The healthcare industry has been talking about automation for decades.
The 2025 CAQH Index quantified just how much progress has been made: an estimated $258 billion in administrative costs avoided through electronic transactions and improved data exchange. That’s a number worth celebrating.
But the same report documents something that should stop every CDO, CIO, and operations leader mid-sentence: a remaining $21 billion in savings still waiting to be unlocked through full automation of manual and partially manual transactions.
That number gets cited in board presentations. It shows up in vendor pitches. It anchors capital investment proposals. What gets cited far less often is why that $21 billion keeps remaining out of reach, year after year, report after report.
The Technology Isn’t the Problem
The disconnect isn’t a shortage of tools. AI platforms, automation engines, claims processing systems, and workflow optimization software are readily available and widely deployed across the industry.
Experian Health’s 2025 State of Claims survey found that 67% of providers believe AI can improve the claims process. Only 14% have actually implemented AI tools. That gap isn’t primarily about skepticism or budget. It’s about a more fundamental problem: organizations know that the data feeding these systems isn’t trustworthy enough to automate against.
AI systems don’t fail because of bad algorithms. They fail because of bad inputs.
Harvard Business School research put it plainly: “if you put in garbage, the AI tool will produce something that’s suboptimal.” The study found that when input data was flawed, managers overrode AI-generated decisions on 84% of tasks, not because the AI was poorly designed, but because the data it was given couldn’t be trusted. Every bad input triggered a ripple of manual corrections that consumed more time than the automation saved.
Healthcare organizations are living inside that exact pattern right now.
The Most Consequential Bad Input Is Provider Data
When AI and automation tools ingest credentialing records, roster files, directory data, and claims histories, they’re working with provider data as their foundation. And that foundation has a well-documented accuracy problem.
The 2025 CAQH Index found that the industry spends enormous resources on administrative transactions that remain manual or partially manual, not because automation is technically impossible, but because the underlying data can’t reliably support automated decisions. When provider records contain invalid or inaccurate NPIs, mismatched taxonomy codes, outdated network participation status, and conflicting practice addresses across systems, automation doesn’t eliminate the manual work. It relocates it. Claims fall out of auto-adjudication and into queues. Verification requests trigger staff intervention. Exceptions accumulate. The manual labor that was supposed to disappear shows up downstream, more expensive and more difficult to trace back to its source.
This is what the $21 billion gap looks like from the inside. It’s not a technology gap. It’s a data quality gap wearing a technology uniform.
Why Manual Work Persists Despite Available Automation
The CAQH Index has documented more than $90 billion in annual administrative spending tied to routine transactions. The reason so much of that work remains manual isn’t a lack of automation tools. It’s that manual verification is being used to compensate for data that can’t be trusted.
Staff are checking, re-checking, and manually reconciling because the underlying provider records haven’t been unified, validated, and continuously synchronized across systems. The automation exists. The confidence in the data to drive that automation doesn’t.
Every percentage point of auto-adjudication that requires manual intervention is a signal. It means a provider record somewhere in the system didn’t resolve cleanly. The NPI didn’t match. The taxonomy code didn’t align with the credentialed specialty. The network status was current in one system and stale in another. The automation found the inconsistency and routed it to a human. That human costs far more per transaction than any subscription to a data infrastructure platform.
The Organizations Making Real Gains
The pattern is consistent among health plans and healthcare operators that have made meaningful progress on automation ROI: they addressed the data foundation first.
Clean, unified, continuously verified provider data isn’t a downstream benefit of automation. It’s a prerequisite. Every automation initiative built on fragmented provider data will eventually hit the same ceiling, the point where the data can’t support the decisions the technology is being asked to make. The ROI math on that investment never closes. The exceptions never fully clear. The manual queue never fully empties.
As Article 12 in this series addressed, a 0.1 to 0.3% lift in auto-adjudication rates typically offsets the full annual cost of a provider data management subscription. Most health plans are leaving multiple percentage points on the table, not because of claims system limitations, but because the provider data flowing into those systems hasn’t been continuously validated, reconciled, and trusted.
The Reframe That Changes the Investment Decision
The $21 billion automation opportunity isn’t waiting on better technology. It’s waiting on better data.
For CDOs, CIOs, and operations leaders being asked to deliver automation ROI, the question isn’t whether to invest in AI or workflow automation. It’s whether to build that automation on a foundation that will hold.
Provider data infrastructure is that foundation. When it’s right, when provider records are continuously validated against authoritative sources, when identities are resolved across credentialing systems, claims platforms, and directories, when every field carries a confidence score and a provenance trail, the automation that was already purchased starts to perform. Claims that should process automatically, process automatically. Exceptions shrink. Queues drain. The ROI that’s been promised for years becomes measurable.
The $21 billion is real. So is the reason it keeps slipping. The organizations that will close the gap are the ones that stop buying more automation and start fixing the data underneath it.
Ready to make your automation investment count? Schedule a 30-min Data Strategy Session with one of our data experts today. Not a sales pitch, but a working conversation about your provider data posture, verification cadence, and what’s standing between your current auto-adjudication rate and where it should be.
We also invite you to learn more through these key third-party sources:
- Experian Health 2025 State of Claims: AI Adoption at 14% Despite 67% Believing It Can Improve Claims
- Datagence: The Hidden Tax on Your Revenue Cycle