In the last few years, artificial intelligence (AI) in healthcare operations has become a default talking point across nearly every health system, shaping how leaders approach technology and strategy.
Consultants use it. Vendors promise it. Leaders repeat it like a mantra.
Nearly everyone claims to be AI-first, often with a confidence that far outpaces results.
But being AI-first does not deliver the advantage some predictors promise. After decades in healthcare and operational leadership, I’ve noticed a clear pattern in how organizations adopt new technology, and it applies directly to the challenges AI presents payers today. Organizations that adopt AI-first healthcare strategies often struggle to translate ambition into operational results. The technology does not fall short; instead, the definition and application of AI implementations create challenges.
When teams prioritize AI, the AI system becomes the starting point rather than a capability layered onto proven operations. Experience, discipline, and ownership of outcomes are treated as downstream concerns rather than the foundation that makes technology effective. The result is sophistication without performance.
If you’re an operations leader at a health payer, this distinction matters a lot. AI-first is a buzzword, not a strategy. And strategies built on buzzwords rarely survive contact with real operational complexity.
In this post, I explain why AI-first strategies fail and what works instead in healthcare operations.
The hype cycle and healthcare’s love affair with terminology
Buzzwords have a way of promising what we want to happen rather than what will happen.
In healthcare, AI-first has become shorthand for administrative automation, predictive modeling, risk stratification, and workflow optimization.
These are real challenges for payer operations that need a solution. But for all the talk, relatively few organizations have moved beyond AI pitch decks into consistent operational improvements.
This is because most AI-first pitches position the solution in this order:
- Technology comes first
- Data magically congeals into insights
- Outcomes follow
However, that’s not the recipe for transformation that the healthcare industry needs. Many leaders assume modernization must be disruptive. They make a similar mistake by believing AI can come before experience. It can’t.
AI has a place in operations, but it isn’t first.
AI without institutional memory is just automation theater
Let’s be clear—I have enormous respect for what AI can do. Machine learning, natural language processing, and automation are powerful. They can eliminate low-value work, surface patterns humans miss, and accelerate decision cycles.
Beyond deploying AI tools, success depends on understanding the problems and having the experience to interpret results responsibly.
AI amplifies what already exists. It’s the critical missing piece in most AI-first strategies. If your data is noisy, incomplete, or not tied to operational outcomes, AI amplifies the noise.
AI does not fix foundational issues. It magnifies those issues when teams leave them unresolved.
This is a point I’ve emphasized when working with teams transitioning to new technology. Without operational discipline and clear metrics, tools yield chaos rather than leverage. Too often, companies chase shiny solutions without first tightening the fundamentals.
And when that happens, you get tech that looks smart but doesn’t help.
Experience beats algorithms without context
The value in AI and data science comes from experience-informed interpretation. AI on its own won’t get the industry very far. AI requires human expertise to turn insights into meaningful outcomes.
Consider an algorithm that can predict which claim might be recoverable. Highly valuable. But it cannot consistently understand and apply complex legal frameworks, navigate nuanced member communications, adjust strategy based on shifting payer policies, or prioritize timeliness over theoretical yield.
Those require human judgment informed by decades of operational experience. The right AI amplifies the human experts already in your organization. It identifies opportunities so that experts can validate, contextualize, and align them with what actually moves the needle.
Without that context, AI is detailed but directionless.
The danger of proof by performance
If you’ve ever sat in a vendor demo, you’ve probably heard language like:
- “We recover 10x more than market average.”
- “We increase ROI by 500%.”
- “Our AI discovers cases no one else can.”
Those claims sound compelling. The numbers would mean significant savings for your organization. But they leave out a critical operational question: What baseline are you comparing against?
In my experience, the boldest claims often rely on narrow samples, cherry-picked populations, or misleading denominators. They look good in a slide deck, but they don’t hold up when you dig into the real business context.
For operations leaders, the priority should not be the biggest number, but the most credible number.
It seems simple, and it should be. Numbers should withstand scrutiny, apply across portfolios, and reflect complete cycles from identification to recovery to reconciliation.
That’s why, if you’re evaluating vendors, you should always interrogate how they construct their metrics.
Where AI creates value when it’s applied correctly
AI proves effective when teams apply AI-driven, AI-powered approaches to support clear goals that improve patient outcomes. It’s not a replacement for rigorous operations.
Here’s what successful applications look like:
1. AI that frees humans to add human judgment
In many clinical workflows, from claims review to member outreach, the low-value work is repetitive, rules-based, and predictable. This same pattern shows where AI in clinical settings delivers value. It also shows how clinicians review and prioritize data from medical devices. The high-value decisions, like prioritization, negotiation, and judgment calls, still require experienced professionals.
That’s the sweet spot.
Instead of asking how to automate, ask:
- What parts of this workflow do not require human judgment?
- What parts do require it?
- How can AI remove friction without removing accountability?
When we think like that, we reduce resistance and increase adoption.
2. AI anchored in reliable data
I’ve seen companies build AI on data that is inconsistent or poorly documented. This is dangerous because AI amplifies what already exists, not just from an operations standpoint but within a dataset.
A model trained on poorly labeled outcomes will predict poorly. That’s why you should ask every AI vendor what data trained their model and how they validated the results.
If the answer is vague, incomplete, or unavailable, be skeptical. Leading AI adopters in healthcare should spend as much time refining their data pipelines as they do training their models.
I explore how automation makes healthcare experts more valuable in another blog post.
3. AI aligned with clear operational objectives
Too often, organizations adopt technology for technology’s sake. They want to signal innovation. They want to avoid falling behind. That’s not a strategy.
A strategy starts with a clear objective such as doubling recoveries, reducing cycle times, reducing costs, or improving member satisfaction. Then, it asks how AI can help to achieve that objective with measurable fidelity.
That anchor prevents technology for technology’s sake and positions AI as a tool for delivering on the mission of the organization.
Why most AI disappointments aren't technical
When an AI initiative fails to deliver impact, leaders usually jump to technical explanations. Maybe the model wasn’t sophisticated enough or we didn’t have enough data. Maybe the vendor’s tech wasn’t mature.
In my experience, the real culprits are almost always organizational:
- Misaligned expectations
AI is a lever, but you still need a fulcrum. AI amplifies capability, speed, and scale only when clear processes, skilled teams, defined ownership, and outcome discipline already exist. - Data without operational grounding
If your data isn’t tied to real outcomes and rigorously vetted, your model’s predictions won’t be either. - Lack of end-to-end accountability
AI can produce insights, but teams must be empowered to act on them and measured on results in order to capture its efficacy.
A better framework for AI adoption
The most successful AI implementations treat AI technologies as operational accelerators, not substitutes for discipline or accountability. So, what should operations leaders focus on?
Here’s a framework that has worked consistently from my vantage:
1. Define the operational outcomes first
Before you implement anything, clarify the outcome you want without prescribing the technology you hope will get you there.
Is it higher recovery rates? Lower rework? Faster cycle times? Fewer manual touches?
Name it. Then measure it.
2. Clean and contextualize your data
AI is only as good as the training data you give it. Prioritize complete outcomes, consistent labeling, and contextual metadata.
This often takes longer than most payers expect. It’s necessary. It yields more precise, meaningful outcomes.
3. Automate with judgment
Automation for its own sake is an easy trap to fall into. A good rule of thumb is to automate drudgery, but don’t replace expertise. That’s how you drive adoption and reduce churn.
4. Measure performance in operational terms
Whether a model is accurate should be a given. A better question is whether it improves the outcomes that matter to your organization.
How this connects to recoveries and workforce strategy
At its core, the issue with AI-first approaches is conceptual. If you believe technology is the strategy, you’ll optimize for technology. If you believe operational execution is the strategy, you’ll optimize for outcomes, with technology as the amplifier.
That distinction is crucial for two reasons:
1. Recoveries are not won by algorithms alone
Identifying recoverable opportunities is important, but it’s only the first step. Converting those opportunities into real dollars matters even more.
AI can help flag cases, but it takes disciplined workflows, legal expertise, and operational prioritization to realize value.
2. Human work still matters
When discussing technology deployment, I make it clear that technology should free analysts and operators from routine. Removing these low-value tasks allows them to use their expertise to exercise judgment.
Automation done right makes experts more valuable, not obsolete. The future of AI in healthcare operations depends on the thoughtful development of AI to strengthen human expertise. Not replace it.
How to use AI for healthcare operations
AI improves healthcare operations when it is applied after operational outcomes, workflows, and accountability are clearly defined. Organizations that treat AI as a starting point typically see high technology spend and lower operational impact.
Why AI-first fails in healthcare operations
Healthcare organizations that adopt an AI-first strategy (considering AI ahead of operations) often encounter:
- Advanced tools without measurable performance improvement
- Teams struggling to integrate AI into real workflows
- Leadership misalignment on accountability and return on investment (ROI)
This happens because AI is an enabling capability, not an operational foundation. Without clarity on outcomes and execution, AI has nothing to optimize.
What comes before AI in healthcare operations
Successful AI adoption requires operational readiness. That includes:
- Clearly defined clinical, financial, or administrative outcomes
- Ownership for acting on insights
- Processes that reflect how work is actually done
- Metrics that track performance, not activity
AI amplifies these elements, but it cannot replace them.
How AI should be used in practice
In effective healthcare operations, AI is used to:
- Surface patterns and opportunities within defined workflows
- Prioritize work based on operational and financial impact
- Accelerate decision-making where judgment already exists
- Scale execution that is already working
When introduced this way, AI increases speed and consistency without disrupting accountability.
It’s good news for our talented experts. Organizations that pair AI with operational discipline and outcome ownership will shape the future of healthcare operations. Leaders who understand this will win because they treat technology as a tool, not a strategy.
Machinify partners with healthcare payers to modernize their technology so that every claim is paid right the first time, on time, every time. To learn how to get it right the first time, follow our series.
Ryan Little—President, Right Payer Solutions—brings broad experience across both healthcare and client advisory services, spanning a variety of leadership roles, including operations, finance, and mergers and acquisitions. Ryan held several leadership roles at The Rawlings Group, including Executive Vice President and Chief Executive Officer.
He also served as Senior Vice President at the Presbyterian Foundation Group and Managing Director in Stern Stewart & Company’s New York and Tokyo offices as a corporate finance consultant across mergers and acquisitions, corporate governance, and financing strategies.
Ryan began his career in investment banking with Alex, Brown, & Sons. Ryan graduated from the University of Virginia and earned an Master of Business Administration from the University of Louisville.
The content on this page is subject to our Terms of Use.
