Jefferson Health’s Colleen Mallozzi argues nurses must co-develop the rationale before ambient AI reaches the bedside. This piece uses a qualitative analysis of ambient AI adoption to show what researchers and UX teams should code for, measure, and iterate on when piloting ambient documentation. Key, evidence-based numbers: pilots began July 2026, Jefferson reports roughly 2.5 hours saved per nurse shift in potential documentation time, and 55% of their nursing staff have five years or less of experience. Read the original reporting at Health System CIO and see how to turn those signals into reproducible themes and stakeholder-ready evidence using Evidano. In the first 5–10 minutes you will get a checklist to run a nurse-led pilot, a short codebook blueprint, and the exact Evidano features to accelerate adoption without coercion.
Key Takeaways
Jefferson Health’s volunteer-first, co-design approach to ambient documentation caused different adoption patterns than top-down mandates, generating advocates and reducing churn in early pilots.
A qualitative analysis of ambient AI adoption should prioritize ownership language, skill-development narratives, and workflow fit, and then translate those themes into vendor requests and governance steps.
- Volunteer-led pilots produced advocates and slowed the churn seen in forced rollouts.
- Pilots reported roughly 2.5 hours returned per nurse shift and involved dozens of nurses in July 2026 pilots.
- Code for ownership, psychological safety (newer nurses show higher hesitancy), and mapping vocalized phrases to structured flowsheet fields.
Fast take + source
Fast take: Jefferson Health put nurses in the driver’s seat for ambient documentation through volunteer pilots, and that co-design approach changed adoption patterns in early pilots.
Full story: Health System CIO.
- Why it matters: Forced rollouts risk quiet abandonment, co-developed 'why' increases ownership.
- Evidence: Jefferson framed ambient capture as practice transformation; early pilots used Abridge and Microsoft Dragon integrated with Epic instances.
Findings snapshot
| Metric | Value | Source / Note |
|---|---|---|
| Estimated time returned per nurse shift | ≈ 2.5 hours | Mallozzi quote, reported July 6, 2026 |
| Percent of nurses ≤5 years experience | 55% | Jefferson internal workforce stat (reported in interview) |
| Pilot vendors | Abridge; Microsoft Dragon; Epic (testing) | Pilots across Lehigh and Philadelphia regions |
| Pilot scale | Dozens of users (early) | Not yet hundreds/thousands on nursing side |
| Published | July 6, 2026 | healthsystemcio.com article |
What happened: design, rollout, and frontline truths
What happened: Jefferson treated ambient documentation as a practice transformation and ran volunteer nurse pilots rather than top-down mandates.
Jefferson’s volunteer model produced advocates and reduced the churn that follows forced whole-unit rollouts.
- Nursing documentation is structured (flowsheets) unlike physician prose, vendors must map ambient capture to structured fields.
- A major human change: nurses must verbalize internal clinical reasoning, newer nurses are often more hesitant to vocalize uncertainty.
- Jefferson used 'real talk' sessions where nurses pointed out UX friction that fed back to vendors, a reversible, learning-oriented pilot posture.
Implications for qualitative analysis of ambient AI adoption (research & UX teams)
What to code for
Code for ownership language, skill development narratives, and workflow mismatches when analyzing ambient AI adoption.
Ownership language (volunteer vs. mandated): tag comments that indicate agency or resistance.
Skill development narratives: capture sequences where nurses describe learning to verbalize assessments.
Workflow mismatches: map vocalized phrases to flowsheet fields and log misaligned captures as friction points.
Segment-by-segment signals
Segment analysis should surface differences by experience and unit to target interventions.
Newer nurses (≤5 years): higher hesitancy, prioritize psychological safety themes.
Seasoned nurses: rapid adoption when invited to co-design, code for advocacy and peer influence.
Innovation units vs. volunteer units: compare drop-off rates and narrative framing around expectations.
Outcome metrics to pair with themes
Pair qualitative themes with adoption persistence, error rates, and time metrics to make findings actionable.
Adoption persistence (weeks post-pilot), errors attributed to mis-capture, time-to-complete shift documentation, and self-reported time for patient interaction.
Do more, faster with Evidano (map to this use case)
Evidano definition
Evidano is an AI-powered qualitative data analysis platform that ingests transcripts, vendor logs, and survey responses, supports transcription with custom dictionaries and PII redaction, and provides AI-assisted coding and cross-segment analysis.
Ingest the full pilot corpus
Ingest the full pilot corpus by uploading transcripts, vendor logs, and survey/spreadsheet responses to Evidano for unified analysis.
Evidano supports transcription with custom dictionaries and PII redaction, vital for clinician audio.
Build a reproducible nursing codebook
Build a reproducible nursing codebook by starting with the themes above and refining codes in Evidano.
Start with the themes above, import or refine codes in Evidano, then run AI-assisted coding to surface subthemes (for example 'psychological safety' → 'fear of judgement').
Compare segments and visualize the risks
Compare segments and visualize risks using Evidano’s cross-segment analysis and co-occurrence networks to see which friction points cluster with newer nurses or certain units.
Turn those visuals into change requests for vendors.
Close the loop with vendors
Close the loop with vendors by exporting time-stamped quotes and structured findings so vendors can reproduce issues.
Evidano’s translation and transcription options help when pilots span regions or accents.
Security & governance
Prioritize security and governance: keep data encrypted and private and ensure it is not used to train third-party models.
Evidano’s models are tuned for qualitative research and your data is not used to train third-party models, a critical reassurance for governance committees.
7-step checklist: run a nurse-led ambient AI pilot (research workflow)
This 7-step checklist turns observation into action and evidence for nurse-led ambient AI pilots.
- 1) Convene volunteer cohort and co-write the pilot 'why' (ownership criterion).
- 2) Define cohorts: new nurses, seasoned nurses, innovation unit, control unit.
- 3) Instrument capture: collect audio, flowsheet logs, and short post-shift surveys.
- 4) Transcribe with custom dictionary (drug names, local jargon) and redact PII before analysis.
- 5) Run thematic + cross-segment analysis to surface friction, safety signals, and time-savings.
- 6) Share actionable artifacts with vendors (time-stamped quotes, co-occurrence maps).
- 7) Iterate: retrain volunteers as peer advocates, repeat measurement at 30/60/90 days.
FAQ: qualitative analysis of ambient AI adoption
How do I compare units reliably?
Compare units reliably by defining the same exposure windows and normalizing by shift length.
Define the same exposure windows and normalize by shift length; use Evidano cross-segment frequency analysis to compare theme prevalence per 1000 words or per shift.
What about psychological safety?
Address psychological safety by coding for language that indicates fear, embarrassment, or reassurance.
Code for language indicating fear, embarrassment, or reassurance and pair qualitative tags with adoption persistence metrics to quantify impact.
Is this research clinical advice?
No, this guidance is operational and UX research methods, not clinical advice.
If using clinical outcomes, consult clinical governance and ethics review where required.
Conclusion & next steps
Jefferson’s approach shows ambient AI adoption is as much cultural and practice-based as technical.
A qualitative analysis of ambient AI adoption should center ownership, skill development, and structured workflow fit, then translate those themes into vendor requests and governance steps.
- Ready to run this pilot? Upload your transcripts and survey spreadsheets to Evidano by visiting Try Evidano for free to auto-generate themes, compare segments, and export evidence packets vendors can act on.
- Read the full report that inspired this post: Health System CIO.