A new HIMSS Global Market Insights study just confirmed what many of us in the industry already feel: healthcare AI is advancing – but unevenly, and for most organizations, too slowly.

The study surveyed 345 healthcare executives and IT leaders across North America, EMEA, and Asia Pacific. The headline finding is both striking and familiar: two-thirds of provider organizations are still in the early stages of AI adoption – either piloting tools or trying to scale pilots that weren’t designed to scale in the first place.

Only 10% of organizations globally have reached advanced, enterprise-wide AI deployment.

They have a name for what the other 90% are experiencing: pilot paralysis.

The Pilot Graveyard Is Real

Over the past several years, healthcare organizations have collectively run thousands of AI proof-of-concepts. Some were genuinely impressive demonstrations. Most never made it to production.

Why? The usual suspects: legacy systems that weren’t built for AI integration. Data that’s siloed, inconsistent, or locked behind compliance walls. Governance frameworks that didn’t exist when the pilot launched and weren’t retrofitted before it tried to scale. And underneath it all, a fundamental question that wasn’t asked early enough:

“Where is the value?”

That question – typically the first one a CFO asks – has quietly killed more AI initiatives than any technical failure ever did. Because when you can’t answer it clearly, you can’t justify the investment. And when you can’t justify the investment, the pilot stays a pilot forever.

The HIMSS study provides data to back this reality. It found that while 71% of organizations run pilots before full rollout, most lack the unified governance model, enterprise data foundation, and operating model needed to scale. As Atlantic Health CIO Sunil Dadlani put it in the report: “The real challenge is not a lack of ideas; it is the inability to scale them.”

The Mindset Shift That Changes Everything

The organizations that are breaking out of pilot paralysis share a common trait: they stopped asking “What can AI do?” and started asking “What problem are we solving?”

That sounds simple. It isn’t. The first question leads you toward technology exploration – demos, sandboxes, vendor showcases. The second question leads you toward operational outcomes. And operational outcomes are the only thing that survives contact with a CFO, a Board, or a health system under real budget pressure.

Success in AI deployment isn’t measured in models deployed or APIs integrated. It’s measured in three things:

• Efficiency gained – fewer manual steps, less staff time on low-value tasks
• Revenue protected or recovered – reduced no-shows, faster scheduling, better prior auth
• Patient outcomes improved – access, follow-through, engagement

Everything else is a demo.

The Most Profitable AI Is Also the Most Boring

Here’s something the industry doesn’t talk about enough: the AI implementations generating the most durable value aren’t the ones making headlines.

Nobody writes a press release about invoice reconciliation automation. Nobody posts a LinkedIn celebration about a 3% reduction in logistics costs. But in a health system processing millions of transactions, 3% is millions of dollars. It’s real. It’s measurable. It compounds over time.

The same principle applies in patient-facing operations. Contact center optimization – automating the routine calls, routing the complex ones, reducing handle time, increasing first-contact resolution – doesn’t look exciting on a conference slide. But it directly reduces operational cost, improves patient experience, and frees clinical staff to do what only humans can do.

This is operational AI. Not glamorous. Not AGI. Not the stuff of TED talks. But it’s what actually works at scale, today, in real health systems with real constraints.

AGI may arrive someday. But healthcare organizations have budgets to manage, staff to retain, and patients to serve right now. The winners in this market won’t be the organizations that chased the most sophisticated model. They’ll be the ones who embedded AI into workflows where it could deliver consistent, measurable, and repeatable value.

What the HIMSS Data Tells Us About What’s Next

The HIMSS report highlights another important finding: 96% of organizations keep clinicians as the final decision-makers. AI in healthcare, at least for now, is assistive – surfacing information, reducing administrative burden, handling the transactional layer so clinical expertise can be applied where it matters most.

That’s not a limitation. That’s the correct design.

The report also notes that nearly half of all organizations (48%) are pursuing a hybrid approach – mixing vendor tools with internal builds. That makes sense. No single vendor can solve every problem, and no internal team has infinite bandwidth. The smart move is to buy what accelerates standard workflows and build (or configure) what differentiates your organization.

The governance gap, however, remains the most urgent issue. Only a third of organizations globally have formally implemented AI policies, and very few have achieved a mature, audited governance state. For North American health systems, this is the work to do now: not more pilots, but governance, data infrastructure, and operating models that enable existing pilots to become production systems.

Where Zappix Fits

At Zappix, we’ve built our platform around a simple conviction: AI that doesn’t deploy at scale isn’t AI. It’s a slide deck.

Everything we do is oriented toward operational AI that works at scale – practical, measurable, and deployed in real health system environments. Our focus is the contact center and patient engagement layer: the high-volume, high-friction workflows where automation has the clearest ROI and the shortest path from pilot to production.

We’ve seen firsthand what moves the needle. At one Zappix healthcare client, our automation doubled booked appointments and reduced no-show rates by more than 50%. At a second client, patients completed digital intake forms at an 88.6% rate – not because the technology was impressive, but because it was designed around the actual workflow.

That’s the standard we hold ourselves to. Not “What can AI do?” but “What problem are we solving, and can we prove we solved it?”

The HIMSS data confirms that most of the industry is still figuring out how to make the same shift. We think that’s an opportunity – and a responsibility.

Source: HIMSS Global Market Insights, “From Pilot to Production: AI Governance, Risk Management and Regulatory Readiness Across Global Healthcare,” March 2026.