AI Agents for Business Users in Healthcare: Rethinking Scale and Adoption

The promise of generative AI is everywhere, yet a staggering 95 percent of enterprise projects fail to deliver tangible value. This disconnect between hype and reality stems from a fundamental misunderstanding: treating a consumer-grade technology as a plug-and-play enterprise solution.

At the Indegene Digital Summit 2025, Jonathan Wray, Co-founder and COO of Aible, delivered a keynote presentation that offered a pragmatic roadmap for rethinking AI scale and adoption. His message was clear: it’s time to move beyond one-off pilots and start building AI agents for healthcare that are tailored, business-led, and capable of solving problems at superhuman scale.

The first challenge Wray addressed is that large language models (LLMs) were never designed for enterprise use. They are built for consumers and trained on public internet data. When applied to enterprise artificial intelligence contexts such as healthcare, they lack the organizational, clinical, and compliance knowledge needed for precision and trust.

This gap leads to errors, hallucinations, and what the Harvard Business Review has called “workslop” or outputs that look polished but are inaccurate and require extensive human review. Many companies compound the problem by developing custom AI systems from scratch, creating a Generative AI life cycle problem: every three to six months, as models evolve, teams must rebuild, revalidate, and resecure their systems.

In healthcare, this results in a scalability nightmare, where constant redevelopment slows innovation and adds significant regulatory friction.

Why it matters: To move beyond AI fatigue, healthcare enterprises need standardized, modular frameworks that enable scalable upgrades without restarting every development cycle.

Wray urged leaders to flip the AI adoption model: make it business-led, not technology-led. “You can’t code your way to strategy,” he said, arguing for the creation of a dedicated AI Product Manager role to bridge technical and commercial priorities. This new leader must speak both the language of data and the language of business impact.

He also called for replacing traditional waterfall development with rapid, agile prototyping. The goal: build something testable in 48 hours, gather real feedback, and iterate fast. “If you fail in hours instead of months, you’re learning at enterprise speed,” he noted. Within 30 to 60 days, teams should be able to prove or pivot an idea at low cost and high speed.

Why it matters: Speed to learning defines success in AI. Agile methods let teams experiment responsibly, reducing waste and aligning solutions with measurable outcomes.

Most AI deployments today focus on natural language query interfaces that answer user questions. While intuitive, this model limits the technology’s value because it depends on what users already know to ask.

Wray advocated for connecting AI agents directly to enterprise data through information models. Systems that integrate structured and unstructured data across the organization. This transforms the AI from a reactive assistant into a proactive intelligence layer that surfaces insights, correlations, and risks humans might never see.

For example, in LLMs in healthcare, AI agents could analyze clinical notes, supply chain data, and patient feedback simultaneously to uncover unseen bottlenecks or adherence issues.

Why it matters: The future of AI agents for healthcare lies not in answering questions faster but in discovering insights no one thought to ask.

Wray cautioned that trust is the single biggest barrier to enterprise AI adoption. “You can’t trust what you can’t verify,” he said. A polished AI-generated report without sources is meaningless because users must still spend hours checking for accuracy.

His prescription: integrate deterministic loopback mechanisms that automatically verify AI outputs against the original data. Effective agents should make verification seamless, allowing users to hover over a statement and instantly see the underlying source. This approach eliminates guesswork and builds the governance backbone required for scaling AI safely across healthcare and pharma environments.

Why it matters: Trust and traceability are prerequisites for scaling AI. Without embedded verification, even the most elegant tools will fail in regulated enterprises.

Wray’s bluntest critique was reserved for organizations chasing small gains. “If your AI saves someone five hours a week, it’s not worth the effort,” he said. Instead, the focus should be on superhuman scale: using AI agents to tackle challenges that are impossible to process manually.

He shared a striking example: an AI system that analyzed customer service data and discovered $12 million in annual costs wasted on the least profitable customer tier. No one had ever spotted it because the data volume was too vast for humans to analyze.

This is where AI adoption and the consumerization of healthcare intersect. By revealing hidden inefficiencies, unmet patient needs, or emerging risks, AI gives healthcare leaders a panoramic view of the enterprise. The goal is not marginal efficiency, rather it’s transformation.

Why it matters: The biggest value of AI is in solving what humans cannot scale. That’s the difference between an automation tool and a business revolution.

A critical theme throughout Wray’s keynote was sustainability. Every few months, new models and frameworks emerge, tempting organizations to start over. Instead, he proposed viewing AI adoption as an evolving Generative AI life cycle—a continuum of learning, adapting, and improving.

Enterprises must design systems that can absorb change without breaking. This means decoupling model upgrades from business logic, automating validation, and continuously training AI agents on new, relevant internal data.

Why it matters: Treating AI as a living system rather than a one-off deployment helps organizations stay resilient amid constant technological change.

Collectively, the keynote reinforced that AI adoption in healthcare is not about chasing the latest model. It’s about building scalable, secure, and business-led systems. It is a shift from deploying tools to designing intelligence, from asking questions to enabling discovery.

As Wray concluded, “The fundamentals of business haven’t changed. What’s changing is how deeply we can understand what’s going on inside our organizations.” For healthcare and pharma leaders, that understanding could redefine scale, decision-making, and the future of intelligent enterprise operations.