CommonSpirit CIO Daniel Barchi: Don’t forget the ‘people side’ of AI

Daniel Barchi, CIO of CommonSpirit Health, has a hierarchy for everything, including AI adoption.

Speaking at the Becker’s 16th Annual Meeting, Mr. Barchi said CommonSpirit’s approach to artificial intelligence through the lens of Maslow’s hierarchy of needs, with core technology infrastructure at the base and AI-enabled “transcendent healthcare” accessible only after every foundational layer is in place. The core message? Tools don’t transform organizations. People, process and a disciplined foundation do.

“Healthcare technology is really 80% people,” Mr. Barchi said. “It’s about 15% process, and it’s really only about 5% technology. And that’s true even with AI. AI operating on its own is not going to deliver better care. It’s not going to have a better outcome for that patient with the nurse standing by the patient. It’s only when we’ve got all those other things working in concert are we going to use AI in a really thoughtful way.”

CommonSpirit is one of the largest health systems in the country — roughly $40 billion in revenue, 150-plus hospitals across 24 states and about 150,000 employees. At that scale, the choices made about AI matter. As of fiscal year 2026, the system has approximately 250 active AI tools, up from around 60 in fiscal year 2023. But that number is not the point.

“250 is not better when it’s 400, and it’s not even better than that when it’s 800,” he said. “At some point we’ve got enough tools. We need to go deeper with them, we need to get enterprisewide. We need to start getting better outcomes for our patients with those tools.”

He said the system deliberately shifted away from vendor point solutions, the small startup contracts for individual hospitals, toward embedded AI tools within platforms CommonSpirit already operates.

“It is so much easier to govern, to use, to train and get going in the workflow,” he said.

Internally developed tools are also growing as a share of the portfolio, though Mr. Barchi acknowledged that CommonSpirit is not a software development shop. The internal builds are guided by a simple question: what’s good enough?

“We’ve got a length-of-stay problem in one of our regions,” he said. “We want to go out and buy a small tool that’ll do it for $5 million for five of our hospitals. Let’s just go to our team and say, ‘Hey, look, we’ve got the EHR, we’ve got the data, we’ve got our nursing leaders who know how to do this. What they need is X. Can you develop X?’ The team said, ‘Give me three weeks.'”

The resulting tool may last 18 to 24 months before needing revision. It will not be sold. But it solves a real operational problem without the overhead of a formal procurement.

Among the clinical AI use cases Mr. Barchi highlighted, one stood out. CommonSpirit began piloting an AI-driven tool to screen emergency department patients for signs of human trafficking, starting in a handful of Arizona hospitals and now accelerating to more sites across the system.

“We see 2 million-plus patients every year. We see the best of humanity and the worst of humanity,” he said. “Very, very vulnerable people come through every one of our health systems, especially in our EDs. We would like to say that we screen every one of these patients for risk factors that they might be a victim of human trafficking. And we were only able to get to a certain percent of that before now using AI.”

The system has since identified patients who were being trafficked, including minors.

“Yes, it’s nice to save money, it’s nice to make ourselves more efficient, it’s nice to have a better patient interaction,” Mr. Barchi said. “But when we’re actually impacting lives and human safety in that way, it makes you proud of what you’re doing every day.”

Other clinical use cases include an internally developed sepsis detection model that has contributed to saving thousands of lives over more than five years, AI-assisted lung, colon and breast cancer screening that automates candidate identification for ordering physicians, and third-party tools to accelerate stroke pathway evaluation. On the administrative side, CommonSpirit has deployed AI for talent acquisition, automating candidate screening and conducting pre-screening interviews before handing the top candidates to human recruiters.

One area that hasn’t delivered on the promised financial benefit is ambient listening technology. While clear soft ROI improves clinician satisfaction and “pajama time,” the potential financial benefits haven’t materialized.

“I find it really interesting that this is where you hear a lot of people talking about it, and people thought that there was going to be a lot of financial benefit from this,” he said. “Sure, there’s some financial benefit, but really it frees up our clinicians to be more focused on the patients right in front of them, automate a lot of the back end, make them happier, more engaged clinicians. We’re not finding or looking for a lot of financial benefit. Although a lot of people will try to sell ambient AI as having a financial impact, we find it’s better for our patients and our clinicians.”

The area where AI is driving more measurable financial return is in clinical documentation and revenue cycle, specifically charge capture and accounts receivable mapping. But Mr. Barchi’s framing consistently subordinated cost outcomes to mission outcomes, echoing CommonSpirit CEO Wright Lassiter’s stated priorities: safety and quality first, then caregiver experience, then consumer experience, and cost reduction last.

CommonSpirit uses a five-gate process to validate AI ROI claims before attributing savings to any tool: the business case, evidence review, finance sign-off, implementation and post-implementation measurement.

“Instead of handwaving it — gosh, we saved $10 million using AI — you can actually go in and say for every one of the AI tools and programs you’re putting into place, what have your clinicians, your finance people and your operators actually said you were going to achieve, and you actually achieved afterwards,” Mr. Barchi said. “That’s a discipline we often don’t have when we get excited about technology.”

To keep AI usage within guardrails, CommonSpirit built an internal large language model interface called Insightly, available to all 150,000 employees, that allows staff to use models from OpenAI, Google and Anthropic without exposing patient data to public-facing tools. The system also blocked employee access to consumer AI platforms and flagged certain tools, including some China-based products, as security risks early on.

Governance is organized through an enterprise data and AI governance committee, known internally as EDAG, that meets biweekly and includes representatives from nursing, privacy and compliance, security, legal, ethics, IT and physician leadership. The committee reviews every tool and decides whether to move forward.

He mentioned Moderna as a cautionary tale. The pharma company, energized by AI momentum post-pandemic, eventually found itself managing 1,400 AI tools running simultaneously.

“Imagine trying to govern 1,400 tools,” Mr. Barchi said. “Imagine going to your board and saying, hey, we’re using AI. And they’re like, what are you doing? I don’t know. We’ve got 1,400 things that are going.”

CommonSpirit has also drawn a firm line it will not cross: no autonomous diagnosis or action without a clinician in the loop.

“There will always be a clinician between our AI and the patient,” Mr. Barchi said. “We never want to get to the point where there’s autonomous diagnosis and action, where a human, especially a well-trained doctor or nurse, is not in the loop making the decision.”

But he also pushed back against what he characterized as overindexing on AI risk.

“Probably more people have been killed in hospital parking lots than have been harmed by AI in the past two years — certainly in the past five years,” said Mr. Barchi. “But think about that. We don’t talk about parking lot harms, even though we all have parking lots and parking garages. But we’re over-indexing on the risks of AI.”

The second half of Mr. Barchi’s remarks focused on people. He challenged every person in the room to audit their own AI usage, offering a rough formula: 1,000 divided by years of experience plus 10 equals the minimum number of times per week a healthcare leader should be using AI. By that math, a senior leader with 35 years of experience should be using AI roughly 20 times per week. Mr. Barchi said he used it about 30 times the previous week.

“Either you’re in the quadrant that is leaning into AI and going to be one of those leaders, or you’re lagging and it’s going to impact yourself and your career,” he said. He said in a recent conversation with a vice president on his IT team who asked for three new analysts to analyze spreadsheets and develop a plan, and didn’t know AI could handle the task.

“Don’t assume that we’re surrounded by people who know AI like the back of our hands,” he said. “We’re all learning.”

To build baseline competency across its workforce, CommonSpirit launched an AI Academy with a capability pathway — AI literate, practitioner, activator — and is offering online courses and digital badges to employees at all levels. The system also held a senior leadership session where executives were walked through basic use cases on Insightly, including prompts as simple as writing a recipe from pantry ingredients.

“When I say basic learning, it’s okay to not be too proud about the fact we’re doing it,” Mr. Barchi said. “Not everybody’s using it at an advanced level. Getting that basic understanding as leaders allows everybody to lean in and get the long-term benefits that we need.”

CommonSpirit has made its AI operational governance dashboard public, tracking the number of tools, agentic agents, solutions in development, and other metrics, and he said he would welcome wider adoption of a similar framework across the industry.

“We’re being very open and transparent about how we’re using it, both internally and with our patients, because we want to be absolutely true to ourselves and to our mission,” Mr. Barchi.

The post CommonSpirit CIO Daniel Barchi: Don’t forget the ‘people side’ of AI appeared first on Becker's Hospital Review | Healthcare News & Analysis.