Kristi Stice
Director of Ambulatory Pharmacy Services & 340B Compliance · University Health
Built the operating model from the ground up.
Intelligent Pharmacy Centralization Playbook
Intelligent Pharmacy Centralization Playbook
How leading health systems centralize pharmacy to grow, deliver better care, and lay the foundation for AI.
We started Latent to automate prior authorization (PA). We found, almost immediately, that the leaders we work with were facing a much bigger problem. They were rebuilding specialty pharmacy for a less healthy population and a tougher market, and prior authorization was just where the pain showed up first.
The real challenge was shifting to a centralized operating model. Decentralization was creating medication access problems, delaying patient care, and causing both prescriptions and patients to leak out of the system.
As we partnered with more systems working through centralization, a pattern emerged. The systems that get it right redesign the work first. They put one team in charge, they standardize the process, and only then do they bring in AI.
Because if you automate a fragmented process, all you get is automated fragmentation.
The leaders in this playbook taught us these lessons. I hope what they've learned is as useful to you as it has been to us.
Director of Ambulatory Pharmacy Services & 340B Compliance · University Health
Built the operating model from the ground up.
Medication Access Supervisor · Cone Health
Centralized 15+ clinics in a single quarter.
Senior Director & ACPO – Ambulatory, Community, & Specialty Pharmacy Services · St. Luke's Health System
Redesigned medication access as an enterprise growth lever.
Director of Pharmacy Services · Akron Children's Hospital
Is charting a path that ties together data and stakeholders.
University Health built a centralization playbook from scratch.
At University Health, prior authorization was consuming roughly 400 nursing review hours and 200 pharmacy review hours every week. The labor cost was the symptom of a structural problem: no one owned the work. Kristi Stice's team solved it by rebuilding the operating model, step by step.
Scattered work. No owner. Volume leaks out of the system.
Quantify volume, review time, turnaround, denial patterns.
Move the highest-friction work into a pharmacy-led queue.
Consistent intake, evidence, escalation, appeals.
One workflow at a time, on the same model.
Layer AI inside the workflow. Only introduce once the model is stable.
A simple example proves the power of the approach. When an infusion center nurse who handled PAs full-time announced she was leaving, the team didn't backfill her. They borrowed an FTE from the 340B group to standardize the infusions workflow before she left, then moved a PA technician from the already-centralized dermatology queue into infusions. A staffing gap became a starting point for centralization. A clinic FTE became a central-team FTE.
From there, expansion followed a simple decision rule, applied service line by service line: where are denials high, and does our capacity allow us to take the volume on?
Gastroenterology came next. Community retail followed.
Each migration generated baseline data that the team used as leverage. First to anchor the broader centralization case with leadership, then to frame the technology decision itself: keep adding FTEs to scale the model, or add fewer FTEs plus AI. They chose fewer FTEs supported by AI and selected Latent to read the chart inside Oracle (formerly Cerner), surfacing the diagnoses, labs, and notes a submission needs, all while keeping a person in control.
Centralization also shifted denial data from a scoreboard to a prevention strategy. The team could finally see where denials were and weren't a problem and act upstream, building the required sequence into the submission from the start. In their words, the team transformed from data hunters to data validators.
Centralization was less about efficiency. It was more about giving the prior authorization work a clear owner so we could see all of it, standardize it, and improve it. Once we had that foundation, AI made the whole thing faster, but the order mattered. We built the model first, then used AI to scale it.
Cone Health moved prior authorization out of the margins of clinical work, and scaled the new model lean.
Before centralization, care coordinators and clinic staff at Cone Health worked PAs between patient visits, documentation, and the daily run of the clinic. Prior authorization was what they did when they had time. Patients could wait one to two weeks before anyone even started on their request.
Centralization gave the work a home. But the lesson from Cone Health isn't that it centralized. It's the design choices they made during the process.
The team is organized into disease-state pods. Each pod develops deep familiarity with a small set of conditions, answering the objection every chief pharmacy officer (CPO) hears: a central team can't match clinic-level expertise.
The team sees requests the moment they arrive and works alongside the clinic's own doctors and nurses in the record. A live operation, not a back-office queue. Latent surfaces the clinical evidence a reviewer needs no matter which clinic the request came from.
The centralized model means requests get reached sooner. Latent makes each review faster once reached. Measured separately, leadership always knows which lever to pull next.
The design scaled. Cone Health hit its quarterly centralization target in two weeks, and centralized 15-plus clinics in a single quarter with four fewer FTEs than projected.
At Cone Health, prior authorizations are not transactions. Behind every authorization is a patient waiting for their medication and a provider trusting a team of pharmacy technicians to clear the way forward.
St. Luke's stopped asking how to make PAs faster and started asking how to design better medication access. The ROI followed.
Josh Weber wasn't trying to fix prior authorization. He was rebuilding the entire pharmacy. So, instead of asking how to make PAs faster, he focused on a deeper problem: how should medication access be designed so it improves patient access, adherence, and enterprise growth?
St. Luke's centralized execution under a single Medication Access Team, redesigned the workflow into pods of three to five people, then deployed Latent across pharmacy and medical benefit.
Monthly prior authorizations cleared at the same FTE base. The added 1,200 is capture, not headcount: faster access keeps prescriptions inside the system's own pharmacy, refill after refill.
The same team now clears 1,200 more prior authorizations every month, but the return was never really about labor savings. It was about capture. Faster medication access means more prescriptions stay within the health system and more patients remain on therapy, refill after refill. ROI shows up as multi-fold returns inside the first year, with payback inside the first quarter. Weber didn't redesign medication access to spend less. He redesigned it to grow.
The future of pharmacy will not be defined solely by how efficiently we process prescriptions, but by how intelligently we orchestrate patient access, clinical outcomes, and enterprise growth.
more prior authorizations cleared every month. The same team.
5,500 → 6,700 monthly · same FTE base · capture, not headcountFour systems ran one playbook from four different starting points. Diagnose, concentrate, standardize, expand, then layer in intelligence: the order is crucial.
But centralizing operations is a foundation. Once the work lives in one queue, measured one way, even more high-value possibilities open.
Centralization makes it possible to see where prescriptions are leaking to external vendors, and to capture them. Across Latent's customers, that reaches tens of millions of dollars in specialty pharmacy revenue per health system. You cannot recapture what you cannot see.
Specialty pharmacy is becoming a source of clinical intelligence as much as a source of revenue. The CPO has a seat at the C-suite table because the enterprise needs pharmacy visibility across cost, margin, and continuity of care. Centralization is how that visibility gets built.
With the work in one queue, standardized and measured one way, intelligence finally has something stable to scale. The operating model comes first. The AI compounds it. That order is the whole playbook.
Kyle Finnerty, Director of Pharmacy Services at Akron Children's, is in the discovery phase of centralization.
At Akron today, authorization sits in two places. Pharmacy owns outpatient injectables. A separate financial group owns the rest. The financial group has the authorization expertise but lacks the clinical context to interpret the order and process it well. The goal is to align clinical, operational, and financial work under one owner. This is how they're getting started.
Akron began by quantifying the opportunity. The first job was proving to the rest of the system that centralization was worth the effort. Kyle's team dug into denial rates, site-of-care restrictions, and the inventory implications of bringing the work in-house.
Centralization at Akron crosses revenue cycle, finance, pharmacy, and the existing authorization group. Kyle's team is mapping who does what before proposing who should do what. The organizational work happens before the operational work.
How does a referral come in? How do teams communicate? What does it look like once approval comes through? Akron is documenting the current state in detail. The playbook the other three systems ran starts in the same place.
At Akron, centralization merged with a separate organizational mandate around agentic AI, becoming a single conversation about modernizing the operating model and turning back the leakage of patients to outside vendors.
Akron's specialty pharmacy tracks clinical outcomes against the 80% of patients who come in from outside vendors. The majority of those patients see better clinical outcomes after they move to Akron.
We had to go to leadership and show them denial rates, where patients were going, what we were losing in inventory. Until they saw it laid out, centralization was just an idea.
A playbook for pharmacy centralization at scale, built from the experience of four health systems: University Health (Kansas City), Cone Health, St. Luke's Health System, and Akron Children's Hospital. The core thesis: operating model first, intelligence second. The systems that get centralization right redesign the work first, with one team in charge and a standardized process, and only then bring in AI, because automating a fragmented process only produces automated fragmentation.
1. Diagnose the burden: quantify volume, review time, turnaround, and denial patterns. 2. Concentrate ownership: put one team in charge of the work. 3. Standardize the workflow. 4. Expand by service line, using a simple decision rule: where are denials high, and does capacity allow taking the volume on? 5. Layer in AI once the operating model is stable.
University Health, Kansas City (built the operating model from the ground up); Cone Health (centralized 15+ clinics in a single quarter using disease-state pods); St. Luke's Health System (redesigned medication access as an enterprise growth lever); and Akron Children's Hospital (in the discovery phase, aligning clinical, operational, and financial ownership).
University Health: 2× weekly prior authorization submissions (144 to 312), 44% reduction in overall PA review time (14.7 to 8.2 minutes), and a 50% reduction in clinic-administered medication denials. Cone Health: +60% monthly PA volume, 15+ clinics centralized in a single quarter with four fewer FTEs than projected, and a 78% reduction in turnaround time. St. Luke's Health System: 57% reduction in PA turnaround time (over 7 days to about 3), 72% reduction in average review time (about 18 minutes to about 5), and 1,200 more prior authorizations cleared per month on the same FTE base (5,500 to 6,700).
Because if you automate a fragmented process, all you get is automated fragmentation. With the work in one queue, standardized and measured one way, intelligence finally has something stable to scale. The operating model comes first; the AI compounds it. That order is the whole playbook.
Pharmacy spent decades being told to shrink. The leaders running it now have been asked to grow. Centralization is how the operating model catches up to the mandate.
The destination was never faster prior authorizations. It was a pharmacy that could operate as a business, a clinical partner, and a growth engine at the same time. The leaders in this playbook didn't centralize to save minutes. They centralized to build organizations capable of delivering better care at a larger scale.