A global pharmaceutical company already has the competitive report it needs. The problem is that the report is buried somewhere within the organisation’s information stores, and the marketing specialist who needs it cannot find it.
They spend the next two days recreating work that someone else completed months earlier. What began as a knowledge-access issue becomes a measurable productivity cost.
I call this research archaeology: skilled people digging through disconnected systems for knowledge their own organisation already owns. Talk to almost any pharma leader, and a variation of this story quickly emerges.
The strange part is how the industry got here. Pharma organisations have spent decades building an extraordinary store of information: clinical research, regulatory filings, commercial data and market intelligence. And in the right hands, it’s a competitive advantage.
In practice, valuable information is often buried in disconnected systems and is wildly inaccessible to the people who would truly benefit from it. Which is why so many organisations struggle to put that knowledge to work. At the Health Management Academy’s Spring 2026 CIO Forum, healthcare leaders identified making existing data usable as a top priority. Not collecting more information, but using what already exists.
Four problems that reinforce each other
The challenges are rarely technological alone. In most organisations, four interconnected factors make it difficult to turn information into usable knowledge.
The first is silos. Across pharma, valuable knowledge sits in CRM platforms, market research repositories, medical information databases, content management systems and regional reporting tools. Most were built independently, with little thought for how they would work together. Each contains part of the picture, but because information moves poorly between them, the picture rarely comes together.
The result is duplicated work, slower decisions and missed opportunities. The global economic cost of data silos is estimated at around $3.1 trillion annually.
The second is legacy infrastructure. Many organisations still rely on technology estates built over years of growth, mergers and acquisitions. These stacks pile up over decades, with each new system added faster than old ones are retired. Replacing them is costly and disruptive, so organisations often work around the problem instead. Over time, those workarounds have a habit of hardening into permanent architecture.
The third is regulation, and it rarely stops at national borders. Compliant campaign data in the UK may need to be handled differently under Swiss data protection law, and differently again elsewhere in Europe. What should be a straightforward effort to build a single view of performance across markets becomes a patchwork of local workarounds. Slow at best and blocked at worst.
The fourth is scale. The volume of information generated by pharmaceutical organisations continues to grow faster than teams can realistically manage. More reports, more content and more analytics do not automatically create better decisions. In fact, without the right systems, they’ll create nothing but more complexity.
Together, these challenges mean people spend far too much time looking for knowledge and not enough time using it.
What agentic AI changes, and where it fails
Agentic AI is the loudest promise in enterprise software right now, but the signal is often buried beneath the hype.
ZS’s recent research on pharma and biotech found agentic AI scaling fastest in IT operations and research and development (R&D), while teams closer to patients and customers move more carefully. That caution is not slowness but the correct reading of the risk.
These teams handle sensitive information, and a badly built agent does the opposite of what the brochure promises: it invents answers, adds checking work and erodes trust faster than any legacy system ever did.
Avoiding that outcome is less about the technology itself than how organisations implement it, because the order of operations decides everything.
Data first. Organisations need a trusted knowledge foundation that brings together campaign materials, market research, analytics, regulatory documentation and institutional knowledge in a single, structured environment. Without that foundation, an agent sees only fragments of information and answers as though it has seen the whole picture.
Governance next. Role-based access controls, audit trails, source attribution, clear accountability, and testing against real queries before anyone outside the project touches it. These are not optional features in pharma, they’re what allow organisations to trust the answers they receive and understand where those answers came from.
In practice, governance is what accelerates scale. The organisations moving fastest with AI are the ones building enough confidence in their systems that teams, regulators and leaders are willing to use them widely.
Only then should organisations focus on workflows. The best starting points are rarely the most ambitious ones, but the activities that consume time and money every day: the overpriced intelligence subscription, the campaign briefs written from scratch, the field teams preparing for client visits by hand, the recurring information requests that pull experts away from higher-value work.
The goal is not simply to automate tasks but to make organisational knowledge available wherever it is needed. A regional team should be able to draw on research gathered anywhere in the business, adapted to local requirements, rather than relying on information that happens to sit within its own market.
We saw this while building a platform for a global, multi-regional enterprise. Prior to working with us, the company was spending roughly $80,000 a year on a competitive intelligence subscription and still digging for data manually. We built a governed knowledge layer and applied AI across it, replacing the subscription, improving access to marketing intelligence across regions, and establishing an audit trail behind every answer.
There was no magic in it. The breakthrough came from making existing knowledge accessible.
The bottleneck isn’t AI
By now, most pharma organisations have moved past small pilots and are working out how to run these systems at scale. The technology is ready and the data already exists, but the organisations that benefit most will be those that can connect those two realities safely and effectively.
AI will not solve a knowledge problem that already exists. If information remains fragmented and inaccessible, AI will simply surface those weaknesses faster. The key takeaway for pharma here is that governance does not slow scale, it’s what makes scale possible.
Alina Piddubna is Portfolio Director at Intellias – an AI-enabled product engineering and digital solutions partner. She oversees the delivery of GenAI and agentic AI solutions for pharma, healthcare and life sciences clients across Europe, with a focus on EU AI Act compliance and AI governance advisory.
