ACEN-accreditedCase Study
Executive Summary
In the first 60 days of adoption with South Florida College of Nursing, HealthTasks Vision AI graded 1,403 skills checkoffs and reviewed more than 83 hours of student video. The result was more than 200 faculty hours saved, equal to 25 full workdays recovered.
Based on this early usage pattern, total savings reached approximately $10,018 and projected annual savings reached approximately $55,584, assuming a $50 per hour rate for an expert-level faculty member. Just as important, the ACEN-accredited partner's adoption curve accelerated sharply after launch, showing how quickly AI-enabled video skills checkoffs can become part of normal program operations.
Through October 13, 2026, faculty-reviewed item scores show 99.75% operational agreement (Cohen's κ = 0.99) across 45,597 rubric-item grades, and 99.09% agreement (Cohen's κ = 0.91) where educators scored. Dual-rated differences were a divergence of interpretation, not a split vote: faculty awarded credit the AI did not, and never took credit away from an AI satisfactory mark.
First 60 Days: ROI Snapshot
Checkoffs Graded
Total Video Time
Time Saved
Workdays Saved
Total Savings
Yearly Savings
Savings projection is based on a $50 per hour rate for an expert-level faculty member, supported by adjunct nursing instructor compensation benchmarks published by Salary.com.
Adoption Curve
The month-over-month trend shows a classic early adoption ramp: low initial volume during launch, followed by a steep increase as the partner operationalized AI grading in regular skills checkoff workflows.
What The Numbers Suggest
Rapid workflow adoption
Usage moved from a January baseline to 167 graded checkoffs in February, then to 1,240 in March. That pattern suggests once faculty and students enter the workflow, adoption can compound quickly.
Meaningful faculty time recovery
Across 1,403 graded submissions in the first 60 days, the partner recovered more than 200 faculty hours. That equates to roughly 8.6 minutes saved per graded checkoff.
Higher leverage than raw video duration
The platform processed more than 83 hours of student video while saving more than 200 hours of faculty time. That gap is directionally consistent with the reality that manual grading includes more than watching video alone. If faculty spend a conservative additional 5 minutes per checkoff on grading entry, feedback, and related administration, total manual effort rises well beyond raw video duration.
Competency Uplift Across Core Skills
Vision AI Skills Checkoffs generated significant average improvements across core procedures. Most foundational and safety-critical skills improved between 25 and 43 percentage points.
PPE Donning and Doffing
CPR Checkoff
Oxygen Therapy
OB APGAR
Manual Blood Pressure
IV Catheter Insertion
Sterile Glove Application
Ambu Mask Ventilation
These improvements occurred within days, and in some cases within hours, through a closed-loop feedback system. Students used structured AI-generated feedback to complete reattempt cycles and correct performance without requiring direct faculty re-evaluation for every iteration.
The pattern matters. The largest gains appeared in lower-baseline skills, indicating that structured remediation and effective feedback loops were driving performance improvement. Higher starting proficiency produced smaller deltas, which is consistent with normal learning-curve behavior.
Improvement was observed across infection control, airway management, maternal assessment, cardiovascular measurement, and IV procedural skills. This is not workflow digitization. It is measurable competency lift.
Inter-Rater Reliability at Scale
After the first 60 days, the same partner continued scoring satisfactory/unsatisfactory skills checkoffs through October 13, 2026. The unit of analysis is each rubric item on a completed submission. Two submissions still processing were excluded. No production data was changed. Items are graded 0 (unsatisfactory) or 1 (satisfactory).
Educators can intervene on every checkoff. When they leave the AI score in place, that is a real production decision, not a gap in the study. When they score, that is a second rater on the same item. Both views matter, so we report both. Agreement is defined as specified by the partner: a missing educator score counts as agreement, the same score counts as agreement, and different scores count as disagreement.
Completed submissions
Jan 25–Oct 13, 2026
Rubric-item grades
1,041 distinct items across 76 templates
Operational agreement
Cohen's κ = 0.99; 45,481 of 45,597 items; faculty could review every item
Dual-rated agreement
Cohen's κ = 0.91 on 12,770 educator-scored items
Finalized among completed submissions
Submissions with any educator item score
Submissions with any item overturn
Educator scoring is all-or-nothing in practice: 935 of 936 reviewed submissions have an educator score on every item. κ = 0.99 is how the workflow runs: AI first pass, faculty able to change any grade. κ = 0.91 is how the two raters compare when faculty do score. Both sit in the almost-perfect range on a 0/1 scale.
Where the educator actually scored (n = 12,770 items), agreement is 99.09% with Cohen's κ = 0.91. Chance agreement is high because most dual-rated items are satisfactory. Of those items, 12,654 match the AI score and 116 differ.
Confusion matrix
Completed rubric-item grades. Bold cells are where the two raters differed.
| AI \ Educator | 0 (unsatisfactory) | 1 (satisfactory) | (none) |
|---|---|---|---|
| 0 unsatisfactory | 617 | 116 | 8,791 |
| 1 satisfactory | 0 | 12,037 | 23,807 |
| null | 0 | 0 | 229 |
Divergence of interpretation
Not a split vote
Every dual-rated difference went the same direction. Educators upgraded an AI unsatisfactory mark to satisfactory on 116 items. They never downgraded an AI satisfactory mark. Among dual-rated items where AI marked unsatisfactory, educators awarded credit on 116 of 733 (15.8%). Faculty and Vision AI differ on whether a step was shown. When they differ, faculty award credit the AI did not.
Overturns are rare and clustered
51 of 76 templates have 100% agreement under the partner's rule. When any flip happens, it is concentrated: 45 submissions, mean 2.58 items flipped, maximum 14 in one submission. 97 of 1,041 rubric items have at least one disagreement; 944 have none.
Those items are documentation, reporting, verbalization, and holistic or abstract assessment steps—criteria that are easy to miss on video—not the core psychomotor sequence.
Why This Matters For Healthcare Education Programs
Faculty capacity expands without adding headcount
Recovering 25 workdays in just 60 days changes what faculty teams can sustain. Programs can redirect time toward coaching, remediation, curriculum improvement, and student support instead of repetitive grading workflows.
ROI appears early, not only at scale
Even in an early implementation window with a single partner, the operational and financial impact was already measurable. That shortens the timeline for leaders evaluating whether AI assessment workflows justify investment.
Competency lift strengthens the academic case
The value is not limited to operational efficiency. With foundational and safety-critical skills improving by as much as 25 to 43 percentage points, programs can point to measurable learning gains alongside faculty time savings.
Video review becomes more operationally realistic
As student video volume grows, manual review models become harder to sustain. Vision AI helps programs absorb that growth while keeping evaluation workflows timely and consistent.
Divergence of interpretation, not a split vote
When faculty and Vision AI differ, they differ on whether a step was shown. Faculty award credit the AI did not. They never take credit away from an AI satisfactory mark.
Financial value supports strategic adoption
Approximately $55,584 in projected annual savings at a $50 per hour expert-level faculty rate creates a concrete business case for scaling AI-enabled skills checkoffs across courses, cohorts, and additional competencies.
Conclusion
This early adoption case study shows that HealthTasks Vision AI can create measurable operational ROI and measurable competency lift quickly. In just the first 60 days with South Florida College of Nursing, the platform supported 1,403 graded skills checkoffs, reviewed more than 83 hours of student video, and saved more than 200 hours of faculty time.
The larger story is not only labor savings. Across core procedures, foundational and safety-critical skills improved by as much as 25 to 43 percentage points, showing that structured AI feedback and rapid reattempt cycles can improve performance while scaling assessment workflows. Faculty scoring through October 13, 2026 adds a reliability finding at production volume: Cohen's κ = 0.99 in the live workflow and κ = 0.91 where educators scored. Dual-rated differences were a divergence of interpretation, not a split vote.
For leaders evaluating AI in clinical education, this case points to a practical conclusion: Vision AI does not need years of deployment to demonstrate value. It can start delivering time savings, workflow leverage, stronger competency outcomes, and a credible financial return within the first semester of use.