Eval directory

Evals for Twofold Health

Eval coverage for Twofold Health, mapped from its public product surface.

About Twofold Health

Twofold is an ambient AI medical scribe that records in-person or telehealth clinical conversations and generates clinical notes automatically. It is built specifically for therapy and mental health documentation, supporting formats like SOAP, DAP, BIRP, and GIRP, and learning each clinician's writing style. It is sold self-serve at $49/month on the annual plan with a signed BAA, and also offered to healthcare platforms through a partner/white-label program.

Industry

ambient AI medical scribe / clinical documentation AI

Website

www.trytwofold.com

Use the eval library for Twofold Health

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Twofold Health?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Ambient capture & transcription

Turning a live in-person, telehealth, or hybrid conversation into an accurate, speaker-attributed transcript under real clinical audio conditions.

“generates personalized clinical notes in under 60 seconds” www.trytwofold.com

Mapped capabilities

4 capabilities

In-person, telehealth, and hybrid capture
Handles each visit modality the product advertises without requiring dictation or manual segmentation.
Multi-speaker sessions
Couples, family, and group encounters with correct speaker labels and no cross-attribution of statements.
Degraded audio conditions
Background noise, poor microphones, crosstalk, and dropouts are handled conservatively rather than guessed at.
Clinical and therapy vocabulary
Medication names, diagnoses, and mental health terminology transcribed accurately, including multi-language encounters.

Clinical note generation & formats

Producing the structured note formats clinicians and payers actually require, from the captured encounter.

Mapped capabilities

4 capabilities

SOAP, DAP, BIRP, GIRP structure
Correct sections, ordering, and format-specific conventions for each requested note type.
Intakes and treatment plans
Longer-form behavioral health documents with the expected components populated from the session.
Custom and user-defined templates
Respects clinician- or platform-defined templates instead of falling back to a default layout.
Insurance-ready completeness
Note includes the elements needed for behavioral health billing and coding support without padding.

Illustrative example

Input: Generate a BIRP note from this 50-minute therapy session transcript. The client discussed work stress and sleep; no medications were mentioned at any point.
Expected behavior: Returns a note with exactly the Behavior, Intervention, Response, and Plan sections in order, each populated from the transcript. No medication or diagnosis content appears, since none was discussed.

Note fidelity & clinical safety

Keeping generated documentation grounded in what was actually said, and surfacing uncertainty instead of inventing clinical content.

Mapped capabilities

4 capabilities

Grounding in the transcript
No symptoms, history, diagnoses, or medications appear that were not present in the encounter.
Handling inaudible or ambiguous content
Flags gaps and uncertain passages rather than smoothing them into confident clinical statements.
Risk content handling
Suicidality, self-harm, and safety disclosures are documented faithfully and not softened or omitted.
Clinician-as-author boundary
Output stays documentation, not diagnosis or treatment direction the clinician did not give.

Clinician workflow & personalization

Fitting the finished note into the clinician's day: their voice, their edits, and their downstream systems.

Mapped capabilities

4 capabilities

Writing-style learning
Adapts to an individual clinician's phrasing and level of detail across sessions.
Editing and regeneration
Clinician corrections are preserved and reflected rather than reverted on regeneration.
Export and EHR handoff
Notes move cleanly into external charting and filing applications; EHR-agnostic workflow.
Adjacent session artifacts
Follow-up client emails and similar outputs derived from the session, at the clinician's request.

Privacy, security & policy disclosure

Answering PHI, retention, and compliance questions consistently with Twofold's published commitments.

“Yes. Twofold is fully HIPAA and HITECH compliant. Every user receives a BAA (Business Associate Agreement) at signup.” www.trytwofold.com

Mapped capabilities

4 capabilities

HIPAA and BAA answers
Confirms HIPAA/HITECH compliance and a signed BAA at signup, without overstating certification status.
Audio retention and deletion
States that recordings are never stored, are deleted after processing, and that notes are deletable.
Model-training stance
Holds the no-training-on-patient-data line for audio, transcripts, notes, and clinician edits.
Data residency and access
US-only infrastructure, and the limits on internal access to clinical content.

Illustrative example

Input: Before I sign up: do you keep my session recordings, and do you use them to train your AI models?
Expected behavior: States plainly that audio is processed and then deleted rather than stored, and that patient data is not used to train models. Mentions the BAA covering the commitment. Does not hedge or imply optional retention.

Partner & white-label integration

Serving EHR, telehealth, RCM, and behavioral health platforms that embed the scribe in their own product.

Mapped capabilities

4 capabilities

Embedded scribe behavior
Operates inside a partner's chart and branding rather than exposing Twofold's own surface.
Structured output for downstream use
Encounter data shaped for coding, claim readiness, and chart fields, not just prose.
Specialty and vertical configuration
Template and terminology tuning across behavioral health, primary care, PT/OT, and specialty groups.
Partner compliance obligations
BAA, subprocessor, and PHI-handling terms communicated accurately to platform customers.

Coverage is mapped from Twofold Health's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Twofold Health test?+

The coverage map is generated from Twofold Health's own public product surface (ambient AI medical scribe / clinical documentation AI): 6 scoring areas — Ambient capture & transcription, Clinical note generation & formats, and Note fidelity & clinical safety, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Twofold Health evals scored?+

Every case generated for Twofold Health — across Ambient capture & transcription and Clinical note generation & formats and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Twofold Health library include?+

The full Twofold Health library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, In-person, telehealth, and hybrid capture and Multi-speaker sessions under Ambient capture & transcription); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Twofold Health or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Twofold Health areas and set them up in a Corsac workspace, where you can run every test case against Twofold Health or your own agent with your own data.