The removal of humans from caregiving could potentially jeopardize outcomes.

      The typical pitch deck for digital health goes as follows: AI substitutes the clinician, costs decrease, access increases, outcomes improve, and everyone benefits. This pitch has proven effective, attracting billions in venture capital to companies that claim removing humans from the care process is feasible and advantageous. However, there’s a significant issue with this premise, highlighted by a specific statistic.

      Ruben Sandoval Davila, the co-founder and CTO of Avena Health, presents the concern: just two percent of active users remain on a clinical nutrition platform after three months. This platform, Avena Health, had previously managed to retain 40% of its patients for eleven months or more, according to its internal data. Sandoval, who created the automation and subsequently redesigned the system after its failure, shares this story as a personal lesson. While the data hasn't been independently verified, the precision of the claim, coupled with Sandoval's candid acknowledgment of his error, adds credibility that broader statistics might lack.

      This data point is unsettling for the industry as it comes from within. Sandoval didn't critique another's automation attempts; he conducted the experiment using his product and users, yielding clear results: full automation led to a loss of patient engagement. While the AI executed its clinical tasks accurately, patients disengaged.

      “Healthcare doesn’t scale because providers are bottlenecked by time, not demand,” Sandoval stated. “Our approach alleviates that limitation. A provider's capacity is not restricted by their skills or willingness, but by the administrative burdens associated with each patient interaction. By reducing that overhead intelligently, providers can manage more patients without compromising the quality of each interaction.”

      At first glance, this sounds like a call for automation, but it actually advocates for a more nuanced and complex solution. Sandoval argues that the bottleneck is real, the AI claim is partially accurate, and the fundamental flaw most companies make is the belief that efficiency and human involvement are inversely related. His data indicates that is not the case. The most efficient version of his platform, which achieved the forty percent long-term retention rate, involved keeping human specialists engaged at key clinical interactions while automating other processes.

      “Fully automated AI systems often struggle with retention,” he affirmed. “The advantage lies in merging hyper-personalization at the patient level with expert guidance.”

      The key insight is not that humans outperform AI in clinical care, but that patients behave differently when a human is part of their care, even if the AI handles most tasks. The forty percent retention rate reflects patient behavior rather than clinical quality, and that behavior directly affects whether a digital health platform can survive beyond its first year.

      This issue is critical because the digital health sector faces a retention problem it rarely discusses openly. Various consumer health apps—fitness trackers, mental health platforms, nutrition programs, chronic disease management tools—share a lifecycle of rapid uptake followed by swift abandonment. A 2023 analysis by the IQVIA Institute found that the average digital health app loses over seventy-five percent of its users within the initial two weeks. Many venture-backed companies favor reporting user acquisition numbers over retention numbers, likely because the latter do not reflect well on performance.

      In response to this challenge, the industry often resorts to strategies borrowed from consumer tech, such as gamification, streak counters, push notifications, and behavioral nudges. Sandoval took a different approach. He questioned whether the issue was truly engagement or rather a more fundamental lack of human connection in the care experience.

      His system design incorporates human specialists as strategic assets rather than cost centers to be eliminated. Specialists focus on actions that foster the ongoing engagement the platform requires. Meanwhile, the AI manages everything else.

      “Agencies capture attention through marketing,” Sandoval pointed out. “Experts produce results. The leverage comes from equipping each expert with the tools to control and scale their impact.”

      This perspective clarifies why the platform centers on the specialist rather than the patient or the platform itself. The specialist acts as the retention driver, while the AI serves as the efficiency tool. This architecture balances both elements, having been crafted by Sandoval based on his previous failure analysis and refined over years of practical application.

      The implications of this approach reach beyond any single organization. If the primary unresolved issue in digital health is not clinical capacity but sustained patient engagement, then companies emphasizing full automation may be misidentifying the problem. They are developing products that are efficient, scalable, yet lack substance. While automation functions, users tend to depart.

      Sandoval is now preparing to validate this theory in the market where AI hype is strongest and the retention issue is often overlooked. Alva Health, the new platform his team plans to launch in the U.S., is based on the same hybrid architecture that improved Avena's retention rate after the automation setback. The U.S. digital health landscape is filled with products that aim to eliminate human involvement. Sandoval's strategy counters this by asserting that the absence of human connection disrupts the care cycle.

      The two percent statistic reflects