The Phone Call Nobody Should be Making

How an agentic AI agent is giving case managers their time back, one discharge at a time

Every hospital case manager knows this moment too well.

It's 2:15 PM. You have six patients who need to be discharged by Friday. Two families haven't responded to your messages. A prior authorization is stuck in purgatory somewhere inside a payer portal. And you are on hold — again — with a skilled nursing facility (SNF) that may or may not have a bed available for a 71-year-old woman with congestive heart failure, Medicare Advantage, and a daughter who speaks only Tagalog.

The hold music plays.

You already called three facilities this morning. Two didn't pick up. One picked up, but the admissions coordinator wasn't in yet. You have a master's degree in social work or nursing. You are credentialed, experienced, and deeply skilled at navigating the most complex human transitions in medicine — the moment a patient moves from the acute hospital to the next phase of their recovery. And right now, you are on hold.

This is the quiet crisis inside American hospitals. Not the one that makes the front page, but the one that costs health systems an estimated $2,500 to $3,300 per avoidable hospital day, extends length of stay, strains clinical staff, and, most importantly, delays care for patients at their most vulnerable moment.

World Wide Technology (WWT) built something to fix it.

The discharge maze

To understand why this matters, you have to understand what case managers actually do.

Discharge planning is not a single task. It is a cascade of interdependent decisions and coordination touchpoints that can span five to ten days for a complex patient. A case manager must simultaneously assess a patient's clinical status, functional needs and social circumstances; navigate payer rules and prior authorization requirements; educate patients and families — often across language and literacy barriers; coordinate with physicians, therapists, social workers and pharmacists; and find a post-acute placement that matches the patient's payer, clinical needs, geographic preferences and — increasingly — cultural and language requirements.

The American Case Management Association's biennial National Case Management and Transitions of Care Survey, which benchmarks more than 400 hospital case management departments nationwide, documents that case managers routinely spend a significant share of their time on administrative and coordination tasks rather than direct patient care. Of those administrative tasks, post-acute facility outreach — calling SNFs, long-term acute care hospitals (LTACHs), and home health agencies to check availability, verify capabilities and confirm payer acceptance — is among the most time-consuming and least value-added.

Think about what that phone call actually involves. The case manager filters facilities by payer network. She eliminates those too far from the patient's family. She factors in whether the facility has Spanish-speaking staff for a patient whose wife doesn't speak English. Then she calls. She may get a voicemail. She calls back. When she finally reaches someone, she asks the same questions she has asked a hundred times: Do you have a bed available? Does your facility accept Humana Gold Plus? Do you have the capability to manage a patient with an NG tube and a new Afib diagnosis requiring anticoagulation initiation? Do you have Spanish-speaking nursing staff on the unit?

— California Hospital Association, 2024

A 2024 report by the California Hospital Association estimated 1 million days of unnecessary inpatient care annually, totaling $3.25 billion in avoidable costs — with SNF placement difficulty and payer delays identified as the most consistent driver. Multiply a single outreach call across five facilities per patient, then across a typical case manager caseload of 15 to 25 patients, and the arithmetic is sobering. This is hours of work per day — work that requires no clinical judgment, produces no clinical insight, and could be done by something else entirely.

Now it is.

What WWT built

At WWT's AI Proving Ground — our research and innovation environment where we design, test and validate AI solutions in conditions that mirror real-world healthcare deployments — our team built an agentic AI agent purpose-built for this exact problem.

The agent doesn't just search a database. It makes calls.

Here is how the solution works in practice.

When a case manager initiates a discharge placement workflow for a patient, the agent begins by pulling the patient's relevant parameters: payer and plan type, geographic radius, clinical care requirements (such as NG tube management, IV antibiotic capability, ventilator weaning, or memory care), and demographic preferences including language spoken and cultural needs. It cross-references these against a curated, dynamically maintained network of post-acute facilities — SNFs, LTACHs, inpatient rehabilitation facilities, and home health agencies — pre-filtered to surface only those in-network for the patient's specific plan.

Then it calls them.

Using a voice AI agent with natural language capability, the system contacts each facility's admissions line and conducts a structured intake conversation. It confirms current bed availability. It verifies that the facility can support the patient's specific clinical requirements. It asks about staffing capabilities — including language and cultural competency where flagged — and logs the response in a structured, timestamped format back into the case manager's workflow dashboard.

If a facility is unavailable, the agent moves to the next on the list — automatically, without any hold music, without any callback tag, and without any case manager time. When it has a match, it surfaces the result with a confidence-weighted summary: which facilities have confirmed availability, which have the required clinical capabilities, and which are ready to accept a referral.

The case manager reviews the results. She makes the relationship call to her top choice. She confirms clinical acceptance with her clinical judgment intact. Then she moves on to the patient who actually needs her — the one whose family is in crisis, whose insurance situation is complicated, whose discharge plan requires a human being with expertise and empathy to navigate.

Why this moment is different

Artificial intelligence has been circling healthcare for years. We've had predictive models, natural language processing tools, clinical decision support algorithms, and documentation assistants. Some of it has been genuinely transformative. Much of it has been narrowly deployed or poorly integrated.

What makes agentic AI different, and what makes this use case a signal worth paying attention to, is that the agent doesn't just surface information. It acts.

The distinction matters. A traditional AI tool might tell a case manager: "Here are three SNFs within five miles that accept Humana Gold Plus." The agentic system makes calls, asks the right questions and returns with: "Kindred at Downey has two beds available, confirmed Spanish-speaking nursing staff, confirmed NG tube management capability, and is ready to receive a referral packet." That is not a search result. That is a completed task. 

It's the difference between a GPS that shows you the map and one that drives the car.

Healthcare is full of high-stakes, time-sensitive workflows that follow a structured logic — workflows where the rules are knowable, the data is accessible, and the right questions are consistent. Discharge placement is one of the clearest examples. The criteria for a good SNF match don't change from patient to patient in their structure. What changes is the content — the specific payer, the specific clinical needs, the specific language, the specific geography. An agentic agent can hold all that context and execute at a scale and speed no human scheduler can match.

The clinical stakes

This isn't an efficiency story in isolation. It is a patient safety and health equity story.

Every day a patient remains in an acute hospital bed after they are clinically ready for discharge carries real risk. Research published in the International Journal of Pharmaceutical and Healthcare Marketing found that extending a patient's hospital stay by a single day increases the probability of acquiring a healthcare-associated infection by 1.37%. A systematic review published in the American Journal of Infection Control found that hospital-acquired conditions — including falls, pressure ulcers and central line infections — added an average of 5.2 to 22.1 additional days to the length of stay for affected patients. For an elderly patient recovering from a stroke or a hip fracture, an extra two days in the hospital is not a neutral event.

A 2020 multicenter study of 1,543 traumatic brain injury patients found that 14% experienced a discharge delay — and that the two most common causes were insurance authorization (52%) and lack of an accepting bed (41%). Patients destined for SNF placement had more than 10 times the odds of experiencing a discharge delay compared to those going home. Those are precisely the barriers this agent is designed to eliminate.

And for Spanish-speaking patients — or Tagalog-speaking patients, or Somali-speaking patients, or any patient whose language and cultural background narrow the viable facility pool — the discharge gap is even wider. Case managers doing manual outreach can only call so many facilities before time and cognitive bandwidth run out. The agent doesn't run out. It works through the full network, surfaces bilingual options, and ensures language access isn't inadvertently sacrificed for expediency.

The equity implications of that are not small.

The WWT difference

WWT is not a software company with a healthcare vertical. We are a global technology solutions provider with more than 35 years of experience helping the world's largest organizations design, build and operate complex technology at scale. In healthcare, that means we sit at the intersection of the clinical, operational and technical in a way few organizations can.

Our AI Proving Ground makes that intersection concrete. It is a validated, vendor-neutral environment where healthcare organizations can see AI solutions working against real use cases — not demos, not slide decks — before making enterprise commitments. For discharge planning automation, that means a health system can watch the agent make calls, review the structured outputs, audit the accuracy of what it captures, and measure the time savings against their own case managers' workflows.

We also understand that no agentic AI solution exists in isolation. The value is in the integration — into the EHR workflow, the case management platform, the payer authorization ecosystem, and the broader post-acute network. WWT's depth across Epic, Oracle Health, and major healthcare IT platforms means we can connect this capability to the systems your case managers are already living in, not ask them to learn something new.

And we understand the governance side. Healthcare AI that acts — that makes calls, submits data, initiates workflows — operates in a regulated environment with HIPAA implications, patient consent considerations, and clinical oversight requirements. WWT's AI governance frameworks are built for that reality. Every agent action is logged. Every human override is preserved. Every escalation point is defined. The case manager stays in control; the agent does the work she shouldn't have to do.

What case managers are saying

When we walk case managers through this capability, the reaction is consistent. The first response is often skepticism — can it really understand the questions well enough to get accurate answers? The second, after they see it in action, is something closer to relief.

That is the right instinct. The case manager's irreplaceable value is not in asking whether a facility has a bed. It is in the family meeting conducted through a telephone interpreter with a frightened wife who thinks her husband is being sent to a nursing home to die. It is in recognizing that a patient's stated preference for a facility close to home is actually a signal about social isolation and caregiver burden. It is in knowing which SNF director will actually pick up the phone when there's a clinical problem at 10 PM and which one won't.

Those things cannot be automated. They shouldn't be. The agent exists to protect the space in which they happen.

The road ahead

Discharge placement is the beachhead, not the ceiling.

The same agentic architecture that handles SNF intake calls can extend across the discharge workflow: prior authorization submission and tracking, clinical referral packet assembly, DME order coordination, transportation booking, post-discharge follow-up outreach, and readmission risk monitoring. Each is a structured, rules-based, data-intensive task that consumes case manager hours and creates friction in the patient journey.

Across a health system discharging 200 patients per day, the cumulative time recapture from even a subset of these automations runs into thousands of hours per month. The downstream effects — on length of stay, on avoidable readmissions, on staff retention in a workforce where 43% of physicians report burnout symptoms and case managers face comparable pressures — are measurable and significant.

The technology to do this exists today. The integration expertise to deploy it responsibly in healthcare exists today. WWT has both.

The question is no longer whether AI can help case managers do their jobs better. The question is how quickly health systems are willing to stop paying clinical professionals to wait on hold.

Or connect with WWT's Healthcare Practice to schedule a live demonstration at our AI Proving Ground — where we prove solutions before you commit to them.

About the authors

Eric Quiñones, MD, is Chief Healthcare Advisor at World Wide Technology, where he works at the intersection of clinical strategy and enterprise technology to help health systems navigate the AI transformation of care delivery. He can be reached through wwt.com.

Ina Johnson, is a Data Science Technical Solution Architect at World Wide Technology, where she helps customers pinpoint where their data and AI capabilities fall short, then build the path from those gaps to real ROI. She can be reached through wwt.com.

Disclosure

This article was authored by a World Wide Technology employee and reflects WWT's view of the healthcare AI landscape based on publicly available research and WWT's AI Proving Ground work. All referenced studies are independent third-party research. Statistical claims are drawn from cited sources and have not been independently validated by WWT.