Agentic AI in Healthcare: The Rehumanizing Effect

Sixty-eight percent of healthcare providers have already deployed AI agents. The organizations that are getting measurable outcomes are not using them to replace clinical staff — they are using them to eliminate administrative work that was consuming clinical time.

What Happened

MIT Technology Review profiled Hospital for Special Surgery (HSS), a New York academic medical center, which partnered with enterprise agentic AI developer Ema Unlimited to automate two categories of administrative work: insurance claims processing and patient scheduling.

The claims results are specific. Before agentic AI, processing 1,100 claims monthly required significant manual effort with an appeal success rate of 65% and an average handling time of 45 minutes per claim. After deployment, the same volume processes with an appeal success rate of 100% and a handling time of 5 minutes — a 9x time reduction on a task that previously required trained staff to navigate payer documentation, prior authorization rules, and appeal templates.

The scheduling deployment runs as a 24/7 conversational AI that books appointments based on patient condition, location, insurance coverage, and physician availability. The WHO has warned of an 11 million healthcare worker shortfall by 2030. Dr. Ashis Barad, HSS Chief Digital Officer, characterized AI’s role through the infrastructure lens: “It’s a general-purpose technology, analogous to electricity.” His projection is that 90% of non-clinical tasks could eventually be AI-administered.

KPMG research cited in the article puts overall healthcare AI agent adoption at 68%, with 84% of providers reporting they are comfortable delegating specific process decisions to AI agents.

Why It Matters

The HSS results illustrate a pattern that healthcare AI has been promising but struggling to demonstrate at production scale: measurable administrative efficiency gains that free clinical staff for patient interaction rather than documentation. The 100% claim appeal success rate is particularly notable — it suggests the agent is not just faster but more consistent, applying the same documentation standards across every claim rather than depending on staff experience or attention levels.

The “rehumanizing” framing from the article’s title is accurate in a specific sense: by eliminating keyboard work from clinicians’ days, these systems shift what Dr. Barad calls “white-glove work” — complex, sensitive patient interactions — back toward the humans whose judgment and presence actually matters there.

For agentic AI builders outside healthcare, the HSS deployment demonstrates that process automation at this level requires deep integration with domain-specific data sources (payer rules, authorization databases, scheduling constraints) and cannot be accomplished with general-purpose agents out of the box. The meaningful results came from a purpose-built integration, not a generic workflow tool pointed at healthcare.

Source: MIT Technology Review

Content created with AI assistance and reviewed for accuracy.