Menu
Close
.svg)
I had the opportunity this week to sit down with Arvind to discuss the current and future impact of Artificial Intelligence (AI) on the pharmaceutical industry and pharmaceutical marketing research in particular.
Arvind leads the Commercial Insights & Analytics (I&A) group at Regeneron, overseeing the implementation of deep insight frameworks and analytical capabilities to help bring the power of science and new medicines to patients who need them. Prior to Regeneron, he held positions at Sanofi, Johnson & Johnson, Bristol-Myers Squibb, and Pfizer, mostly on the early pipeline and launch side of the business. Arvind is a recipient of the 2024 Axtria Ignite Leadership Award and the 2023 AI100 Award in the Early AI Adopter category in recognition of his achievements to advance the field of Analytics & AI in organizations. In 2023, Arvind was also cited as one of the 10 Most Influential Data & Analytics Leaders in the US Pharma Sector by AIM Research.
Arvind is a past President of the Pharmaceutical Management Science Association (PMSA) and sits on the Advisory Board of the Pharmaceutical Market Research Conference (PMRC). He is also a member of the Regeneron Artificial Intelligence Advisory Committee and has been a featured speaker at key meetings such as the AI Summit, CDO Vision, The MachineCon, the Digital Marketing World Forum, Informa TMRE, the ARF Attribution and Analytics Accelerator Conference and several other forums.
Noah: It's great to chat with you today, Arvind!
Arvind: My pleasure!
N: How do you see AI reshaping how we do market research? What is the right balance among AI, analytics and the human intelligence gathering and analysis components?
A: I’m an AI optimist, but a guarded optimist. There’s often confusion about what AI can genuinely achieve and how to leverage it to identify meaningful insights. With data and content exploding and becoming increasingly granular, it’s clear that relying solely on human processing is no longer sufficient. However, would we trust machines to fully replace human judgment? Unlikely. AI’s true value lies in augmenting human thinking, not in replacing it and taking over.
AI and machine learning today are powerful tools for generating hypotheses, pattern identification and providing a starting point, but they aren’t replacements for human insight. Extracting deep, nuanced understanding still requires human interpretation, especially when navigating behavioral subtleties. The journey from data to information to knowledge and ultimately to wisdom remains fundamentally human, though AI enhances our agility to synthesize and categorize this journey, and reduces the cumulative legwork involved.
In a world of overwhelming information, AI empowers us to stay focused on what matters—moving closer to wisdom without losing depth or going down an irrelevant rabbit hole.
N: Where should pharmaceutical company marketing research departments start to harness AI’s potential for insights generation, especially since many companies are approaching AI with some degree of caution and prudence?
A: Privacy is paramount, especially in our industry, where protecting respondents and research integrity is non-negotiable. But today’s technologies are unlocking possibilities we’ve never had before. For the first time, we're breaking into the world of unstructured data. Traditionally, insights and analytics has straddled two separate worlds—numbers and qualitative insights—but it’s been tough to bridge them into an all-encompassing narrative. Numbers have always been easier to work with, while true insights have been harder to capture and connect it to the granular level.
Now, AI is changing the game. With tools like natural language processing, we can dive deeper into sentiment, tonality, emotion, and intent—capturing nuances that go far beyond what a transcript alone could convey. I’ve seen this firsthand: reading an interview (text) transcript about a parent describing their child’s pain can’t fully capture the raw emotion of their voice and the intensity of the pain that they are trying to communicate. But with AI’s ability to analyze unstructured data like voice and images, we may be finally able to preserve those powerful emotional and expressive layers that are crucial in understanding and improving patient outcomes. Obviously, we must ensure that these modalities like voice and image do not in any way compromise the anonymity of respondents or their privacy, which is paramount.
This shift means we’re not just reading words; we’re finally able to hear the full story as it is intended.
N: Could you dive deeper into the concept of unstructured data?
A: Before the advent of AI, most of our analytics focused on structured data within conventional databases. In these systems, information could be organized into logical tables, making it straightforward to extract insights. However, working with unstructured data, like language, images, or audio, presented unique challenges since they don’t fit neatly into these structured formats.
Now, with AI, we have the capability to represent unstructured data, such as images, as large vectors. This new approach enables us to parse and analyze information that previously lacked structure. Although still in its nascent stages, this technology marks significant progress. My optimism for AI’s potential, especially in healthcare, stems from the possibility of capturing complex human expressions—such as deep emotions, agony, suffering, and contempt—that go beyond what words can convey. Understanding these nuanced expressions is essential for deeper insights into human behavior and communication.
I'm excited about AI’s potential to blend human insight with machine processing, enabling us to handle vast amounts of nuanced information. We can still develop hypotheses based on data, but now we can strengthen them by using AI to validate and uncover patterns we might have overlooked—whether due to data overload or our own biases. With unstructured data, AI can reveal subtleties in tone, phrasing, or emphasis that may convey intentions and meanings we’d traditionally miss.
In healthcare, for example, understanding expressions of deep emotions—like agony, suffering, or contempt—requires more than text alone. AI lets us capture these layers, helping us interpret the full range of human expression and intent. This results in richer insights, still grounded in human interpretation to unlock their true meaning. It moves us beyond traditional interviews, offering a deeper, more comprehensive way to understand intent and emotion.
N: How can the pharmaceutical industry leverage AI to build a more patient-centric healthcare experience?
A: People keep asking what’s changed in the healthcare journey. We throw around buzzwords like ‘omnichannel’ and ‘orchestration,’ but at its core, the shift is about moving from passive brand perception to a fundamentally more active brand experience and engagement. Ten years ago, brand perception was more static—how people felt about a brand didn’t evolve in real time, and we wanted to fully understand their mindsets and how they envisioned our brand and benefit against a competitive frame. This is still important, but today, with smartphones and the internet, it’s also about dynamic, continuous engagement. There is an “always-on” brand stickiness that is needed in the brand journey in the omnichannel ecosystem.
Old methods like brand positioning or segmentation studies worked well for understanding product perceptions, but they fall short in capturing the real-time, interactive, and adaptive nature of today's brand experiences. There is also a need for personalization that requires a more granular understanding of personas and engagement. This is why AI is essential. We need systems that can process while constantly updating information to keep up with the speed of our lives, driven by our devices. If you are not adapting your message and mix to these real-time changes, your business is not as agile as it needs to be.
Medicine, once far removed from consumer brands like Nike or Apple, is no longer exempt from this requirement. As science becomes more complex, we’re tasked with making it more accessible and engaging to patients. Simplifying language, communicating benefits clearly, and maintaining a constant dialogue have become critical. And understanding how this impacts behavior continuously is paramount.
The sheer volume of data and touchpoints now makes it impossible to rely on old methods alone. AI offers a way to filter what’s relevant, enabling pharma companies to focus on what shapes patient experiences versus what’s just noise. The challenge is to continually understand and influence brand perception in an ever-evolving landscape, and AI can play a powerful role in that transformation. Think about this as a movie reel versus static snapshots of scenes from the movie. You need to be in the whole movie, not just in select scenes from the movie!
N: At which stage or stages of clinical or commercial development should pharmaceutical companies start integrating AI into market research, and are there specific types of projects where it adds the most value?
A: It depends on the project. I see AI as an enhancer, not the core of the insights process. Insights and analytics today are under pressure to deliver real value. It's not about collecting endless data anymore; it’s about quality over quantity. We’re drowning in data, so the focus has shifted to generating impactful, progressive insights that build real value—whether top-line or bottom-line. If your insights are not driving efficiencies or helping the business become more effective, you need to think twice about why you want to do the project at all.
AI can be a powerful tool early in the process, especially for competitive intelligence to understand the disease landscape for example. With content exploding, AI helps with summarization and pattern recognition, uncovering insights we might miss otherwise. But while AI can reveal threads and patterns, human synthesis is still essential to connect those pieces into a coherent whole. Determining whether information is strategic, and beneficial in a launch environment embedded with uncertainty, still needs human involvement. Otherwise, it is just well collated information with nebulous business impact. We’re in a new content landscape, and keeping an open mind on how AI can help us drive better decisions is key.
N: What considerations might lead pharmaceutical executives to approach AI with caution and skepticism?
A: Several factors fuel skepticism around AI, but it’s less about the technology itself and more about data-related biases, privacy and copyright concerns, and potential misuse. These are crucial issues, especially in healthcare, where strict standards like HIPAA demand a cautious approach. Early AI systems are only as unbiased as the data they’re trained on, so ensuring unbiased data in the future is key to reducing bias in AI outcomes.
As AI advances with tools like synthetic data, we’re beginning to see ways to leverage patient insights without compromising privacy. Obviously, these capabilities will first have to pass the muster of compliance and legal, before we can start trying to pilot these initiatives. For instance, in a secure environment, synthesized patient images (which go beyond anonymizing data or using pseudonyms) could reveal nuances about usage or patient types and presentations that traditional data might miss—all without potentially violating privacy.
While these ideas are still in early in AI evolution, they highlight the potential for how AI could potentially enhance patient care. Moving forward, the pharma industry will need to carefully balance the promise of AI-driven insights with the critical responsibility of safeguarding patient data. This is why compliance and legal are essential business partners in the forward AI journey in insights and analytics.
N: Could you elaborate on the concept of synthetic data, particularly in the context of its application to rare diseases?
A: Synthetic data works similarly to generative AI, like ChatGPT. In generative AI, a model is trained on a dataset and then uses those learned patterns to generate new, contextually relevant outputs that extend beyond the original data. Unlike traditional machine learning, which stays within the bounds of its training data, generative AI creates new content that maintains the relationships in the original data.
With synthetic data, the model learns and recreates patterns from the original dataset, generating new, data that preserves key relationships without privacy concerns. Think about it this way: you can anonymize, pseudonymize, or synthesize this original data. In anonymization, you remove or alter all identifiable markers. But this still leaves a residual risk of re-identification.
In pseudonymization, you use pseudonyms to replace identifiers. But even here, the set up does not get you out of data protection mandates. In synthesized data, you create artificial data using Gen AI. It is entirely disconnected from the source data in that it is different data. It just preserves some of the properties in that data. This is why it is being given consideration across a range of industries. Once it is blessed by legal and compliance, it might hold the key to new insights based on data relationships that are currently available but cannot be tapped into as is. This opens up new insights previously restricted by privacy limitations.
In the context of rare diseases, synthetic data can be especially valuable for finding new patients who will probably benefit from a product offering. Traditional machine learning uses look-alike models to find similar patients based on existing data, but remains within the confines of the trained data set. Synthetic data can vary certain characteristics while preserving core patterns, allowing for broader patient identification in the broader patient pool. Additionally, in many rare diseases, you don’t have enough patients to even meet the sample requirements for lookalike modeling - but synthetic data might provide a way forward. This can improve efforts to locate and treat rare disease patients faster and more comprehensively. Privacy considerations remain crucial, but synthetic data shows promise in making treatments more accessible to those who need them.
N: With the adoption of AI, do you anticipate that pharmaceutical companies might aim to optimize costs in gathering insights? Do you think leadership may view AI as a means to achieve deeper customer insights with a more efficient investment in market research?
A: I don’t believe AI will lead to a reduction in investment in insights. Instead, insights and analytics are increasingly evaluated based on the tangible value they bring—their impact on decisions and business outcomes. The real shift will be in which providers can deliver the depth of insight (beyond what a machine can sift through) that genuinely drives value, and helping curate away from data and facts that are nice to have but don’t unlock any existing tension to constitute real insight.
The core process of generating insights hasn’t changed, but AI has introduced an essential layer of efficiency. If anything, AI will force all practitioners of insights, whether on the client-side or the supplier-side, to a higher value standard. From an accountability perspective, that is not necessarily a bad thing. Viewed correctly, this is an active opportunity to reposition the true value of our craft, rather than a passive resignation to its eventual redundancy.
One area of transformation is the need for curation. Previously, we didn’t focus as much on framing the right problem at the outset. Now, with AI in the mix, it’s even more critical to ask the right questions. If we’re being measured on value, we need to ensure we’re not just generating answers but addressing the questions that truly matter. Insights practitioners must engage earlier in the decision-making process, as they’re increasingly invited to the table from the beginning to frame the conversation earlier. This early involvement allows them to guide the process, pushing back on distractions and aligning on the right objectives. It is a privilege, and we must reinforce our ability to provide value in this way.
In this sense, AI hasn’t made insight work simpler or less necessary - it has elevated the expectations for those delivering insights. Those who can clearly articulate and demonstrate the value they bring will be in higher demand, while those who struggle to differentiate their contributions may find themselves less utilized. AI hasn’t reduced the need for expertise; it has, in fact, highlighted the importance of skilled professionals who can extract and curate meaningful insights from data.
N: Looking 5 to 10 years ahead, how do you see AI's role evolving in insights? Are there emerging AI developments that you believe will become more prominent in shaping the future of insights?
A: I believe the real impact of AI will come from advances in multimodal learning. This approach goes beyond siloed storytelling within structured or unstructured data sources by helping create more comprehensive, overarching narratives. It will enable the integration of multiple unstructured sources, like text, audio, rich media, and immersive content. These diverse modalities are complex, but as we shift from passive engagement to active experience, understanding these nuances together or creating concepts that incorporate information from many modalities simultaneously becomes crucial—and that’s where AI will be truly transformative.
For practitioners, the key is to anticipate this shift and equip ourselves with enough knowledge about AI to engage thoughtfully. By understanding where AI can enhance the depth, agility, and value of insights, we position ourselves to be smarter, more effective professionals. Rather than fearing AI, we should welcome its potential to elevate our craft.
The introduction of any new technology often brings concerns about job displacement. But our field has always focused on efficiency, effectiveness, and accountability. The human element has never been eliminated, and I don’t believe it ever will be. Instead, the demand for us to sharpen our skills will only increase. We should actively embrace the intelligence that technology offers, recognizing that our value lies in leveraging AI to become smarter, not in fearing it as a threat to our roles.
N: Do you think companies that are more selective and quicker to embrace AI to become smarter will outperform their competitors commercially?
A: Yes, any investment in AI must deliver value beyond just proof of concept. The key questions are: is the project scalable, and what were the insights gained? Also, why are the insights better than traditional methods for the use case? AI is maturing, and we're beginning to see which applications can offer a positive return. Some technologies, like multimodal large language models (LLMs) and foundational models, are still early-stage, as are deep learning, natural language processing, and certain neural network architectures that remain 'black boxes.'
As AI progresses, we’re gradually unraveling how neural networks function. Techniques like forward and backward propagation are fascinating and reminiscent of traditional methods like principal component and factor analysis, where we condensed data to uncover knowledge hierarchies. Today, neural networks aim to compress vast amounts of data into pockets of insight, though we don’t always understand the connections.
Companies like Anthropic are working on reducing network complexity to address the issue of ‘superposition” in neural networks ',' which causes the black-box effect. The superposition problem stems from the fact that a neural network often represents more “features” than dimensions, which introduces interpretive problems. By compressing dense networks (i.e. networks with many neuronal layers) into simpler structures, they’re trying to evaluate each layer's contribution to eventual learning from input to output. While it’s early days, this progress may help us better grasp how these networks operate. Additionally, we are now hearing about how reasoning features in the latest versions of foundation models, although the jury is still out on this one.
The alternative to adopting AI is to risk being overwhelmed by the flood of information. I believe it’s far better to be a cautious, discerning adopter than to avoid AI altogether. This approach allows us to become skilled curators of information, adding real value to our work by refining what constitutes meaningful insight. That’s why I’m an optimist. The judicious use of AI empowers us to contribute impactful insights and value to our fields every day.
N: Traditionally, insights professionals have either psychology, business/MBA, or other social sciences backgrounds. Given the vision you’ve described, what background and skill set do you envision for the market researcher of the future - working alongside AI?
A: I believe certain foundational skills in market research will only grow more essential. Understanding human behavior—its nuances, irrationalities, and choices—requires a strong grounding in psychology and an intuitive grasp of conversational techniques. We’ll always need those core skills in conducting thorough and insightful conversations, reading facial cues, managing group dynamics, and interpreting intent. These deeply human aspects of interaction are irreplaceable.
Where I see an evolution is in how we synthesize and communicate insights. Storytelling and persuasion are becoming critical, as researchers not only gather information but weave it into narratives that drive impact. The ability to connect diverse insights and present a cohesive story will set the best researchers apart.
Looking ahead, I can imagine real-time AI assistance enhancing our work, perhaps dynamically adjusting interview guides based on a respondent’s responses. This kind of adaptability could make interviews more personalized and efficient, potentially producing richer insights. But these tools would only augment, not replace, the human skills at the heart of meaningful research.
So, while the foundational knowledge won’t change, the competencies will evolve. Researchers of the future will need to master conversational flow, probe effectively, and link findings to a broader narrative—all while adapting seamlessly to new technologies. We’re already beginning to see this shift, and it’s exciting.
N: Any closing thoughts before we finish up today?
A: I believe that progress in this field will come through partnerships. Effective partnerships allow us to explore AI applications deeply within our business while benefiting from insights across different brands and industries. This blend of perspectives is essential for learning and innovation. In the realm of AI and insights, collaboration will be key to moving forward, and I look forward to seeing how we can achieve that together.
Learning and adapting a new capability comes from perspectives beyond one’s own use case. It must incorporate learnings of others in a variety of industries and applications. Only this will validate the robustness of the capability and how consistently it helps deliver an impact. Suppliers provide this wide lens because of where they sit and inform beyond any single company’s walls. It is true of all insights, and matters as much for our path forward with AI in the insights practice.
