Before a nursing AI pilot, plan a local evaluation
What happened
The 2025 FUTURE-AI consensus guideline recommends evaluating local clinical validity, usability in real workflows, human oversight, and ongoing monitoring. WHO’s 2024 guidance announcement also identifies inaccurate or biased generative AI outputs and automation bias as concerns.
This is an explainer drawing on existing guidance, not a report of a new study or product release. Neither source establishes that a particular nursing AI tool improves patient outcomes.
Why it matters for a DNP project
A project proposal needs a specific task and an evaluation plan. “Use AI to improve patient education” leaves important questions unanswered: Which education task? Who reviews the draft? What would count as a useful improvement, and what would make the team stop?
For discussion with your advisor and site team, start with one narrowly defined workflow. Confirm the site’s approval, privacy requirements, and permitted technology before using any patient information. Do not treat this article as authorization to introduce an AI tool into care.
What to measure or check
The following are DNP Blueprint planning examples, not a validated checklist or a complete safety assessment.
- Local fit: Identify the intended population, setting, language, and task. Record where the available evidence differs from that intended use.
- Human review: Identify the reviewer and escalation route. In a hypothetical patient-education drafting pilot, define a substantive factual correction before collecting data. Record drafts needing such correction out of all drafts reviewed, rather than reporting a count without its denominator.
- Workflow: Define when review timing starts and stops. Record review time, missing required education elements, and staff-reported workflow problems. Account for correction work when considering potential time savings.
Agree in advance on who reviews problems and what finding would pause the pilot. Select measures with the clinical team and advisor; these examples do not replace tool validation or institutional review. A change in staff satisfaction or drafting speed alone should not be described as evidence of improved patient outcomes.
Source: WHO guidance announcement, 2024