Weekly digest

Digest Issue #4

The Big Finding, Product Watch, Clinical Reality Check, Head-to-Head, and One Action This Week.

Last updated 2026-10-28

Published 2026-10-28. Vendor facts last checked 2026-10-26. Five sections, every claim sourced and labeled. See the methodology.

The Big Finding

An NHS trust is rolling out an AI scribe in older people's mental health services, led by clinicians, not by a vendor pitch. Rotherham, Doncaster and South Humber NHS Foundation Trust is deploying Accurx Scribe, built on Tandem Health's AI, in its older people's mental health services, as part of a trust-wide rollout reported October 20 (ResultSense / Digital Health, Oct 20). Appointments are long and sensitive (mood, memory, family situation, medicines), and each generates substantial documentation. The trust's clinical lead said the trust wants technology to "earn its place" and is using Scribe "carefully," with clinicians involved from the outset. Drafts are checked and signed off by clinicians before anything enters the patient record. Why this is important: older-adult psychiatry is where a scribe's failure modes (misheard collateral, a muddled timeline, a softened risk statement) carry the most weight, and where the documentation burden is heaviest. The coverage does not claim hours saved; clinicians are reported to be noticing a practical difference in day-to-day work.

Product Watch

  • Suki: Raised $90M Series E led by Healthier Capital (October 13), projecting $60 to $80M in 2026 revenue. On our scorecard Suki (48/100, #15) ranks where it does because public documentation scores 0/20 on pricing transparency and 0/20 on specialty and psychiatry fit. Full Suki review. A large raise is not a documentation score; if you are evaluating Suki, ask for published pricing and psychiatry-specific documentation its public pages do not yet show. Source, checked 2026-10-26. Documentation reviewed
  • VA ambient-scribe IDIQ: The September 22 multiple-award ceiling of $775.72M is a ceiling shared across vendors, not committed spend, covering SaaS scribe services across 9 million veterans and 1,380 facilities on VistA/CPRS and Oracle Health. Contract shape as reported in an ibl.ai analysis, not from the award notice itself. For behavioral health, this is the largest single buyer signal for ambient documentation in US care. Source, checked 2026-10-26. Documentation reviewed
  • Tebra AI Note Assist: Providers generated 1,065,000+ notes as of August (Business Wire, October 15). Tebra claims documentation time fell from about 23 to about 14 minutes per note, notes completed rising from 7.2% to 30.9%, and marathon sessions (30+ minutes) down 58%. Supports SOAP, Therapist Initial, Therapist Progress, and Psych Initial formats. Caveat: vendor-reported customer survey (2,297 paired responses, self-reported time ranges, not directly measured). EHR-integrated tool, not a scored scorecard scribe. Source, checked 2026-10-26. Documentation reviewed

Clinical Reality Check

Your signature owns whatever the scribe wrote. CMS fact sheet MLN905364 (July 2025) states: "If you use a scribe, including artificial intelligence technology, sign the entry to authenticate the documents and the care you provided or ordered" (CMS MLN905364). The Medicare Program Integrity Manual, chapter 3, section 3.3.2.4, adds that the signature affirms the note adequately documents the care provided. A healthcare compliance expert witness writing October 22 puts it plainly: CMS folded machine-written notes into the old rule, so the signer owns the note line by line (Medium, Oct 22). Reviewer patterns flagged: invented findings, copy-identical notes, complexity drift, and batch signing. Education, not legal advice. Pairs with our methodology; our Signing AI-Scribed Notes post covers the same ground.

A spoken "maybe" can become a charted diagnosis. On KevinMD (October 18), caregiver advocate Nicole Drapeau Gillen (not a clinician) argues an AI scribe in psychiatry can chart a spoken differential as a settled diagnosis, merge family collateral into the patient's history, and blur a spoken question with an answer (KevinMD, Oct 18). Underneath: a BMJ Innovations study of TORTUS across 5,719 consultations found that with both safety mechanisms on, unevidenced chunks fell from 3.3% to 1.6% and drift risk from 2.9% to 1.2% (BMJ Innovations, Oct 8). Caveats: general care, not psychiatry; three of four authors at TORTUS AI; drift judged by LLM-as-a-judge. A May 2025 medRxiv preprint also reported lower likelihood of any psychiatric intervention in AI-scribed visits (aOR 0.83); preprint, not settled (medRxiv).

Head-to-Head

TheraNest vs CharmHealth vs OptiMantra for psychiatric EHRs. Practice Stack, not the scribe scorecard. TheraNest (Ensora Health) leads at 74, CharmHealth and OptiMantra tied at 73, across 13 EHRs. Estimated monthly cost for one prescriber: TheraNest about $104, OptiMantra about $137, CharmHealth about $221. Choose TheraNest if published cost leads and you accept prescribing as a paid add-on with no documented lab or API integrations. Choose OptiMantra if you want the lower documented monthly costs and add-ons with e-prescribing included (PDMP workflow manual by default; integrated PDMP optional). Choose CharmHealth if you want a free edition path and EPCS bundled at $250 per provider per year, and can pay the higher monthly total. Read the full psychiatric EHR comparison.

One Action This Week

Before you sign your next AI-drafted note, run a psychiatry pre-sign check. Read the diagnosis line against what you said: differential or diagnosis? Check that family collateral is attributed to the family member. Check that spoken questions did not become answers. Then sign, knowing the signature affirms the note documents the care provided. Our psychiatry scribe checklist lists demo questions. If your vendor's template makes those checks hard, that is a demo question.

Also new on the site: first clinical accuracy pilot results are live for six psychiatry AI scribes (protocol v1.0.1): Heidi 96.0, Freed 91.0, JotPsych 90.2, Blueprint 83.2, Mentalyc 79.0 (2 critical), Upheal 63.0 (3 critical). Separate from the research scorecard. Also: four new scribe reviews on October 23 (Berries (70/100, then ranked 7), Medwriter (65/100, then ranked 8), PMHScribe (60/100, then ranked 9), DeepCura (60/100, then ranked 10)); scorecard then 19 tracked / 17 ranked as of 2026-10-23; Berries Pro listed at $79/month on its live pricing page (earlier figure was $99; scores/ranks/verdict unchanged); new telehealth therapy blog (October 23); Practice Stack EHR comparison now covers 13 EHRs with TheraNest #1; Signing AI-Scribed Notes live (2026-10-27).

Who curates this site

Curated by a practicing, board-certified psychiatrist, not named to avoid conflicts. From the site data: 19 vendors tracked, 354 distinct external source pages linked. Last site update 2026-11-20; last targeted vendor verification 2026-11-12; last complete research audit 2026-11-04. How we work · Editorial policy.

Get The Behavioral Health AI Brief

One short, sourced email a week: what changed in AI scribe pricing, privacy terms, and features. Free, unsubscribe anytime.