AI in healthcare analytics: Milliman MedInsight’s approach to delivering trusted answers at speed and scale

Organizations are under growing pressure to make faster, more confident decisions in an increasingly complex healthcare environment. As healthcare continues shifting to value-based care, administrative tasks and data volumes are both growing. Although many leaders see artificial intelligence (AI) as a possible solution, it also brings new challenges. Without access to accurate data, a strong analytics foundation, and robust data governance, AI can actually increase workload and erode trust in data insights.

In this blog, Iyibo Jack, Chief Product Officer at Milliman MedInsight (MedInsight), highlights a key opportunity to scale AI responsibly. He emphasizes the importance of shifting the focus from individual AI tools to delivering verified answers that help organizations manage total cost of care, improve outcomes, and achieve sustained performance improvements.

Key questions we’ll address:

Q. Why does AI matter in healthcare analytics?

AI matters in healthcare analytics because we’re at an inflection point. Costs are climbing, analytics teams are stretched, regulations are getting more complex, and the data keeps multiplying. There’s a widening gap between the number of decisions organizations need to make and their capacity to make them well. AI is how we close that gap.

In practical terms, we see AI speeding up work on three fronts. It predicts outcomes and resource needs before they hit, such as identifying rising‑risk populations or forecasting ED utilization. It takes on administrative burden so clinicians and analysts can focus on higher‑value work instead of manual chart review or ad hoc reporting. And it allows people to interact with complex data in plain language, so a care manager can ask, “Which members are trending toward readmission next month and why?” and get a clear answer without waiting in a queue for custom reports.

For MedInsight’s customers, AI matters for three specific reasons. First, better inputs drive better outputs. Whether you’re using an enterprise LLM, clinical summarization, or a predictive model, the reliability of the result depends on the quality of the underlying data. Our peer‑reviewed methodologies and rigorous governance improve the consistency and reliability of answers you receive, grounded in credible data.

Second, domain intelligence beats generic approaches. Our clinical groupers, risk models, and episode classifications embed healthcare context from the start. A model working with clinically meaningful structures will outperform one trying to make sense of raw claims on its own.

Third, the bar is rising quickly. As clinical AI from major tech providers becomes table stakes, expectations around instant summarization, evidence‑based reasoning, and natural language interaction are only growing. The organizations ready to meet that bar are the ones with a trusted, well‑governed data layer already in place.

So for us, AI isn’t about the latest tools. It’s about trust, speed to insight, and helping teams move from information to action with confidence.

Q. What factors are accelerating AI adoption in healthcare and what are the obstacles?

AI is transforming the industry by predicting outcomes and resource needs ahead of time, automating administrative tasks to allow staff to focus on higher-value work, and enabling natural language interaction with complex data so insights are accessible without specialized technical skills.

At MedInsight, we’re observing that healthcare organizations largely view AI as a critical future investment to reach their clinical and financial goals. In a recent survey, about a quarter of our provider and payer customers reported plans to invest in AI within the next 12 to 18 months. Their primary goal is to leverage AI-powered analytics to process vast amounts of data with enhanced speed and accuracy. By applying advanced technologies, these organizations aim to uncover patterns, predict trends, and identify risks and opportunities that would otherwise go unnoticed, enabling organizations to better anticipate patient needs, allocate resources efficiently, and intervene earlier to improve outcomes while managing costs.

However, organizations are also realizing that the true value of AI depends on having trustworthy data, with nearly 60% of payer customer survey respondents identifying integration with legacy systems as a major barrier to scaling AI across the enterprise. Most healthcare organizations operate with a mix of old and new technologies, and their most critical data often resides in legacy platforms. For AI models to generate meaningful insights, they require access to comprehensive, high-quality data from across the organization. Data integration, however, must be approached with a focus on data quality, transparency, and governance. Without these foundations, even the most advanced AI solutions will struggle to deliver their promised benefits. As we support our customers on their AI journeys, we emphasize the importance of reliable data to pave the way for truly transformative analytics.

Q. What are the most impactful current use cases of AI in healthcare analytics?

For us, AI should deliver answers, not just tools. Our vision is what we call “Answers as a Service”: using AI to turn complex data into verified, context-rich guidance delivered directly to payers, providers, and teams across the organization. It closes the gap between “what happened” and “what should we do about it.” Instead of asking users to wade through reports, filters, or dashboards, we surface concise, actionable insights at the moment of need. That reduces friction, builds trust, and lets people spend more time making informed decisions and driving outcomes, not interpreting results.

For example, AI empowers executives and operations leadership with immediate answers to strategic questions, shifting from delayed reporting to real-time insights with clear root causes and recommended actions. In MedInsight, a centralized Model Zoo provides data scientists and machine learning (ML) teams with ready-to-use healthcare models and clinically validated features, dramatically shortening time to deployment in a native Databricks environment.

Humans must also stay in the loop. In healthcare, AI has to be transparent, validated, and subject to oversight. That’s why we design every capability so outputs can be traced back to the underlying data, the reasoning is auditable, and clinical judgment stays with the experts. At MedInsight, AI supports decisions—it doesn’t make them.

Ultimately, our AI roadmap is designed to deliver four core benefits to our customers: faster insights, more reliable answers, personalized experiences, and trust through transparency and verifiable outputs. With AI available directly in the workflow, users will be able to receive precision answers tailored to their role, context, and task. This not only broadens the reach of analytics across actuarial, clinical, quality, and network teams, but also ensures consistency and trust in the underlying data.

Q. How is MedInsight enabling AI today, and what does the future look like?

At MedInsight, we don’t just bolt AI onto existing systems. We build the conditions for AI to work reliably. That starts with the data. Our ingestion, normalization, and governance processes ensure information is accurate, consistent, and well-understood before any model touches it. Paired with peer-reviewed methodologies and full transparency, that gives our customers a trusted data layer that reduces noise, bias, and ambiguity, exactly what you need for dependable machine learning and generative AI.

As payers and providers experiment with enterprise and clinical GenAI, we’re very clear: while there are many powerful models available today, their effectiveness depends on high-quality data and meaningful context. That’s the foundation of our AI strategy. It’s not about applying the latest tools to the problem; it’s about making sure customers can trust the answers they receive, move quickly from information to action, and spend less time deciphering data. We reduce complexity and make analytics accessible so teams can focus on outcomes.

We’re also moving beyond the traditional drag‑and‑drop approach to ad hoc analysis. By integrating natural language with cloud‑based AI, users can describe what they need in plain language, and the system handles the complexity, returning clear, actionable results. That accelerates time to insight and provides a more personalized experience for many types of users within an organization.

All of this is bringing our “Answers as a Service” vision to life across the MedInsight ecosystem. We’re delivering a more intuitive, intelligent, and impactful analytics experience, surfacing trusted, context‑relevant answers to complex questions and streamlining how people interact with data, so organizations can tackle today’s challenges and be ready for tomorrow’s opportunities.

What resources are available to learn more about MedInsight’s AI approach?

Contact us to learn more about healthcare data analytics