One of the biggest mistakes organizations can make with AI and automation is automating a bad process.
Automation can make work faster. AI can add judgment, pattern recognition, and interpretation. But neither one automatically makes the underlying process better.
If the workflow is unnecessary, inconsistent, poorly designed, or filled with avoidable handoffs, technology may simply allow the organization to do the wrong work more efficiently.
Start With the Work, Not the Tool
Before deciding where to apply automation or AI, I like to work through a simpler sequence:
- Eliminate work that does not need to happen.
- Simplify the workflow and remove avoidable handoffs.
- Standardize the parts of the process that should operate consistently.
- Automate repetitive, rules-based work where the process is understood.
- Apply AI where judgment, interpretation, prediction, or pattern recognition adds value.
The sequence matters. Technology should be applied to a process that has already been challenged—not used as a substitute for challenging it.
Revenue Cycle Has Plenty of Automation Opportunities
RCM contains large volumes of repetitive work: eligibility checks, claim edits, work-queue routing, coding support, denial classification, follow-up prioritization, documentation review, and reconciliation.
Those are attractive targets for technology. But the first question should still be: why does this work exist in its current form?
If a team is manually correcting the same eligibility defect every day, automating the correction may help. But eliminating the upstream defect may create far more value. If claims are being routed through multiple queues because of inconsistent rules, automating the routing does not necessarily solve the process design problem.
Automation and AI Are Not the Same Thing
It is also useful to distinguish traditional automation from AI.
Automation is often strongest when the rule is clear: when X happens, perform Y. AI becomes more useful when the task involves interpretation, prioritization, classification, or decisions where patterns are more complex than a fixed rule set.
Both can be valuable. The mistake is starting with “Where can we use AI?” rather than “What problem are we trying to solve, and what is the simplest effective way to solve it?”
Measure the Work You Remove
A successful technology initiative should not only improve throughput. It should reduce unnecessary touches, shorten cycle time, improve quality, create capacity, or prevent defects from entering the system in the first place.
That means measuring more than whether the tool technically works. Leaders should ask:
- How many manual touches disappeared?
- Did service-to-bill or claim-processing time improve?
- Did error or denial rates decline?
- Was staff capacity redirected to higher-value work?
- Did the technology eliminate a problem or simply move it somewhere else?
The Goal Is Less Unnecessary Work
AI should not make inefficient work happen faster.
The better objective is to build a revenue cycle that requires less unnecessary work in the first place—and then use automation and AI to make the remaining work more accurate, scalable, and effective.
Revenue Cycle Insight Lab focuses on practical ways to improve RCM operations, performance, technology, and decision-making.
