Zocdoc CEO: I’ve watched Walmart, IBM, and others try to ‘disrupt’ healthcare. Here’s why they failed

A Zocdoc veteran argues that healthcare cannot be effectively disrupted from outside the existing ecosystem, warning that the current AI cycle could amplify entrenched incentive problems rather than solve them. The article cites failed healthcare ventures including Haven's 2021 closure, Walmart Health's shutdown of 51 clinics in 2024, IBM's sale of Watson Health for about $1B after roughly $4B of investment, and Babylon Health's 2023 U.S. bankruptcy after a valuation above $4B. It advocates for integration-led innovation, noting Zocdoc connects more than 200,000 providers, 10,000 insurance plans and 175 calendar integrations, while patients wait an average of 31 days and 20%-30% of provider appointment openings go unused.

Analysis

The investable read-through is not broad bearishness on healthcare AI; it is a premium shift toward incumbents and vendors that monetize workflow integration, compliance, and distribution rather than consumer-acquisition-led care models. Oracle (ORCL), Epic-private ecosystem partners, UnitedHealth (UNH), CVS (CVS), and Elevance (ELV) possess embedded claims, provider, and clinical-workflow data that make them likely toll collectors if AI reduces administrative friction. The constraint is that value capture may accrue to payers/providers through lower labor intensity, while software vendors face implementation-heavy revenue recognition and elongated sales cycles.

For AMZN and WMT, healthcare should be valued as a strategic retention and adjacency investment, not a near-term standalone margin engine. A renewed consumer-health push would risk incremental losses, regulatory scrutiny, and management distraction unless tied tightly to pharmacy, employer, or existing care-delivery networks; this is more relevant to multiple discipline than to consensus EPS over the next 1-3 months. IBM's historical AI-healthcare experience argues against assigning material upside to broad healthcare-AI narratives without disclosed contract value, renewal rates, deployment scope, and evidence that customers—not vendors—bear integration costs.

Over 6-18 months, AI can improve access and utilization only where it interoperates with scheduling, eligibility, prior authorization, and EHR workflows. That favors administrative automation and revenue-cycle applications over diagnosis-first products, but reimbursement policy and provider liability remain the gating catalysts. Consensus may be too skeptical of incremental productivity gains at scaled incumbents, yet too optimistic on venture-style platforms converting technical capability into durable healthcare economics.

Key Decisions for Investors

• Prefer long ORCL / short a basket of healthcare-AI narrative exposures where available; initiate only after confirmation of healthcare cloud/EHR backlog and margin trajectory. Target a 6-12 month relative-value trade; exit if ORCL's cloud growth decelerates materially or implementation costs pressure operating-margin guidance.
• Maintain a tactical underweight in AMZN and WMT healthcare optionality rather than outright core shorts. Over the next 1-3 months, treat any healthcare expansion announcement as a catalyst to reassess capex and operating-loss exposure; a disclosed partnership with a major payer or health system, rather than a standalone offering, would invalidate the negative read-through.
• Long UNH or ELV versus consumer-facing care disruptors/health-tech ETFs as a 6-18 month quality tilt: scaled payers are positioned to retain a portion of automation savings through claims and prior-authorization workflows. Key risk is adverse medical-cost trend or regulatory action limiting automation-driven utilization management.
• Do not add IBM on a generic healthcare-AI thesis. Upgrade only if management discloses recurring healthcare AI revenue, named enterprise deployments, and measurable services-margin accretion; absent those data, healthcare AI remains insufficient to support a multiple re-rating.