

The second, and in his view the real prerequisite for any agentic AI setup to succeed, is the state of the underlying data:

“Think of AI like a fast train. You cannot deploy a fast train on broken tracks — it won’t move fast. You have to sort the data first. Once you do that, AI and AI agents will work much more effectively on top of it.”

Upskilling, Not a Big-Bang Rollout

Beyond data and governance, Nataliya points to a more human obstacle: literacy. People can’t trust what they don’t understand, and part of that fear is personal — not being replaced by the very system they’re being asked to adopt.

Michal’s answer is continuous upskilling rather than a one-off training event:

“We need to make sure people are upskilled — not as a one-time, big-bang event, but continuously. AI evolves so fast that learning has to be an ongoing process.”

He warns against over-centralizing AI expertise in a single team, since that quickly becomes a bottleneck. A central function can set standards — security, compliance, how the organization approaches AI — but every individual still needs enough fluency to use these tools well, with the autonomy to make smart, justified decisions inside safe guardrails.

The Marketer’s Dream Scenario — Already Real for SMBs

Nataliya shares an idea she heard discussed at the Reuters Pharma US Forum: marketers imagining a future where they simply talk to an agent, and the agent understands intent, pulls in the latest developments, builds the segmentation, designs the customer journeys per persona, and presents it all back — essentially a dream scenario for a brand team.

Michal’s answer: that dream is already live, just not yet at pharma-enterprise scale.

Those dreams came true already, for small and medium businesses. We saw Open Claw — the most successful open-source initiative I’ve seen. Within a month it took off, and it now has a larger community than Linux built over 30 years.

He notes that NVIDIA has released its own more secure flavor of the tool, adding the extra layer of protection that could make the enterprise version of this vision possible: a brand manager with an agent dedicated to every digital campaign, and an orchestrator layer above them making sure campaigns stay aligned with each other and with the company’s overall message to patients and healthcare professionals.

This is already happening at the small and medium business level. The technology is there — it’s just a matter of introducing it securely into the corporate pharma space.

Guardrails Against Miscommunication

Nataliya zeroes in on the word “securely.” The speed and the technology are both there, but pharma is handling information that can directly affect patients and patient engagement, so the guardrail question matters more than in almost any other industry.

Michal frames the answer in two parts. The first is technical: using an “agent as judge” model, where one agent is tasked with evaluating the output of another. The second, and the one he considers essential in a regulated industry, is human oversight:

“Especially in highly regulated industries like pharma, we need a human in the loop. Imagine handing MLR review entirely to AI — that’s probably not the smartest idea. With humans holding the control button, the risk drops dramatically.”

What’s Left for Humans to Do

If agents are judging other agents and taking on more of the operational workflow, Nataliya asks what that leaves for people and the future of pharma content — are we still strategic decision-makers, or just helpers to the system?

Michal is direct that the role isn’t shrinking:

“The dust hasn’t settled yet. If you thought there’d be less work, I don’t think that’s true. The volume of what’s happening in the world is only getting larger. Humans are still driving and setting the strategy and the goals — we need to make sure our AI companions are following the path we set.”

His expectation is that repetitive and low-value work will increasingly move to agents, freeing people to spend more time on the creative and strategic work that actually builds value — which, in his framing, is where humans will keep the advantage for a while yet.

Nataliya connects this back to one of the show’s most-discussed earlier episodes, on whether AI might eventually replace marketers altogether. Her takeaway here is less alarmed and more collaborative: the two are evolving together, and the shape of that partnership is still being worked out in real time.

Hot or Not: Rapid-Fire Verdicts

Nataliya closes with a round of statements; Michal responds hot or not and explains why.

Agentic AI in the pharmaceutical industry is a bigger shift than GenAI ever was — Hot.

We moved from asking questions to autonomous action. That’s a different category of change.

Most pharma companies are not ready for this yet — Not hot.

The gap isn’t really readiness — it’s trust in our own people. Teams are already experimenting and keen to bring AI in. The world is moving fast, and so are they.

Pharma’s regulatory culture is an asset, not a barrier, when it comes to governing AI — Hot.

We have a real advantage here: years of experience setting proper guardrails. Less-regulated industries haven’t had to build that muscle.

Companies that win with agentic AI are the ones that invested in a data foundation three years ago — Hot.

AI adoption is far easier when the data foundation is already there — that’s the fuel it runs on. We put a lot of focus on picking the right model and forget that even the best model needs the right foundation underneath it.

Human oversight of agentic AI in regulated industries is non-negotiable — Hot.

We don’t really have a choice — regulation requires a human in the loop, and that’s not up for debate. It also means we keep learning and improving as we go.

The biggest risk in agentic AI is moving too slowly, not too fast — 50/50.

We don’t want to be left behind — the pace of this technology means we need the courage to jump on. But moving too fast, without the security baked in, carries more risk than the upside is worth.

Final Words

The throughline of the conversation is that agentic AI is not a faster version of GenAI — it’s a different kind of system, one that acts rather than just answers, which means pharma needs to govern it differently: clear objectives and guardrails, a solid data foundation before scaling, human oversight built into the regulated steps, and continuous upskilling so people aren’t left behind by their own tools.

As Michal puts it, the priority now is making sure teams have the skills to actually leverage what agentic AI can do — and as Nataliya adds, that means treating learning and change management with the same seriousness pharma already gives to compliance.

It’s also, in many ways, the thinking behind eVa AI Agent and Viseven’s eWizard platform: content generated inside brand and DAM guardrails, checked against compliance before it ever reaches MLR, with humans still holding the final call.

Enjoyed this conversation? Want to continue the conversation on agentic AI, pharma marketing, and how the two are evolving together? Talk to us, and let’s explore those exciting topics together.

---