AI Not Just an Add-On — The New Operating System for Pharma Analytics

Walk into almost any pharma boardroom today, and the chances of you seeing the same picture are high. Dashboards filled with AI initiatives, dozens of pilots, multiple co-pilots, and a handful of enterprise-ready chatbots. All this might seem impressive on paper, but the reality is that none of them deliver enterprise-wide impact.  

In our recent Pharma Roundtable: AI Not Just an Add-On — The New Operating System for Pharma Analytics, a familiar sentiment surfaced almost immediately: despite the explosion of GenAI activity, very little of it is translating into enterprise-wide transformation.  

As one stat shared during the discussion put it bluntly, nearly 80% of pharma AI pilots never make it past the POC stage. Not because the models don’t work, but because organizations aren’t designing their systems, workflows, and operating models around AI. 

What followed was one of the most candid conversations hosted by Ankit Bansal, Pharma Solutions Architect, where MathCo leaders, Snehamoy (Sneh) Mukherjee, Head of Pharma AI, Ashwin Gopalakrishnan, Head of Life Sciences, Jaideep Allam, Principal – Life Sciences Solutioning, and Silvana Dimitrov, Partner – CPG Practice Lead, spoke about what it truly takes to scale AI in Pharma.  

The Excitement Phase 

Ankit: Let’s set the context. Over the last few years, pharma has been in an “AI excitement phase.” Whether it’s a small biopharma, mid-size, or large pharma – everyone has launched GenAI pilots and use cases. Their own version of chatbots, agents, and copilots have been deployed. What are you seeing? 

Ashwin: It’s evident that adoption has been aggressive. In just over two years of GenAI being in the limelight, we’ve seen hundreds and thousands of pilots and POCs. We’re building enterprise-grade systems ourselves. But there’s a gap between deployment and transformation. 

Ankit: What’s creating that gap? 

Ashwin: The problems that get piloted are often the ones the loudest stakeholders want to solve. A chatbot to answer questions faster. A content generator. Success becomes “we productionized it.” And then people move on. 

Snehamoy: That’s where the MIT statistic comes in: nearly 80% of AI pilots don’t make it past POC. 

Jaideep: It’s not because AI doesn’t work. It’s about how it’s adopted and embedded into the organization. 

The Loudest Voice Problem 

Ankit: So, the issue isn’t the technology, but how organizations decide what to build? 

Ashwin: Exactly. Most solutions are add-ons layered on legacy systems with siloed approaches. There’s also a pinch of wishful thinking that AI is the Holy Grail, and it can do anything or everything, which ultimately is not happening. 

Let me give you an example. A marketing team asks for a content generator. IT builds it. Usage spikes. Success is declared. But then what? Did marketing fundamentally change how they work? Did they reach more physicians? Did they improve engagement rates? Usually, no. It’s just another tool in an already cluttered toolbox. 

Jaideep: That’s the loudest stakeholder problem. Use cases get prioritized based on influence, not enterprise value. 

Ankit: So we end up with twenty-seven pilots, and no transformation. 

Ashwin: Precisely. 

What’s Actually Lacking 

Ankit: What’s the missing piece that stops these pilots from scaling? 

Ashwin: Outcome-first thinking. Instead of asking “Where do we add a copilot?” we should ask “How do we reimagine sales or marketing with AI?” Today, success is defined by deployment, not impact. 

Ankit: So it’s tool-first thinking instead of outcome-first thinking. 

Ashwin: Exactly. The challenge remains that people are talking about it, and they don’t know how to take that next step to make that happen. 

Jaideep: And this is where context becomes critical. Pharma has largely fixed data and infrastructure. Models work. Governance exists. Yet pilots don’t scale. 

The real bottleneck isn’t data, it’s context. 

Without therapeutic, workflow, and compliance context, GenAI produces answers that look right but aren’t useful. 

The Cost Reality Check 

Ankit: Are there practical barriers as well? 

Snehamoy: Cost is a big one. With GenAI, expenses can scale unpredictably, unlike traditional analytics projects. 

Ashwin: That leads to underinvestment. Companies allocate a small slice of their budget to AI instead of asking how AI should reshape how the entire budget is spent. 

Jaideep: If you’re spending a billion dollars on media, the question isn’t “Should I spend $20M on AI?” It’s “How does AI help me deploy that billion dollars better?” 

That shift requires deep context, not more tools. 

The “Next Step” That Gets Missed 

Ankit: So organizations deploy AI, and then what? 

Ashwin: The next step is asking how the capability evolves to drive broader objectives. How do we increase physician reach? Improve engagement? Change outcomes? 

That evolution rarely happens. 

Jaideep: Because success is misunderstood. Success isn’t “we deployed AI.” It’s “we changed how this function operates—and we can measure it.”A sales copilot isn’t successful unless it improves reach, quality of conversations, or prescription outcomes. 

Ankit: Can you give me an example of what that looks like in practice? 

Jaideep: Sure. Let’s take sales enablement. Most organizations say, “We built a copilot for our sales reps. They can now ask questions about products and get instant answers. Success!” 

But the real question should be: “Did our sales reps increase their reach? Are they having higher-quality conversations with physicians? Are we seeing better prescription patterns? Did we reduce the time from first contact to prescription?” 

If you can’t answer those questions, you haven’t succeeded. You’ve just deployed a tool. 

Trust, Adoption, and Training 

Ankit: Is trust still a concern? 

Jaideep: Not really. 70–80% of employees use GenAI daily in their personal lives. Inside enterprises, adoption drops to 20–30%. That’s not fear, it’s relevance. The tools aren’t embedded into real workflows with the right context. 

Ankit: So if it’s not trust, what’s creating that gap? 

Jaideep: The question is: Are organizations really creating the most relevant tools for their employees? Number one. 

And we are seeing through POCs that organizations are definitely taking that step forward. 

The other big piece, however, is: How do they use those tools most effectively in the process or in the context of their day-to-day work? 

That is another important consideration that organizations now have to start thinking about, whether we call it change management, training, or whatever the case may be. That is one big area of challenge that is coming in right now. 

Ashwin: Which brings us to training. Pharma invested heavily in traditional sales training. Now with AI tools and AI adoption coming in, do you fundamentally need to rethink the profile of a trainer? Will that get better adoption or thinking around what people need to be trained on, what tools, how do you use it for customer interactions, and so on, as opposed to trying to train the traditional trainer on how to train the broader organization? 

Jaideep: Exactly. Training now has to focus on working differently because AI exists, not just on how the tools function.

Silvana: And that requires capturing cross-functional and tribal knowledge. Context doesn’t live in datasets; it lives across teams. 

The Real Shift 

Snehamoy: With a strong context layer, AI stops being a chatbot and starts acting like an operating system. Insights become relevant. Workflows connect end-to-end. 

Ashwin: Data was yesterday’s advantage. 

Jaideep: Models are table stakes. 

Silvana: Context is the next competitive battleground. 

Ankit: And the pharma companies that win won’t be the ones with the most pilots—but the ones that finally move from experimentation to real, measurable impact. 

To view the full Roundtable, click here.  

Explore MathCo’s Pharma and Life Sciences Capabilities 

The conversation doesn’t end here. 

At MathCo, we believe AI should be more than an add-on or a collection of pilots. Our focus is on making AI a true enterprise advantage deeply embedded into pharma workflows, grounded in therapeutic and operational context, and designed to drive measurable outcomes. 

If you’re looking to move beyond experimentation and build AI capabilities that actually scale, explore our pharma and life sciences capabilities and learn how we’re helping pharma organizations turn context into competitive advantage.