Governance and transparency

Trust comes from visible constraints, not “AI-powered” claims.

NurseNote is an early-stage documentation writing assistant. This page states how it works, where data flows, what has been tested and what remains unproved.

Current production method

Constrained generative transformation

The user selects Narrative, SBAR or SOAP and a detail level. The de-identified source is sent through Base44’s Core language-model integration with rules to use only supplied facts. Before anything is returned, deterministic markers and an independent model pass compare the draft with its source. A failed check withholds the draft without consuming the daily free use.

Controls in the product

Browser-side and server-side checks for common structured identifiers

Fictional or genuinely de-identified confirmation before submission

A constrained generation prompt that forbids invented clinical facts and decisions

Deterministic number, date, time and ambiguity checks plus a separate model verification pass

Side-by-side source comparison, an editable draft and review confirmation before copying

Structured output separating the draft, changes, checks and ambiguities

No source note or generated draft written to NurseNote application records

Allowlisted, consent-based analytics properties that exclude note text

Known limitations

—Generative output can still be inaccurate, incomplete or overconfident.

—Automated identifier checks cannot reliably detect every name or revealing combination.

—A short or ambiguous source may not contain enough information for a safe transformation.

—Local abbreviations and organisational styles vary, so a human must review the result.

—NurseNote has not been assessed or approved for live organisational deployment with patient data.

What is logged

Not saved to NurseNote history

Source note text and generated draft text are not written to NurseNote entities, product analytics or performance logs.

Minimised operational data

Daily-use counters, keyed one-way abuse-prevention identifiers, selected format, request duration, verification status, success state and consent-based funnel events may be processed.

Model processing is transient to the requested generation but may involve technical processing by Base44 and its configured model provider. This is why identifiable patient information is prohibited even though NurseNote does not build a note history.

Validation status

Automated tests cover identifier blocking, backend confirmation, daily and burst limits, structured response shape, copy-review gating, payment contracts and source-fidelity failure cases. These checks do not establish clinical safety, clinical effectiveness or regulatory approval. Product evidence counts pseudonymous devices or accounts rather than verified individual people, and internal administrator activity is reported separately.

Free proof

Three complete improvements per day

Paid test

£9 once for 30 days

Decision gate

Repeat use and real payments—not sign-ups

See the controls in the result.

Use a fictional example and inspect how NurseNote separates polished wording from human verification.

Improve a note