A cut-paper therapist works beside a recorder and a laptop showing sound waves becoming notes.
AItherapy documentationprivacycompliance

Hathr AI or Session Notes for a Therapy Practice?

Photo of Dr. Sofia Reyes
Dr. Sofia Reyes Clinical Documentation & Compliance Editor 8 min read
Outline

Two different starting points

Therapists comparing AI tools need to know whether they are organizing existing documents or moving a new session into the clinical record. This article settles that choice by tracing what enters each workflow, what the tool retains, and who reviews and signs the result.

See how a session-to-note workflow works. Maya is a fictional client. Her therapist, Jordan, has a care-plan document that already exists and a new visit to document. For the existing document, Jordan begins in a document workspace: Jordan chooses the stored file from the library and enters a question or instruction. The workspace processes that file and may keep it in its library for later use. Jordan remains responsible for checking the output and deciding what belongs in the practice record. Hathr’s current web subscription describes a signed BAA and a stored file library, which can suit repeat work with existing documents.

For the new visit, Jordan begins with the session itself. After the visit, the clinician’s session content is used to draft a progress note, which Jordan reviews, edits and signs. The note belongs in the clinical record; the AI draft is not the signed note. In this workflow, the clinician starts with a new encounter rather than selecting an existing file. Jordan can compare both paths by noting what starts each task and where the reviewed result belongs.

Hathr says its web app is “Powered by the latest Claude AI,” and describes it as powered by Anthropic’s Claude on its pricing page and healthcare page, checked November 2026. Keep that Claude-backed document workspace distinct from a session-to-note workflow: ask whether your staff need help working with stored documents or a repeatable path from a visit to a clinician-reviewed note.

Score the workflow before a trial

Use the trial scorecard to compare both workflows against the same fictional task before testing any client information. Record which information enters the system, what stays in the workspace, who checks each result, and where the approved record belongs. Keeping the test fictional lets your team compare setup and review steps without introducing client data. Add one row for each task your staff perform regularly, then note which steps require clinician judgment or a practice-specific approval. Compare the time needed to locate source material, verify the draft, correct it and record any unresolved questions for follow-up during the pilot before choosing.

Choose a documentation workflow

Use the scorecard this week to compare setup, review, recordkeeping and client-choice requirements.

Free. We'll email the PDF link right away and add you to the therapist toolkit, sent every two weeks. Unsubscribe any time.

Before using client information, have your team compare both options with the same fictional task. Record what each workflow retains and how a clinician verifies its result.

Compare the starting point and record

WorkflowStarts withInformation enteredWhat persistsReview and signature
Hathr web subscriptionExisting files or clinical informationQuestions, summaries or record analysisFile library; Hathr’s pages describe this differently; confirm the current limit with HathrClinician checks output; practice sets record handling
Hathr developer APIApplication data, such as transcripts or bulletsSOAP-note automation and medical NLPZero data retention by default, per API pageDeveloper and clinician review; practice sets record placement
Session-to-note workflowNew visit or session contentClinician-supplied or captured session contentDraft, then clinical recordClinician reviews, edits and signs

Hathr describes the web subscription as a file library with tools to summarize notes, analyze records and organize information into templates. Its developer API is a separate service with SOAP-note automation, billing-code suggestions and lab-test processing. Do not treat API retention terms as web-library terms, or either description as evidence of a clinician-reviewed, signed-record workflow. Sources: Hathr healthcare page and Hathr HIPAA Compliant AI API page, checked November 2026.

Hathr’s pricing page lists a signed BAA and unlimited file uploads and downloads. Its pricing page describes a library of 100 files to choose from, while its homepage describes unlimited documents and up to 100,000 pages. Hathr’s pages describe this differently; confirm the current limit with Hathr. The pricing page lists the API separately. Source: Hathr pricing page and homepage, checked November 2026.

Decide what job staff need done

Start with the object your staff handle most often. If you are evaluating Hathr AI, compare its stored-file workflow with the session-to-note process your clinicians actually need. If they repeatedly ask questions about an existing policy, treatment-plan file or other document, a workspace organized around stored files may be the closer fit. If their recurring task is producing a progress note from a session and signing it into the record, assess a session-to-note workflow instead.

Before a pilot, write down the starting information, what the service retains, where the reviewed result goes, and which clinician checks and signs it. A file processor may return a polished summary while leaving the team to decide where it belongs and whether it is part of the record.

Hathr’s pages describe this differently; confirm the current limit with Hathr. Its privacy page says users can delete files, checked November 2026. Before a pilot, ask about retention behavior for the exact service and data path your practice would use, including backup purge timing. Do not infer web-subscription backup handling from the API’s zero-retention statement.

Set the privacy boundary

Do not test a product with a real note after removing only the client’s name. HIPAA Safe Harbor requires removal of 18 categories of identifiers concerning the individual and relatives, employers or household members, plus no actual knowledge that the remaining information could identify the person. Dates, unusual combinations and identifying details in free text can matter. Expert Determination is a separate method, not a lighter version of Safe Harbor. See the HHS de-identification guidance, checked November 2026.

Before you authorize an AI workflow, use the HIPAA risk assessment tool for small therapy practices to record safeguards, risks and ownership alongside your pilot approval. Explain the intended use to clients, give them a meaningful choice and a non-AI option, and specify how a clinician reviews output. Check local policy, your role and the actual jurisdiction. A product’s eligibility and a signed BAA do not by themselves satisfy HIPAA, professional duties or state law. For a separate question about account eligibility for a named general assistant, see Is ChatGPT HIPAA compliant for client notes?.

Counselors should explain AI use and obtain informed consent, critically evaluate the technology, and continue reviewing accuracy and appropriateness, according to the American Counseling Association’s recommendations for practicing counselors, checked November 2026. This is professional guidance, not a universal law. Explain what the system will do, its purpose and limits, and how a client can decline without losing a non-AI path.

State rules apply to specified professionals and tasks. Check the law that governs your practice before a pilot. Document the applicable requirements and confirm whether they cover your provider, task and jurisdiction. For a closer look at session-to-note steps, see how the workflow works.

Review the documentation itself

Hathr AI’s web workspace and a session-note workflow answer different operational needs. When comparing Hathr AI with a session-to-note process, judge the task and record each workflow actually supports, not a general claim about AI. A useful pilot measures more than whether a draft looks polished. Define what a correct note needs to include, which facts must come from the clinician, what errors would change meaning, and how every draft is checked before signing. If Jordan’s draft leaves out a change in risk or the clinician’s assessment, Jordan must correct it before signing and file the approved note in the practice record. See how notes are drafted and reviewed.

Watch one staff member test the workflow. Can they trace a sentence in the draft back to its source? For example, if the draft says the client denied a symptom, the reviewer should be able to verify that against the session material before signing. If your local policy permits a client-data trial, record the approved workflow, protections, responsible reviewer and stop conditions before using it.

For a generic task, a general assistant may be enough when staff use invented, nonidentifying information. A different job starts when clinicians need the same path from a session to a reviewed, signed note in the clinical record. Choose a service built for the work your team repeats, and check its actual data path rather than relying on broad AI capabilities.

Why Emosapien fits a session-to-note job

Our Scribe Agent drafts a progress note in SOAP, DAP, BIRP, GIRP or PIE format from the session. The treating clinician reviews, edits and signs every note. The workflow is designed for moving from session content to a clinician-reviewed note, rather than managing a library of existing files. Confirm the agreement and protections for the specific plan and workflow your practice selects; a BAA is one part of your obligations, not a legal guarantee.

Keep the decision on file

Before choosing either path, keep a short approval record for your practice:

  • The defined task, data entered and intended recipient.
  • What is stored, where the output goes, and the vendor’s written retention and deletion answers for the specific service.
  • The agreement and configuration that cover the chosen service, plus your review and signing procedure.
  • Your client explanation, consent process where required, refusal path and non-AI alternative.
  • The applicable professional, organizational and jurisdictional rules, with a named owner for reassessment.
  • Pilot criteria, known limitations, escalation route and the decision to continue, change or stop.

Review the record when the task, service, settings, vendor terms or applicable rules change. Keep the pilot narrow enough that clinicians can inspect the output and report problems before expanding use.

Keep the clinical note under a signed BAA

If the work starts with a new visit and ends with a clinician-reviewed, signed progress note, choose a workflow built around that responsibility. Draft the new visit’s note under a signed BAA.

Draft the new visit’s note under a signed BAA

The clinician reviews, edits and signs every progress note.

Start free
  • Free plan
  • No credit card
Session note review
Fictional · Clinician signed

Jordan reviewed

The draft reflects the visit; I corrected the plan before signing.

Clinician review

Every progress note is reviewed, edited and signed by the treating clinician.

Review a session-derived draft →

References

Free PDF: AI Assistant Workspace Approval Cards

Printable decision sheets for checking assistant accounts, features, copies, retention and clinician review before a workflow is approved.

  • Claude Referral Approval Card and Gemini Referral Desk Card
  • Copilot-to-chart worksheet and Psychiatric Referral Check
  • Research-session attachment and deletion worksheet and Chart Proofreading Field and Extension Checklist
  • Document workspace versus session-note trial scorecard

Free. We'll email the PDF link right away and add you to the therapist toolkit, sent every two weeks. Unsubscribe any time.

Ready to transform your practice?

Join 10,000+ therapists using Emosapien.