Dive Brief:
- Healthcare is still embracing AI, but unpredictable costs and increased technical complexity are making it hard to move from pilots to enterprise-wide adoption, according to a new report published by Google Cloud.
- Nearly 60% of surveyed healthcare executives say they’ve faced integration and deployment challenges when trying to shift enterprise AI from the pilot stage to a live system, and 40% cite unpredictable costs as a reason for pausing AI initiatives.
- To Aashima Gupta, global director of healthcare at Google Cloud, addressing those cost concerns as an industry is crucial to successfully scaling AI in healthcare. “As AI moves from pilot to at-scale, the economics of AI comes into the forefront. You don't see that when you're running a 100-person or 200-person pilot,” Gupta said.
Dive Insight:
Healthcare has been notorious for lagging on modern tech adoption compared to other industries — it faces a complex regulatory environment, a reliance on legacy systems and limited budgets. But when AI became generally available, healthcare organizations quickly embraced it, leveraging the tech for everything from back office administrative tasks to clinical documentation.
A few years into the AI craze, models are more complex, use cases have expanded and costs remain variable. Just 30% of healthcare organizations report successfully scaling advanced or agentic AI across their operations, according to Google’s survey of more than 250 healthcare executives conducted by consumer insights firm Hypothesis Group. The sector trailed the benchmark for all industries by 13 percentage points.
Cost and complexity appear to be leading barriers to enterprise-wide adoption, according to the report. More than 80% of respondents cite cost as a key hurdle to reducing dependence on existing AI vendors, which can be a bottleneck for innovation.
And healthcare leaders may be more cautious about taking on AI risk. About 40% of respondents named protection of sensitive data as an especially important consideration when it comes to AI, nine percentage points higher than the benchmark.
Respondents expressed a desire to have full visibility into an AI tool’s governance process, as well as the ability to customize tools.
Gupta says these trends are a sign that a one-model-fits-all approach is not the right solution for healthcare as AI advances.
“Healthcare is not one workflow — healthcare is many workflows. And each workflow has a different level of sophistication,” she said. “You don't need a supersonic plane to do the simple task of going from one town to another.”
As such, Google is pushing an approach that favors using different AI models for different use cases, rather than championing a single model in an industry as complex as healthcare.
The tech giant recently launched Gemini for healthcare, a version of its AI agent designed to be used by providers and payers. It touts the platform’s integrations with third-party vendors like AI scribes and electronic health record vendors.
The platform also works across Google’s own competitors, enabling users to pair AI models across Google, OpenAI and Anthropic under the same governance umbrella.
It’s the company’s latest play to stand out in a competitive market saturated with AI tools — both Anthropic and OpenAI have released healthcare-specific tools too, though Google claims its open model approach sets it apart.
Still, Gupta acknowledges that cost remains a hurdle as AI in healthcare gets more complex.
A tool that seems affordable at the pilot stage might not be when it’s scaled up. And hospitals dealing with thin margins and looming federal funding cuts can’t afford to confront unpredictable costs.
Gupta says healthcare should adopt strategies for “tokenomics,” the discipline of measuring and managing the cost of AI.
Tokens are the units of data that AI models use, and understanding their variable costs can save organizations money. A report from Accenture found that over a two-year period, companies that adopted tokenomics strategies avoided a projected 44% increase in token costs.
“I would even go as far to say healthcare needs a chief token officer. Because who's managing that? You need to have a strategy and budget for it,” Gupta said.
“I believe for the field to advance, we need to get this part right because even if AI will solve problems, it will not be able to scale because of the cost constraints.”