Software & AI: Accelerating Value in Healthcare

IN CONVERSATION WITH

Paul Bleicher, MD, PhD

Chief Data & AI Officer, Ardan Equity

The rapid advancement of artificial intelligence is ushering in a transformative new era, where long-held expectations for efficiency, precision, and innovation in fields like healthcare, life sciences, and daily life are colliding with emerging realities.

To explore these frontiers and foster deeper understanding, Ardan is excited to present the Insight Series — a collection of perspectives, roundtables, and events with leading thinkers, innovators, and practitioners within Ardan, its portfolio and the broader network.

In our inaugural interview, we sit down with Paul Bleicher, MD, PhD. Since 2008, Paul has worked in healthcare AI and machine learning, with peer-reviewed publications in clinical prediction, production AI deployments at Optum and UnitedHealthcare, and two issued U.S. patents in predictive analytics and AI. With his extensive background as former Founder/CEO of Phase Forward and OptumLabs, and as a member of the founding team and Chief Medical Officer of Humedica, Paul is highly valued in his contributions to discussions on machine learning's promise and perils in clinical development and patient care. Paul offers a grounded, expert perspective on how AI is redefining what's possible in healthcare and beyond.

KEY AI OPPORTUNITIES & POTENTIAL RISKS

Where is the most compelling opportunity to apply AI right now?

I believe the core opportunity lies in applying AI to repetitive and time-consuming administrative and clinical workflows. Functions like revenue cycle, prior authorization, chart extraction, quality reporting, and risk adjustment remain costly and data-intensive. I believe AI can materially lower unit costs while improving accuracy and throughput. The key focus is disciplined workflow automation at scale, instead of speculative and incremental clinical innovation.

Cutting through the noise, what are the key areas of disruption for AI?

There are certainly a few clear signals emerging. AI is eroding labor-dependent service models focused on manual review and call-center operations while accelerating analytics, revenue cycle management, care management, and reporting. In my view, companies reliant on linear headcount growth could face compression; however, those embedding automation into workflows have the potential to expand margins and scale more efficiently. Faster iteration may also shorten product development cycles.

How do we distinguish defensible AI from marketing claims?

I believe defensible AI depends on proprietary, high-quality data that compounds in value with scale, deep integration into essential workflows, and objective performance improvements over traditional methods. Vendors that rely primarily on public or third-party models may deliver an incremental benefit, but they may struggle to achieve durable competitive advantage. Structural defensibility arises from systems that continuously learn and improve within proprietary data environments.

AI IN HEALTHCARE & LIFE SCIENCES

How should we think about AI in healthcare operations?

Sometimes AI is glorified to be robots or other jarring images. In reality, AI is reshaping both product and operational processes, the mundane tasks. It accelerates software development cycles, streamlines reporting, automates chart abstraction, optimizes staffing and care coordination, and enables standardized decision-support across clients. Companies that develop high-performing AI capabilities once and deploy them repeatedly across their customer base may have the ability to establish a lasting operational advantage.

What are the main risks and deployment challenges?

People should know that all industries are different. Healthcare has several structures, frameworks and other guidelines, and for good reason. Key risks include data privacy violations, security exposure, dependence on third-party models, and regulatory oversight when AI influences financial or clinical outcomes. Operational barriers, such as EHR integration, workflow disruption, clinician skepticism, and misaligned incentives, often outweigh technical ones. Sustained success requires cross-functional alignment between compliance, IT, and end-users.

What are the limits to AI in healthcare?

AI performance is constrained by data quality, regulatory guardrails, and clinical variability. Many datasets are incomplete, inconsistently coded, or biased by billing incentives, societal and practice norms, all of which limit model transferability across settings. In regulated environments, explainability and auditability are required. AI excels at pattern recognition and structured task automation but has challenges in ambiguous or context-dependent decisions. This is where having a "human in the loop" makes AI viable.

What do you mean by “human-in-the-loop” AI?

We believe human oversight will persist where decisions carry clinical, legal, or financial liability and risk. Routine, operational processes may move toward full automation. However, in our view, winning companies will pair proprietary longitudinal data and deep workflow integration with people at key or ambiguous decision points for measurable outcomes, regulatory discipline, and scalable deployment. Durable value comes from embedding AI into operations as an augmentation to people’s resources, not marketing it as a stand-alone feature.

What is agentic AI, and how is it different from earlier healthcare AI?

Agentic AI extends beyond prediction or content generation. Rather than simply answering a prompt or flagging a pattern, it can carry out multistep workflows in parallel, gathering information, triggering actions, escalating exceptions, and coordinating across systems with limited autonomy. In healthcare, this may be useful in areas such as prior authorization, revenue cycle, care navigation, documentation follow-up, and patient communication. The opportunity is not simply better chat interfaces, but the redesign of fragmented operational processes.

What will determine whether agentic AI creates real value in healthcare?

Agentic AI will create the most value where work is high-volume, partially rules-based, cross-functional, and currently dependent on manual handoffs. Companies will benefit most when agents are tightly integrated into proprietary data, workflow systems, and measurable operating objectives, not deployed as generic assistants. The winners are unlikely to be those with the flashiest demos, but those that can govern autonomous actions, maintain audit trails, and embed agentic capabilities into day-to-day operations without increasing compliance, safety, or liability risk.

HOW AI MAY AFFECT INVESTOR PORTFOLIOS

How can portfolio companies assess AI-related risk?

While not exhaustive, risk assessment encompasses data ownership and rights, privacy and security compliance, dependence on third-party models, and the reliability and reproducibility of model outputs. Companies that use AI internally face different exposure than those embedding AI into externally marketed products for clinical or operational decision-making. In those cases, intellectual property, liability, and regulatory compliance (e.g., under HIPAA and FDA frameworks) are central considerations.

How does AI create shareholder value?

AI may generate value by increasing revenue, reducing costs, and, in some cases, supporting higher exit multiples. It helps automate administrative functions such as claims processing, enhance care management through risk stratification, and improve clinical or operational outcomes when embedded in core products. The crucial distinction is between AI used as a peripheral tool and AI structurally integrated into the company’s data flows and workflows.

What long-term AI themes should LPs monitor?

Investors should focus on where economic value accumulates and the durability of defensible advantages. AI may compress margins in commoditized software but can expand them in data-rich, workflow-integrated platforms. Key drivers to monitor include the concentration of foundation model capabilities, evolving regulatory expectations, and the balance between algorithmic decision-making and clinical judgment. The central investment question is which companies can capture and retain economic benefit from AI in a compliant and scalable manner

ABOUT PAUL BLEICHER, MD, PhD

Paul is the Chief Data & AI Officer at Ardan Equity.

Paul was founding CEO and Chairman of Phase Forward, which over 13 years transformed 10,000 clinical trials for hundreds of life sciences customers from paper to SaaS-based. Paul played key roles at Phase Forward, helping to lead a major merger of equals and preparing the company for its IPO (NASDAQ: PFWD). The company’s acquisition by Oracle—called “the deal of the decade” in life sciences software—was Oracle’s largest in that period. Paul’s 33-year career in healthcare and life sciences software and data began after his internal medicine and dermatology training at Harvard Medical School, and post-doctoral training at the Dana-Farber Cancer Institute.

A life sciences software pioneer, Paul held leadership roles in clinical strategy and development for ethical pharmaceuticals and novel biopharmaceutical products at leading CRO Parexel and subsequently a public biotechnology company. At Phase Forward, Paul led product strategy, built the marketing and regulatory function, and helped lead the acquisition of five companies. Subsequently, Paul joined the founding team of Humedica (Acquired: Optum) as Chief Medical Officer where he was a member of the Board of Directors. Humedica was a pioneer in big data and population health across the provider, payor, and life sciences markets. Following Humedica’s acquisition, Paul became founding CEO of OptumLabs, the research and innovation arm of Optum, part of the UnitedHealth Group. Paul was a faculty member at Harvard Medical school and practiced at the Massachusetts General Hospital where he ran a research lab with high-profile publications.

Dr. Bleicher was a co-author of the National Academy of Medicine's 2022 book entitled: Artificial Intelligence in Health Care: The Hope, the Hype, the Promise, the Peril. He holds eight patents, including two patents in predictive analytics & AI.  

Disclaimer

THE VIEWS AND STATEMENTS EXPRESSED HEREIN ARE THOSE SOLELY OF ARDAN. THIS ARTICLE CONTAINS PRELIMINARY INFORMATION ONLY, IS SUBJECT TO CHANGE AT ANY TIME AND IS NOT, AND SHOULD NOT BE ASSUMED TO BE, COMPLETE. NO REPRESENTATION IS MADE AS TO THE ACCURACY OR COMPLETENESS OF THE INFORMATION SET FORTH HEREIN.

THIS ARTICLE HAS BEEN PROVIDED TO YOU FOR INFORMATION PURPOSES ONLY AND MAY NOT BE USED OR REPRODUCED FOR ANY OTHER PURPOSE. THIS ARTICLE IS FOR INFORMATIONAL PURPOSES ONLY AND ALL INFORMATION CONTAINED HEREIN IS SUBJECT TO REVISION AND COMPLETION. THIS ARTICLE DOES NOT CONSTITUTE OR FORM PART OF AN OFFER TO ISSUE OR SELL, OR OF A SOLICITATION OF AN OFFER TO SUBSCRIBE OR BUY, ANY SECURITIES OR OTHER FINANCIAL INSTRUMENTS, NOR DOES IT CONSTITUTE A FINANCIAL PROMOTION, INVESTMENT ADVICE OR AN INDUCEMENT OR INCITEMENT TO PARTICIPATE IN ANY PRODUCT, OFFERING OR INVESTMENT. THE INFORMATION SET FORTH HEREIN DOES NOT PURPORT TO BE COMPLETE. THE CONTENTS OF THIS PRESENTATION ARE NOT TO BE CONSTRUED AS INVESTMENT, LEGAL, REGULATORY, BUSINESS, ACCOUNTING OR TAX ADVICE. YOU SHOULD CONSULT YOUR OWN INVESTMENT ADVISER, ATTORNEY, BUSINESS ADVISOR, ACCOUNTANT AND TAX ADVISOR AS TO LEGAL, REGULATORY, BUSINESS, ACCOUNTING AND TAX ADVICE.