The Current “SaaS Meltdown” Will Fundamentally Change the Tech Stack by 2027

Today’s healthcare leaders are largely being sold “AI-as-a-service”: SaaS products that solve a specific task and connect to other SaaS products, enhanced with AI, as a wave of startups pushes AI into every layer of healthcare delivery 

That’s why we’re seeing proliferating AI pilots, with many health systems still struggling to move from proof-of-concept to scaled adoption. AI hasn’t changed the legacy health tech model of identifying a gap, buying a product for it and adding another product when the next bottleneck appears – all while relying on staff to absorb the gaps between them. 

But soon, a shift will occur from this legacy model to autonomous AI that truly takes on operational burden. We won’t just have new products, but an entirely different way of working, and CIOs who haven’t prepared will be left behind.

Healthcare leaders must recognize this impending shift now and get ready for it by changing how they evaluate technology and the operational tasks they expect it to solve. 

The state of healthcare AI

I’ve been in health tech for over a decade, and the 2024-2025 era of healthcare AI felt oddly familiar. 

Just like a decade earlier, innovative health systems got excited about the possibilities of AI and started snapping up more and more solutions. The main issue with this model is that staff are still required to own every step of the outcome. Some pieces of workflows get better, but those gains are also counteracted by the burden of having multiple additional systems.

Now, the latest buzzy term is “agentic” AI – or AI that can independently take the right action for a specific task. 

A few years ago, I would have been blown away by the idea of agentic capabilities. But with the rate of change in technology today, a focus on agentic AI is missing the point. The shift is simpler and more consequential. Software is moving from performing tasks to completing work.

The old model:

• Software performs a function
• People connect the steps

The emerging model:

• Systems carry the workflow forward
• People step in when something breaks, requires judgment or falls outside policy

By 2027, health systems will be connecting core systems of record to autonomous layers that go beyond specific tasks and instead complete everything needed to accomplish an outcome. If a system produces an output and waits for a person to act, the work has not changed. If a system moves the workflow forward without intervention, it has.

That’s going to produce bigger change than just a better software product to buy. And that change is what leaders need to prepare for.

Is AI adoption meaningfully advancing health systems?

Right now, many organizations may feel stuck on where to go next. If an organization is like many peer systems, it has some pilots in both clinical and operational AI – some promising, and some not quite worth the effort. But what’s the path forward to meaningfully become an AI-powered organization, and not just keep adding pilots to the pile?

It is important to recognize where the industry is today. There are three common categories of AI that may emerge during vendor evaluation:

• AI in name only – Some vendors use AI as primarily a marketing term. The product is largely unchanged, and the roadmap doesn’t incorporate meaningful transformation, but a tool or two have been supplemented with AI.
• AI as an umbrella term – Some vendors frame any combination of machine learning, GenAI, set workflows and agentic work as AI. It is important to understand whether the AI pieces are driving the vendor’s development and vision, or primarily a way to frame the conversation.
• “Agentic” AI – Agentic AI performs an action flexibly and naturally within parameters, making it the closest thing to autonomous AI that is available on the market today. Confirm whether the vendor is using it as a buzzword, or the technology is actually agentic – and if it is, that it’s deeply EHR-integrated and can accomplish the workflows in scope. 

Most of these vendors aren’t equipped to run full end-to-end workflows for health systems. That’s partially due to their functionality addressing a specific piece of a workflow. For example, parsing a fax but leaving it for staff to process doesn’t take on the coordination calls and follow-up that come next. 

But it’s also reflective of an AI industry that’s changing fast. The fully autonomous workflows expected within the next year are not yet available in the industry. By recognizing that the full iteration of what AI looks like in healthcare is yet to come – but coming soon – organizations will be equipped to steer organizations with positive changes now that will make sense as technology evolves.

What does the “SaaS meltdown” mean for  health systems?

The “SaaS meltdown” will bring a shift in expectations, with systems increasingly judged by what they complete. SaaS hasn’t yet gone away, but we’re in the middle of a period of rapid change. So, what does this mean for health systems right now?

First, current pilots should not be stopped, especially if they’re delivering value. But evaluate whether the vendor can evolve beyond improving a single task, and take on more of the workflow over time. Second, if an organization has AI expertise in-house, establish a group that can track how quickly the technology is evolving and help organizations interpret what matters. Finally, prepare a forward-looking approach to tech stacks that will help organizations evolve.

Forward-thinking AI adoption 

With the current state of healthcare AI in mind, healthcare leaders can prepare by:

• Making a shortlist of vendors to keep an eye on over the coming months. Include any vendors the health system is using today, vendors who approach problems in a way that resonate with the health system and those liked but not chosen in past evaluations. 
• Asking for roadmaps from these vendors and evaluating whether they plan to continue “AI-as-a-service” (solving individual problems, but with AI instead of traditional SaaS) or have a more expansive plan for end-to-end AI.
• Getting in the weeds on AI metrics and how to really know if they’re making a difference. For example, a voice AI vendor might advertise a 90% containment rate for calls but go further by asking what skills that containment rate applies to and running the math on how that reduces labor costs at organizations. 
• Review vendor contracts to ensure that they reflect AI (not SaaS) usage patterns and get familiar with how costs are calculated. As AI embeds into every touchpoint — scheduling, intake, follow-up, referrals — a few cents per interaction across millions of patients becomes significant. 

It’s an exciting time for healthcare technology, but as talk of the SaaS “meltdown” suggests, healthcare isn’t insulated from the seismic shift created by AI. The proactive healthcare leader understands that we’re not at the end of this shift, but in the middle of it, and prepares accordingly.

Photo: J Studios, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.