When a patient calls PFS about a confusing statement, an unexplained denial, or a payment plan that makes no sense, the problem usually did not start in PFS. It started upstream in registration, billing, workflow design, or communication. Amanda Hines, Patient Financial Services Director at Essentia Health, makes the case that revenue cycle leaders should pay closer attention to that downstream signal. In her view, PFS sees the ecosystem more clearly than most departments because it is where upstream defects finally become visible to patients.
One of the best sources of denial-prevention insight in your health system may already be on payroll in PFS. The question is whether the rest of the revenue cycle is set up to hear what those representatives are hearing every day.
Hines started as a collector making outbound calls and offering payment plans. She grew into leadership and started noticing the patterns her collectors were absorbing.
“The longer you’re in it, the more you start to realize that everything really is connected,” said Hines. “I started thinking about revenue cycle a lot less like individual silos or work lanes and more like an actual like ecosystem.”
A registration issue becomes a billing issue later. A poorly worded statement or an unexplained denial becomes a patient experience issue that shows up on the PFS phone line. Hines jokes that her team is "the back end of the back end." The people fielding those calls know - in real time - where the system is breaking.
The operator question is whether that knowledge ever gets structured into a form the rest of the revenue cycle can act on. Without a deliberate mechanism, PFS handles the account, closes the ticket, and moves on. And the upstream owner never learns what the downstream absorber saw. In practice, PFS functions like an early-warning system for the ecosystem, but only if leaders build a channel to receive the signal.
Hines applies a specific screen to AI and automation decisions. She frames it as a question of appropriateness, not capability.
“Can technology really do the work,” asked Hines. “I guess maybe the question is, can technology do the work or should it do the work? Just because AI can do something doesn’t automatically mean we should let [it] do it and without oversight.”
Just because AI can perform a task does not mean it should perform that task without oversight. Her risk threshold is explicit: the closer the decision gets to a patient's wallet, the more governance it requires. Anything that affects a patient's livelihood, ie. how much a patient is supposed to pay, what a payment plan amount should be, is where she draws the line for human judgment. As she puts it: “you can't outsource accountability to a robot.”
Based on Hines's framing, here's what a risk-tiered governance approach could look like in practice. Low-risk repetitive tasks - routing, data entry, status inquiries, informational retrieval - can move to automation with lighter oversight. Anything touching a patient's balance, payment terms, or dispute needs auditing, human review, and defined escalation paths.
Hines is direct about what technology cannot do in patient interactions.
“It can’t use empathy, it doesn’t really have judgement,” said Hines. “It doesn’t have context.”
A chatbot cannot hear confusion in a patient's voice, cannot sense frustration building, cannot advocate for the patient based on the whole story. It can retrieve information. It cannot practice judgment.
The split RCM leaders should consider is between information retrieval and judgment-and-advocacy work. Blur that line, as Hines warns, and "your customer service line is going to be on fire."
If automation absorbs the transactional layer, the work that stays with humans gets harder.
“The work left for people is going to be more complex,” Hines said.
That single sentence is a workforce strategy problem most RCM leaders have not fully priced in. Exception handling, escalation calls, nuanced payment conversations, QA of virtual agent interactions, these require staff with more training, more tenure, and more judgment than the entry-level work automation is absorbing.
Hines has been building toward this. Her structured training program at Essentia runs three to four weeks and includes classroom instruction, department-specific learning, case study examples, call listening, and safe test-environment practice. It replaced the older peer-shadowing model where a new hire sat next to someone experienced, hoped the right scenarios came up, and was wished good luck.
“People are being asked to perform in a really complex environment before they’re probably truly prepared,” said Hines.
Her attrition observation is worth taking seriously. She says poorly trained hires are more likely to leave before going independent or within the first 60 days because the overwhelm and stress become too much. That's an early-attrition signal Hines codes as a training-quality problem, a signal most departments never route back to their training function at all.
The capability the reader needs is a training function that treats onboarding as a workforce stability metric. That can grow senior staff into the QA, exception-handling, and virtual-agent-supervision roles that automation is creating. That capability needs to exist before the automation lands, not after.
The other reason PFS insight matters now is that regulation lands across the ecosystem, not in one department.
Hines operates under federal rules, Minnesota state rules, Minnesota Attorney General requirements, and 501(r). She describes her mindset moving through the 2023 and 2024 regulation waves this way:
“When that happens, it’s not like a little project,” said Hines." "That’s like an actual change management effort.
Her implementation approach is worth cataloging, because a policy update alone is not enough when a regulation touches every part of the revenue cycle. Hines's sequence:
That is enterprise change management, not a memo. And PFS is where you find out fastest whether what you rolled out is actually working when patients encounter it. Hines makes this point plainly: if organizations underestimate the training and communication that has to go with a regulation change, that's where they get themselves in a pickle.
The thesis is straightforward. Patient financial services sees the ecosystem earlier and more clearly than any other function in the revenue cycle. Treating it as a collections function wastes the diagnostic signal. Automating on top of it without listening to it scales the defects it has been absorbing.
Based on Hines's comments, here's what that could look like in practice for a VP of Revenue Cycle or PFS Director walking into a Monday operations meeting.
First, build a monthly PFS root-cause report. Track the top five drivers of inbound patient calls - confusing statement language, denial explanations, registration-driven balance errors, payment plan disputes, insurance mismatches - with fields for issue type, upstream owner, patient impact, volume trend, and corrective action. Route it to registration, authorization, statement design, and denial communication owners on a 30-day cadence.
Second, apply the can-vs.-should test to every automation and AI decision on the roadmap. Anything touching a patient's balance, payment plan, or affordability gets heavier governance - auditing, human review, defined escalation, QA of the interactions. Anything in the information-retrieval category can move faster with lighter oversight.
Third, validate the workflow before you automate it. If PFS has been telling you the process is broken, automating the broken process scales the break. This is a sequencing discipline, not a technology decision.
Fourth, treat early attrition - the 60-day window Hines flags - as a training quality metric. If new hires are leaving before they go independent, examine the training function first.
Fifth, redesign the career path now. The work left for people is going to be more complex. Senior representatives should be moving toward QA, exception handling, and virtual agent supervision. If you wait until automation lands to plan that transition, you will lose your senior bench during the changeover.
Sixth, treat the next regulation as a change management program from day one. Legal, compliance, IT, and operations at the table together. Interpretation, workflow mapping, communication, training, post-go-live metrics. And consider forming a peer working group with providers in your market before you finalize interpretation.
These capabilities can be built internally, acquired through partnership, or run as a hybrid. The organizational form is a downstream decision. The upstream decision is whether you are willing to listen to PFS before you scale on top of it. Hines's team hears the signal every day. The question is whether the rest of the revenue cycle is set up to receive it.