Introduction

SPEAKERS:

- Ajay Chandwani, Head of Market Research and Advanced Analytics, Takeda Oncology
- Inder Singh, Head of Insights & Analytics, Takeda US

KEY TAKEAWAYS:

• Agentic AI consolidates fragmented research into queryable disease repositories
• Internally built to democratize insights, enabling seamless access and enterprise wide scalability across oncology and beyond
• Agentic AI architecture that specialize in different types of qualitative and quantitative data types
• Co-creation with brand teams turned challenges into a wave of adoption and shared success
• Current LLM constraints guide a phased roadmap toward advanced predictive capabilities

Pharma companies invest millions each year in variety of market research studies, yet its value often peaks around the initial presentation before becoming difficult to access. Takeda Oncology saw an opportunity to change that narrative - recognizing that vast repositories of insights were underutilized while teams often sought answers already available across geographies and therapeutic areas. Inder Singh captured the challenge: “Teams naturally hesitate to embrace something new. Why turn to AI when I can just ask my insights team?” The solution required more than organizing files - it called for a bold reimagining of how pharma organizations capture, access, and amplify the value of their knowledge investments through artificial intelligence.

Takeda’s oncology division faced an opportunity to go beyond simple accessibility and truly unlock the potential of its knowledge assets. As Ajay Chandwani noted, market research conducted across functions and brands was often presented and discussed but rarely leveraged to its full potential, leaving “so much of this content underutilized.”

By addressing this challenge, Takeda aimed to transform this opportunity - such as duplicated efforts across geographies - into a foundation for smarter, connected insights that drive global impact. Takeda’s creative solution was the GenAI Disease Area Expert - a groundbreaking platform that unifies market research, publication data, and secondary sources into a single, AI-powered query system. Designed to transform static archives into dynamic intelligence, this innovation reimagines how insights are accessed and applied. The challenge wasn’t just technical - it was cultural. Convincing brand teams that AI could deliver value beyond traditional analytics required bold thinking and a vision for smarter, faster decision-making.

Chandwani’s team embraced a credibility-first strategy, starting with a bold proof of concept: “For the CML disease area, we first built a patient journey using GenAI to show stakeholders what’s possible when you create a knowledge base from existing content. This helps build trust in what we’re trying to achieve.” That initial deliverable became a turning point - transforming theoretical potential into tangible value and sparking confidence in the vision. Singh highlighted the power of co-creation:

“We partnered with brand teams as experts - starting small, showing what was possible, and asking for feedback: Are you getting the insights you need? The right answers? Then we kept improving together.”

This iterative, collaborative approach turned early skeptics into enthusiastic partners driving shared ownership and accelerating adoption.

Takeda embraced an agile, innovation-driven approach - eschewing the traditional “launch and hope” model in favor of continuous refinement. Chandwani explained that by iterating quickly based on real user feedback and engaging early adopters first, the team delivered rapid wins that showcased the platform’s transformative potential. This dynamic strategy not only accelerated adoption but positioned Takeda as a pioneer in reimagining knowledge access through AI. Takeda’s AI journey began with a rapid proof-of-concept on a vendor platform, moving through four stages - POC development, brand rollout, iterative evolution, and impact validation - before a strategic pivot. Rebuilding internally ensured enterprise-scale deployment and allowed full customization to Takeda’s unique needs, overcoming the inherent limitations of vendor-based solutions.

Singh shared the vision: “We built the Agentic AI model in-house, now scaled across multiple oncology disease areas with unlimited access and full control of the tech stack.” This internal build unlocked major advantages - no licensing fees, complete data sovereignty, and a roadmap shaped by Takeda’s priorities. Most importantly, it evolved into a scalable, internally driven enterprise capability with active support of Takeda’s technology team.

The architecture features an orchestrator agent that tailors responses appropriately for different functional roles, supported by specialized nested agents for diverse data types. Today, the market research agent transforms qualitative studies, publications, and other relevant information into actionable intelligence. The modular design fosters creative collaboration across teams while also accelerating innovation by enabling parallel development. Chandwani explained: “My team may focus on market research, while others work on structured or external data agents. Together, we integrate these into a cohesive ecosystem.” This distributed model reimagines analytics as a dynamic, co-created capability for the future.

The platform called AskSofie, inspired by “Sophia,” symbolizing wisdom and knowledge, is built on a secure AWS Aurora foundation, powered by Claude 3.5 and Bedrock to deliver advanced AI capabilities. Already live for chronic myeloid leukemia (CML) and polycythemia vera (PV), AskSofie is driving smart decisions and accelerating innovation, with rapid expansion across oncology on the horizon.

Singh highlighted the evolving landscape: “AI is advancing rapidly - especially for secondary data. While today’s strength lies in text, our roadmap ensures continuous upgrades as technology improves.” This adaptive strategy keeps Takeda at the forefront of innovation. Chandwani explained the current focus: “LLMs excel with qualitative data but aren’t yet optimized for raw quantitative inputs like IQVIA files with rows of data. Structured sources such as claims and EMRs require curation before use.” This reality shapes Takeda’s phased roadmap - prioritizing text-based insights today while building infrastructure for seamless quantitative integration as AI capabilities advance. Human behavior is generally more challenging than technology. Singh noted the core hurdle: existing workflows felt sufficient, and the AI value proposition wasn’t immediately clear to busy brand teams. “Why switch to an AI model when I can just ask my insights team?” Overcoming this mindset required thoughtful change management - showing tangible benefits and building confidence through collaboration. Adoption succeeded through a smart strategy: start with early adopters, deliver quick wins, and build credibility before scaling to the mainstream. Co-creation - treating brand teams as partners, not end users - changed the dynamic entirely. Continuous iteration based on real usage refined the platform and created champions who drove momentum across teams and therapeutic areas. When the platform delivered a clear, well-sourced answer to a complex question, it turned skeptics into advocates. Chandwani highlighted this as a pivotal moment that built stakeholder credibility and accelerated adoption.

Team turnover created knowledge gaps that the centralized platform quickly filled. Singh noted, ‘One repository - one place where everything sits - really hit home during turnover. Teams were hungry for content, and this gave them rapid answers.’ The platform proved invaluable for onboarding, shortening learning curves and preserving institutional knowledge. The initiative also sparked broader strategic questions about analytics organization structure. Chandwani reflected:

"We are thinking how our analytics organization will evolve in the next few years and what we need to do differently to be able to come to that place. Do we need a separate innovation team focused on AI, or can a regular analytics team do these kinds of things?"

The question reflects a fundamental tension facing pharma companies: whether AI capabilities require specialized teams or should be integrated into existing analytics functions.

Takeda's GenAI Disease Area Expert currently functions as an intelligent query system, democratizing access to research insights that previously required analytics team mediation. Brand teams can directly explore the knowledge base, testing hypotheses and triangulating evidence without formal research requests. But the vision extends far beyond information retrieval into predictive and prescriptive analytics.

Chandwani articulated the long-term potential: "If you get this right, you can even ask the question, for example, if I increase my sales and marketing spend by X million dollars, what will the impact be? That's in theory. There's a lot of things you can do. This will become your single source of truth." The platform could evolve to simulate other key development milestones and model resource allocation scenarios with unprecedented speed and sophistication.

Progression goes from descriptive analytics (explaining what happened) to predictive analytics (forecasting possible futures), and finally to prescriptive analytics (recommending best actions).

Each stage requires escalating technical capabilities and organizational trust. A brand team comfortable querying historical research insights may need considerable convincing before trusting AI-generated marketing spend recommendations.

Takeda's strategy of establishing credibility through simpler applications before advancing to higher stakes use cases reflects sophisticated understanding of this maturity curve. The patient journey deliverable that launched the initiative required only information synthesis, not prediction or prescription. As the platform proves its reliability on lower-risk queries, it earns permission to tackle more consequential decisions.

Technical evolution remains critical to this progression. Singh noted that "as the technology evolves and LLM experiences evolve, we'll enhance it even more with the technology," particularly for integrating structured data sources like claims and EMR that current LLMs handle poorly. These quantitative sources are essential for the predictive and prescriptive capabilities Chandwani envisions. Market research provides context and qualitative insight, but estimating marketing impact requires quantitative data and real-world evidence.

The external data agent represents another capability expansion. Rather than relying solely on internal repositories, the platform will automatically monitor predefined external sources - relevant news outlets, clinical trial registries, regulatory filings - extracting relevant information and integrating it into the knowledge base. Chandwani described the vision: "This model will automatically go to external websites that you have predefined. You want to make sure you're going to reliable websites and it will go there automatically. It will search for the right content, extract the content, bring it to you, and then extract the right information from the content." This transforms the platform from a static archive into a continuously updating intelligence system.

The modular agent architecture positions Takeda for distributed innovation at enterprise scale. Different therapeutic area teams can develop specialized agents within their domains - oncology building disease-specific agents while rare disease develops theirs, commercial analytics focusing on market research agents while medical affairs builds publication monitoring - all while maintaining integrated user experience through the orchestrator. This approach could fundamentally reshape how pharma analytics organizations operate, moving from centralized request fulfillment to distributed capability development coordinated through AI infrastructure.

The democratization potential extends beyond efficiency gains. When insights become universally accessible rather than mediated through analytics gatekeepers, brand teams can explore adjacent questions, discover unexpected connections, and develop more sophisticated hypotheses. The platform doesn't replace human expertise; it amplifies it by eliminating the friction between question and answer, between hypothesis and evidence.

Singh addressed the scalability question directly when asked about expanding across therapeutic areas and markets: "Once you have built the tech stack and once you have built the architecture that's running it then it's just a question of bringing all the therapeutic area content in one place. And this model will access it and we will build - think about it as a sandbox, a sandbox for every therapeutic area which is being leveraged by this platform."

The infrastructure investment pays dividends through replication. Each new therapeutic area requires content ingestion but not architectural rebuilding. The orchestrator, security protocols, user interface, and agent coordination mechanisms remain constant while disease-specific knowledge bases expand. This positions Takeda to scale AI-powered insights across its entire portfolio without proportional increases in development effort or cost.

To get you highlights of Pharma Customer Engagement USA faster, we are using generative AI technology to summarise the transcripts of the sessions. If you have any feedback about the summary, please contact lucy.fisher@thomsonreuters.com.

"My team tells me, oh, but we don't have a seat at the table. My advice is to be proactive, go find the table and get a chair and share how our expertise can support shared goals. Maybe they were not aware or did not understand. Maybe you have to make it easier for them."

This partnership model extends beyond technology. "Our customers don't see the different roles that we all represent," Kelly reminds. "Our customers see the company that we represent and the products." Internal functional distinctions become invisible to customers experiencing unified company touchpoints. When medical, commercial, and market access messages arrive uncoordinated, customers perceive organizational dysfunction regardless of the sophisticated rationales justifying each department's independent optimization.

Building Capability for Digital-First Engagement

The shift toward increasingly digital go-to-market models requires systematic capability building across three dimensions. Novo Nordisk started with foundational work: creating a glossary ensuring consistent terminology. "Can we all talk about the same thing in the same manner to ensure that when we're meeting a customer journey means a customer journey and not a pathway, not a digital asset," Gala explains. Without shared vocabulary, cross-functional collaboration stumbles on semantic confusion.

Training extends beyond front-line commercial teams to back-office functions. "This needs to be much more a company mindset," Gala argues. Technology upskilling ensures infrastructure readiness for emerging capabilities. Change management acknowledges the emotional reality of transformation: "I haven't met anyone that says that. Oh, I love not knowing what is going to come. I love not knowing what is going to be my job or how to do it."

Perez Casesmeiro adds perspective on navigating constant change: "The only constant is change. So I don't think there's anything new. The difference here is that it's happening very fast."

From Aspiration to Operation

The transformation journey requires patience and persistence. Organizational change messages need approximately eight repetitions before achieving stickiness. "Someone said that for people, for a message to stick, repeat eight times. So probably do 16," Gala advises. Starting with believers before scaling broadly allows proof points to accumulate. Simple tools, Excel-based editorial calendars, transparent customer journey maps, can drive meaningful coordination improvements without waiting for sophisticated technology solutions.

Hergli shares an example of cross-company collaboration that demonstrates possibilities beyond internal silos. Working with a partner holding commercialization rights in Japan for a Chiesi product, the companies coordinated regulatory filings for a new indication. "We have been helping them from a medical perspective with discussing with the healthcare professional and key opinion leaders, but also from regulatory perspective, from marketing perspective," she explains. The cross-functional, cross-company collaboration succeeded because shared purpose transcended organizational boundaries.

Kelly synthesizes the discussion into actionable principles: have one clear goal, communicate prolifically across functions, ensure everyone has voice in strategic conversations, and use meetings to hold up mirrors when customer-first principles are violated. "I think there are some practical things that we can be doing on a daily basis to achieve this vision," she notes. "We're trying new things and we're having to break down the silos global to regional, regional to local, but also within our functions, within our organization."

The imperative is clear: move from customer-first language to customer-first operations, one coordinated touchpoint at a time. As Perez Casesmeiro reminds, overlapping functions create organizational resilience: "No one operates alone. Shared responsibility is crucial in ensuring we don’t fail. There is too much at the stake for patients and healthcare professionals." The challenge isn't eliminating complexity but learning to navigate it effectively, transforming bureaucratic friction into collaborative strength that ultimately serves patients and healthcare professionals better.

To get you highlights of Pharma Customer Engagement EU faster, we are using generative AI technology to summarise the transcripts of the sessions. If you have any feedback about the summary, please contact lucy.fisher@thomsonreuters.com.

Text

Discover more on this topic at Pharma USA 2026 (March 17-18, Philadelphia) - North America's largest cross-functional pharma gathering. Visit the website here.