Health plans have spent the past decade automating. Workflow systems, robotic process automation, offshore operations, application programming interfaces, cloud data platforms, and now artificial intelligence have all found their way into payer operations. Despite this investment, large parts of the day-to-day work inside health plan data operations still run on people. Analysts review spreadsheets. Teams reconcile conflicting provider records. Staff carry out manual healthcare data processing by hand, moving data between systems, chasing down exceptions, calling providers, reading documents line by line, and cleaning up errors that surface further down the pipeline.
The natural conclusion is that health plans simply have not invested enough in healthcare data automation. Look closer and a different picture emerges. Automation has been layered on top of provider, claims, and clinical data that is fragmented, inconsistent, and constantly changing. Rules-based tools can move that data faster. They cannot make it trustworthy on their own.
Health plans are overspending because so many of their health plan data operations still depend on people to compensate for what their underlying data and systems cannot resolve. Understanding why that gap persists, and where it costs the most, is the first step toward closing it.
The Hidden Economics of Manual Data Operations
Direct labor is only the visible cost
The most visible line item behind rising healthcare administrative costs is headcount: internal operations teams, outsourced or offshore resources, data analysts, provider operations staff, clinical reviewers, quality teams, and customer service staff who spend their day interpreting and correcting data rather than acting on it. This is the cost that shows up in a budget review. It is rarely the largest one.
The cost multiplies downstream
A single data problem rarely stays contained to one department. Incomplete or inaccurate data triggers a manual investigation, which delays processing, creates rework, produces downstream errors, and triggers another manual intervention somewhere else in the organization. Along the way, health plans absorb the cost of repeated validation, duplicate work, escalations, correction cycles, delayed decisions, provider abrasion, member friction, compliance exposure, and sometimes lost revenue or unnecessary spend on services that should have been caught earlier.
A ten dollar manual task rarely stays a ten dollar problem. When the same underlying data issue surfaces in credentialing, claims, network management, and member services, the true cost compounds across every team that has to touch it, driving healthcare administrative costs higher across the board. The CAQH Index, the industry's longest-running benchmark for administrative transaction costs, found that provider time spent on administrative transactions rose 14 percent on average that year, adding roughly $21 billion in cost to the medical industry. The nine transaction types the index tracks already account for $89 billion of the roughly $400 billion the country spends annually on administrative complexity.
Why Manual Operations Persist Despite Years of Automation
Health plan automation has advanced steadily over the past decade, yet three structural issues keep pulling teams back into manual work.
1. Healthcare data doesn't arrive clean and standardized
Health plans ingest information from providers, health systems, clearinghouses, claims, electronic health records, rosters, directories, clinical records, government databases, and third-party vendors. Each source has its own format, update frequency, level of completeness, and definition of the same field, which makes consistent healthcare data management difficult from the outset. A rules engine built to expect one structure struggles the moment a second, differently formatted source enters the pipeline.
2. The same entity can look different across systems
A single provider can appear under slightly different names, practice locations, specialties, affiliations, and identifiers depending on which system recorded them. Contracts describe the same relationship in different language. Clinical concepts get coded inconsistently across settings. Humans end up serving as the reconciliation layer, manually deciding whether two records describe the same person, place, or agreement.
This is not a minor inconvenience. A JAMA study that compared publicly available directories across five large national insurers found accuracy of around 59 percent for providers practicing at a single location, and only 19.4 percent for providers with multiple practice locations. The same provider, recorded slightly differently across systems, becomes progressively harder to reconcile the more places they appear.
3. Traditional automation works best when the rules are predictable
Robotic process automation and deterministic rules engines excel at repeatable, well-defined steps, the backbone of most payer workflow automation to date. They struggle with unstructured documents, ambiguous information, missing fields, conflicting sources, exceptions, and data that keeps changing shape. Moving a record through a workflow faster does not resolve the underlying question of what that record actually means. Automating the movement of data is a different exercise from automating the judgment required to interpret it, and most legacy automation was built to do the former
The Five Places Health Plans are Still Paying the "Manual Data Tax"
Five workflows illustrate where health plan data operations absorb the most manual effort today.
1. Provider data management
Roster ingestion, demographic validation, directory updates, affiliation reconciliation, specialty verification, provider outreach, and one-off data corrections consume a disproportionate share of operations time. This is where payer data management shows its biggest gaps. The consequences reach well beyond the data team, feeding directly into directory accuracy, claims adjudication, credentialing, contracting, and network management, and ultimately determining whether a member can find and reach a provider who is available.
The scale of the problem is well documented. A federal review of Medicare Advantage online directories found that 48.74 percent of provider directory locations contained at least one inaccuracy, ranging from wrong phone numbers to providers who were not actually located at the listed address. Stronger payer data management would catch most of these errors before they ever reach a member. For members seeking a specific type of care, particularly behavioral health, the impact can be severe. A Senate Finance Committee secret shopper study of Medicare Advantage mental health listings found that staff could schedule an appointment only 18 percent of the time, with more than 80 percent of listed providers turning out to be unreachable, not accepting new patients, or not actually in network. A report from the New York Attorney General's office found a similar pattern, with 86 percent of listed in-network mental health providers turning out to be ghosts.
2. Credentialing
Source verification, document collection, application review, data entry, discrepancy resolution, and recredentialing remain some of the most stubborn examples of manual healthcare data processing at most health plans and provider organizations, even though much of the underlying information already exists somewhere in the organization's own systems. Every discrepancy that a reviewer has to chase down by phone or fax adds time to a process that determines when a provider can start seeing members and generating revenue.
3. Contract data management
Rates, terms, and obligations that determine how a health plan pays and manages its network are frequently trapped inside PDFs and other unstructured documents. Abstracting contract terms, comparing amendments, tracking obligations, and validating that operational systems reflect what a contract actually says remains a manual, document-by-document exercise for many plans.
4. Clinical data operations
Chart abstraction, quality measure reporting, risk adjustment, utilization management, and clinical gap identification all depend on interpreting clinical information that is rarely structured in a way systems can act on directly. Reviewers spend hours reading notes to find the handful of data points that matter for a given measure or determination, work that is essential to healthcare data quality but does not require a clinician's judgment on every single record.
5. Provider outreach
When systems cannot confidently resolve a question about a provider, a plan, or a claim, the fallback is almost always a phone call, an email, or a fax. Teams still rely on manual outreach to resolve information that automated systems flag but cannot settle on their own.
Across these workflows, the underlying pattern in health plan data operations repeats. People spend their time finding, interpreting, validating, reconciling, and correcting data because the systems around them cannot yet do that work reliably.
Why Throwing More People at the Problem Stops Working
The traditional response to rising transaction volume has been to add capacity: more people, more outsourced or offshore resources, higher operating cost. This approach has real limits. Costs scale roughly with workload. Institutional knowledge sits with individual employees rather than the organization. Quality varies by reviewer. Training becomes a constant, ongoing expense. Turnover introduces operational risk every time an experienced employee moves out. Exceptions pile up faster than teams can clear them, and service-level pressure often encourages tactical fixes that resolve a single case without addressing why the case existed in the first place, the exact gap that healthcare administrative automation is meant to close.
This is linear scaling in practice. When transaction or data volume rises 30 percent, an operation that depends on human intervention typically needs a proportional increase in capacity just to keep pace, and that additional capacity shows up directly in the budget. Effective health plan automation offers a different curve. Done well, it decouples operational capacity from headcount, so growth in membership, provider networks, and data volume no longer requires a matching increase in staff.
The Automation Paradox: Why RPA and Workflow Tools didn't Eliminate the Work
Traditional automation was never designed to solve this problem, and it would be a mistake to call it a failure. Robotic process automation and workflow tools were built to handle the doing: moving files, populating fields, routing work, triggering the next step, and executing clearly defined rules. They have done that well.
What health plans increasingly need is automation that can handle the understanding. Are these two records describing the same provider. Which source should be trusted when two records disagree. Is this information still current. What does a particular contract clause actually require. Does a clinical note contain evidence relevant to a specific quality measure. Is a flagged exception really an error, or a false positive.
These are judgment calls, and deterministic rules struggle to make them consistently at scale. The next wave of healthcare data automation centers less on automating individual tasks and more on automating the data interpretation and decision making that has to happen before a task can even be completed. Health plans that only automate the movement of work, without upgrading the payer workflow automation underneath it, will keep running into the same wall: fast processes built on top of data nobody has actually resolved.
From Workflow Automation to Data Intelligence
The natural progression runs from manual operations, to rules-based automation, to genuinely intelligent healthcare data management. This is what healthcare data intelligence looks like in practice: modern AI can increasingly help health plans ingest structured and unstructured information, normalize data, identify entities, resolve conflicting records, detect anomalies, infer missing relationships, extract information from documents, prioritize exceptions, recommend actions, and learn from outcomes over time.
Humans Should Manage Exceptions, Not Every Transaction
This should be the operating principle behind any modernization effort. Rather than having reviewers examine every record that comes through the door, intelligent systems can resolve high-confidence cases automatically and route only the ambiguous or high-risk exceptions for human review. This is the clearest practical expression of AI in health plan operations: a pipeline where AI processes, validates, resolves, and scores confidence, and people step in only where the system's confidence is genuinely low. Shifting effort this way moves skilled staff away from repetitive data handling and toward the judgment calls that need them.
What Health Plans Should Measure Instead of Headcount
Executives evaluating these workflows often default to a single question: how many full-time employees does this process require? A more useful gauge of payer operational efficiency looks at the data itself, through questions such as:
- What percentage of transactions require human intervention
- What is the cost per successfully resolved record
- How many records require rework
- What percentage of exceptions are automatically resolved
- How long does data take to become operationally usable
- How often does the same data issue trigger work in multiple departments
- What percentage of operations can run straight through
- What is the accuracy rate after automation
Human touch rate, the share of transactions that still require a person, is an especially useful metric for tracking progress over time. The goal is not zero human involvement. It is reducing the percentage of transactions that require manual intervention while holding or improving healthcare data quality.
The Business Case Goes Beyond Administrative Savings
Cutting full-time headcount is the most obvious argument for health plan automation, and the least complete one. Improving the data underneath payer workflows also affects payer operational efficiency through less manual processing and rework, provider experience through fewer repetitive requests and corrections, member experience through more reliable provider information and smoother access to care, revenue performance through a better ability to identify and act on information hidden in clinical and operational data, compliance and quality through more consistent and traceable processes, and scalability, since membership, provider networks, and data volume can grow without a matching increase in operational headcount.
The scale of what is at stake is significant. Healthcare administrative costs are estimated at 15 to 30 percent of total US health spending, and a Health Affairs review concluded that roughly half of that, or 7.5 to 15 percent of national health spending, is effectively wasteful. A separate JAMA analysis identified administrative complexity as the single largest of six waste categories in the US healthcare system, at roughly $265 billion a year. Data operations sit at the center of that complexity.
The Question Health Plans Should be Asking
For years, the operating question inside most health plan data operations teams has been where to automate another workflow. A more useful question is why a given workflow requires a person in the first place.
The answer usually traces back to the data underneath it. It is not trusted, not standardized, not connected, not structured, not current, or the system simply cannot interpret it.
The larger opportunity is eliminating the repetitive data work that keeps skilled teams from focusing on the decisions, relationships, and outcomes that genuinely need human judgment. Automation done well frees people for that work rather than replacing them. Health plans that invest in healthcare data intelligence and broader healthcare administrative automation across their health plan data operations can change the economics of their operations rather than making today's processes marginally faster.
