Top 10 AI Prompts and Use Cases and in the Healthcare Industry in Charlotte

Charlotte and the broader North Carolina health ecosystem have become early adopters of practical AI - deploying tools that predict risk, cut clinician paperwork, and speed urgent care decisions.

Local examples include lung‑nodule scoring and portal‑message drafting at Atrium Health, AI follow‑up assistants at OrthoCarolina, radiology triage at Novant, and Duke's Sepsis Watch, which the state reports cut sepsis mortality by 31% - while Atrium says its DAX Copilot can save clinicians roughly 40 minutes a day; for a full catalog see North Carolina Health News: 10 ways NC providers are harnessing AI (2025) and read more on Atrium Health DAX Copilot improves clinician documentation.

Those measurable gains - faster triage, fewer messages, documented minutes reclaimed - explain why hospitals, regulators and Charlotte's workforce are all focused on safe, governed AI adoption.

Bootcamp	Length	Early‑bird Cost	Registration
AI Essentials for Work	15 Weeks	$3,582	Register for the AI Essentials for Work bootcamp

“This is just one example of an innovative way to use this technology so that teammates can spend more time with patients and less time in front of a computer.” - David McSwain, UNC Health

Selection prioritized AI use cases with clear North Carolina provenance, measurable clinical or operational impact, and evidence of local governance or safety review: candidates had to be implemented by an NC system (Atrium, Duke, Novant, Wake Forest, UNC, OrthoCarolina) or described in state reporting, show demonstrable outcomes in the field (for example, Duke's Sepsis Watch was reported to reduce sepsis mortality by 31% and WakeMed's drafts cut provider inbox volume by roughly 12–15 messages per day), and surface policy or equity concerns that require oversight.

Sources and public reporting guided weighting: deployment scale and documented results from “10 ways NC health care providers are harnessing AI” and the state's calls for organized oversight in “As AI advances in NC health care, state leaders call for oversight” were used to verify claims and to highlight where human oversight and governance frameworks mattered most.

Selection Criterion	NC Example (source)
Measured clinical impact	Duke Sepsis Watch - 31% mortality reduction (NC Health News article on North Carolina health care harnessing AI (2025))
Operational efficiency	WakeMed/Atrium message drafting - ~12–15 fewer messages/provider/day (NC Health News article on North Carolina health care harnessing AI (2025))
Governance & equity risk	State leadership urging oversight; system vetting highlighted (NC Health News report on state oversight for AI in NC health care (2025))

“If it makes the wrong decision, where's the liability? Who's responsible?” - Sen. Jim Burgin

Atrium Health Wake Forest Baptist has adopted the FDA‑approved Optellum Virtual Nodule Clinic to strengthen early lung‑cancer detection and triage: the AI was trained on more than 70,000 CT scans and produces a Lung Cancer Prediction score that classifies incidental nodules as high, intermediate or low risk, helping clinicians prioritize timely biopsies while reducing unnecessary procedures for low‑risk patients.

Wake Forest Baptist - one of the first U.S. academic centers to deploy this tool - combines the AI scores with multidisciplinary review and robotic bronchoscopy to reach small, hard‑to‑access nodules, and the program evaluates hundreds of nodule patients each year.

The practical result for North Carolina care is clearer pathways from an incidental CT finding to the right next step - faster, more confident referrals for those at highest risk, and fewer invasive follow‑ups for the majority who are benign (see the Wake Forest news release and Optellum's Virtual Nodule Clinic overview for details).

Feature	Detail
AI training data	More than 70,000 CT scans
Risk categories	High / Intermediate / Low
Annual local volume	Wake Forest Baptist assesses >500 pulmonary nodule patients
AI developer	Optellum (Virtual Nodule Clinic)
Adjunct technology	Robotic bronchoscopy (Intuitive)
Program recognition	Screening Center of Excellence; NCI‑designated Comprehensive Cancer Center

“The exciting part of this artificial intelligence lung cancer prediction tool is that it enhances our decision making, helping doctors intervene sooner and treat more lung cancers at an earlier stage.” - Dr. Christina Bellinger

OrthoCarolina - headquartered in Charlotte, N.C. - partnered with healthPrecision to deploy Medical Brain® across a network of more than 300 providers at nearly 40 locations, using a patient‑facing mobile app that delivers 24/7 personalized clinical guidance and an “intelligent extended arm” for clinicians; this continuous, automated postoperative engagement and remote monitoring flags emerging risks and care gaps so teams can intervene earlier, reduce unnecessary in‑person visits, and lower staff inbox and follow‑up burden while advancing value‑based care (see the Medical Brain Orthopaedics partnership announcement and OrthoCarolina's OrthoCarolina strategic partnership release with Medical Brain for deployment details).

Feature	Detail
Deployment scale	~300 providers; ~40 locations (Southeast)
Patient tool	Medical Brain® mobile app - 24/7 personalized guidance
Provider benefits	Automated follow‑up, extended monitoring, reduced workload
Strategic aim	Support value‑based care and improve postoperative recovery

“For decades, OrthoCarolina has been committed to providing patient-first comprehensive care across a wide array of orthopedic specialties, and the integration of Medical Brain® into our care continuum will help us to better meet patients' real-time needs while also accelerating our organizational value-based care goals.” - Dr. Bruce Cohen

Novant Health has layered FDA‑cleared imaging AI into emergency workflows to speed diagnosis and treatment for time‑sensitive conditions: the system was the first in the Carolinas to deploy Viz.ai's LVO triage and care‑coordination tools across Novant Presbyterian and Forsyth Medical Centers, automatically analyzing CT angiograms and pushing suspected large‑vessel occlusion alerts to specialists' phones so teams can act in parallel rather than serially; Viz.ai reporting shows measurable time savings (Novant sites saved up to ~20 minutes per hospital and reported shorter door‑to‑treatment times), and Novant later added Aidoc's FDA‑cleared algorithms (intracranial hemorrhage, pulmonary embolism and others) to flag other acute findings and help shorten ED length‑of‑stay by about an hour in published studies - critical because every minute of untreated stroke costs roughly 2 million brain cells.

Learn more from Novant Health's deployment announcement and Aidoc's partnership overview.

Vendor	Use case	NC deployment	Noted impact
Viz.ai	Automated LVO detection + care coordination	Novant Presbyterian & Forsyth Medical Centers	Up to ~20 minutes saved per hospital; reduced door‑to‑treatment times
Aidoc	FDA‑cleared triage for ICH, PE, other acute findings	Novant Health network (system‑wide rollout)	Flags acute pathology to prioritize reads; Yale study found ~1 hour ED LOS reduction for ICH

“Time is very critical for the brain and we need to shave off minutes every opportunity we can.” - Dr. Laurie McWilliams, Novant Health neurointensivist

Novant Health deployment announcement and Aidoc partnership overview

Atrium Health and WakeMed have deployed AI to triage and draft patient‑portal messages, turning noisy inboxes into curated worklists. WakeMed reported a reduction of about 12–15 patient messages per provider per day after using AI to filter and draft responses, while Atrium's workflow produces initial AI‑generated replies that clinical teams edit before sending, keeping clinicians in control of content and clinical judgment (NC Health News article on North Carolina providers harnessing AI (2025)).

Early evaluations show a tradeoff worth noting: AI drafts can improve understandability and perceived empathy but tend to be longer and sometimes use more complex language, so human review remains essential to preserve nuance, equity, and the doctor–patient relationship - an outcome that matters because shaving routine message volume can directly lower inbox burden and protect clinician time for higher‑value patient care.

Wake Forest University School of Medicine leverages an EHR‑embedded risk signal - centered on its electronic Frailty Index (eFI) - to screen cognitive and functional vulnerability at scale: the eFI combines diagnostic codes, labs, vitals and measures of physical and cognitive function to flag older adults for targeted follow‑up, perioperative planning, or population outreach (Wake Forest electronic Frailty Index (eFI) overview).

That objective signal matters because frailty identified by the eFI strongly predicts utilization and harm (frail individuals had roughly 8× as many hospitalizations and 6× as many injurious falls) and was used during COVID telehealth outreach to contact nearly 700 frail patients, with more than 200 requiring clinic visits - showing the tool turns passive data into actionable care pathways.

Those EHR signals align with growing evidence that non‑pharmacologic interventions can improve cognition (see the U.S. POINTER lifestyle trial) and with machine‑learning work that predicts new cognitive impairment from routine EHR data, together forming a practical, evidence‑linked approach to early detection and intervention in North Carolina care settings (U.S. POINTER lifestyle trial press release, BMC Geriatrics study on prediction of cognitive impairment from EHR data).

A memorable operational payoff: replacing guesswork with eFI data reclassified about 40% of older patients compared with an “eyeball test,” enabling more precise care decisions.

Feature	Detail
Tool	EHR‑embedded electronic Frailty Index (eFI)
Predictive power	~8× hospitalizations; ~6× injurious falls (frail vs. non‑frail)
Operational use	Telehealth outreach contacted ~700 frail patients; >200 needed clinic visits

“My dream is that eFI will be like a geriatrician at your fingertips.” - Kate Callahan, MD, MS

Duke Health's Sepsis Watch pairs a deep‑learning early‑warning model with a disciplined clinical workflow so follow‑up and preventive outreach actually lead to faster treatment: the system - built at the Duke Institute for Health Innovation - was trained on roughly 42,000–50,000 patient encounters (>32 million data points), continuously evaluates 86 variables (sampling EHR data every five minutes), and predicts sepsis a median of about five hours before clinical presentation; that lead time has translated into doubled 3‑hour SEP‑1 bundle compliance and reported sepsis mortality reductions of roughly 27–31% across sources, an operational payoff equivalent to an estimated eight lives saved per month and clearer, prioritized outreach for at‑risk patients (DIHI's platform work also supports automating patient connection for social needs).

Read the Sepsis Watch implementation and outcomes on the Duke Institute for Health Innovation site and Duke's reporting on clinical impact for deployment details.

Metric	Value
Training data	~42,000–50,000 encounters; >32 million data points
Variables monitored	86
Monitoring frequency	Every 5 minutes (EHR sampling)
Median prediction lead time	~5 hours before clinical presentation
Reported mortality reduction	~27–31%
SEP‑1 bundle compliance	Doubled after implementation

“A lot of people develop AI models, but not many are integrating them into clinical practice to improve clinical outcomes. That is a huge differentiator for us at Duke.” - Suresh Balu

Novant Health's Behavioral Health Acuity Risk (BHAR) model was built to strengthen suicide prevention by flagging patients who may not self‑report ideation: a random‑forest model trained on historical EHR data (and external sources) runs near‑real time inside the EHR and displays a numeric risk percentage plus a color‑coded low/medium/high banner so clinicians see actionable risk without waiting for voluntary disclosure - helping teams prioritize outreach and clinical intervention.

The model, developed by Novant Health Cognitive Computing teams in Winston‑Salem and Charlotte, was implemented with clinician partnership and human‑centered workflows; early reports describe improved identification and patient engagement and a plan to share the approach with other systems.

Read the development and validation details in the medRxiv preprint and Novant's implementation profile on CIO for deployment and workflow context.

Feature	Detail
Model	Behavioral Health Acuity Risk (BHAR) medRxiv study and validation
Technique	Random forest machine learning
Deployment	Hosted in EHR, updated near‑real time
Output	Real‑time risk percentage; color‑coded low/medium/high in chart
Purpose	Suicide prevention through clinical interventions and prioritized outreach
NC provenance	Novant Health Cognitive Computing; Forsyth Medical Center; Novant Charlotte Medical Group (Winston‑Salem & Charlotte)

Duke Health applied three machine‑learning models to thousands of surgical cases and found they predict operating‑room duration about 13% more accurately than human schedulers - models trained on and used across more than 33,000 cases are now in active use at Duke University Hospital to tighten schedules, reduce cases running past regular hours, and improve suite utilization; that small uplift translated into an estimated $79,000 reduction in overtime labor over a four‑month period in the study and made scheduling more reliable for clinicians and patients alike (see Duke Health algorithm improves scheduling surgeries at Duke Health and the Duke School of Medicine review: Artificial Intelligence in Health Care - Promise and Pitfalls).

Metric	Value
Accuracy improvement vs. humans	13%
Cases used / deployed on	More than 33,000 cases
Estimated overtime savings	~$79,000 over 4 months
Models	Three machine‑learning models; deployed at Duke University Hospital

“One of the most remarkable things about this finding is that we've been able to apply it immediately and connect patients with the surgical care they need more quickly.” - Daniel Buckland, M.D., Ph.D.

UNC Health has piloted a secure, internal generative‑AI chatbot - built on Microsoft's Azure OpenAI Service and integrated through Epic - to answer system‑specific questions, surface training materials, and draft routine EHR messages so clinicians spend less time hunting documentation and more time with patients; the June 2023 pilot began with roughly 30 clinicians and administrators in a tightly governed environment, with broader waves planned as teams validate safety, clinician review workflows, and prompt controls (UNC Health Azure OpenAI chatbot pilot announcement, coverage of the UNC Health Epic InBasket chatbot pilot and clinician oversight).

The practical payoff for North Carolina systems is concrete: conversational queries over a fed training library reduce time spent sifting hundreds of “how‑to” documents, and early testing emphasizes that AI drafts are tools to accelerate routine replies - not to replace clinician judgment - so human-in-the-loop review and governance are built into rollout plans.

Feature	Detail
Tool (internal name)	Ava (AI virtual assistant, per reporting)
Platform	Microsoft Azure OpenAI Service; Epic integration
Pilot start	June 2023 (small clinician/administrator cohort)
Hosting	Secure, governed internal environment
Primary aims	Reduce time searching training libraries; draft routine portal/in‑basket messages; streamline admin work

“This is just one example of an innovative way to use this technology so that teammates can spend more time with patients and less time in front of a computer.” - David McSwain, MD, Chief Medical Informatics Officer, UNC Health

North Carolina's AI story now reads less like unrestrained experimentation and more like pragmatic rule‑making: lawmakers and system leaders are pushing governance that balances measured benefits (faster triage, inbox relief, lives saved) with clear limits on bias, privacy and liability.

State reporting notes concrete risks - algorithms can underperform for patients of color, vast EHR pools raise privacy flags, and clinicians must avoid over‑reliance - so the NC Medical Board already stresses that physicians remain responsible for AI‑derived decisions and must review AI‑generated notes; meanwhile Sen.

Jim Burgin has signaled legislation to address accountability and a draft like NC S624 would formalize licensing, safety and privacy for chatbots. Health systems are responding with internal vetting and pilots (Duke, UNC, Novant) and workforce planning that shifts routine pre‑reads toward human‑in‑the‑loop validation - an actionable adaptation spotlighted in local training guides for Charlotte tech and clinical teams.

The practical payoff: a “gold‑standard” state approach could let hospitals scale proven tools safely while requiring the checks - human review, equity testing, and clear liability - that protect patients and clinicians alike (North Carolina Health News reporting on state oversight for AI in NC health care, North Carolina Senate Bill S624: proposed AI chatbot licensing, safety, and privacy rules, Local training guide: human‑in‑the‑loop image validation and workforce adaptation in Charlotte).

Cross‑cutting Issue	NC implication / example
Bias & equity	Algorithms can underperform for patients of color; systems must test and monitor (NC Health News)
Privacy	Large EHR datasets and ambient documentation raise data‑use concerns; policy needed
Liability & regulation	Sen. Burgin plans legislation; NC S624 would set licensing, safety and privacy rules
Governance & education	Internal vetting at Duke/UNC/Novant; workforce training to support human‑in‑the‑loop roles

“If it makes the wrong decision, where's the liability? Who's responsible?” - Sen. Jim Burgin

Practical next steps for beginners in Charlotte start small, measure concretely, and build governance into every pilot: first, get practical AI skills through a hands‑on program - Nucamp's 15‑week AI Essentials for Work bootcamp trains nontechnical staff to use tools, craft prompts, and apply AI across business functions (Nucamp AI Essentials for Work bootcamp - 15-week program); second, study safety‑first integration patterns such as AHRQ's prototype guide for ambient digital scribes so documentation pilots preserve privacy and reduce clinician burden (AHRQ guide for ambient digital scribes and documentation safety); third, design a tiny, governed pilot (one clinic or one surgical service) that enforces human‑in‑the‑loop review and equity testing, then track one clear metric - messages/day, minutes saved, door‑to‑treatment time, or a clinical outcome - and iterate (local pilots like Atrium's Project Nursing show how voice capture and small tests reveal practical tradeoffs).

These steps turn hype into readable, auditable gains and prepare teams to scale responsibly.

Step	Action	Resource
Learn	Practical, nontechnical AI skills	Nucamp AI Essentials for Work - 15-week bootcamp program
Study safety	Best practices for ambient scribes and documentation	AHRQ guide for ambient digital scribes and documentation safety
Pilot	Tight, human‑reviewed deployment with one metric	Coverage of Atrium Project Nursing AI pilot (Charlotte Business Journal)

“This is just one example of an innovative way to use this technology so that teammates can spend more time with patients and less time in front of a computer.” - David McSwain, UNC Health