How it works
A therapist holds a lot in their head. The thing a patient said three weeks ago. The brother who keeps surfacing, always in the same flat tone. The week the sleeping got better and nobody clocked why. You carry it, hour after hour, and then someone new sits down at two o'clock and you start carrying theirs.
Intuita carries the long arc with you. After each session it follows one patient's story back across every session before it: the people who keep returning, the threads running quietly underneath, the pattern that only shows itself over weeks. Then it shows you what moved.
Why does that matter? Because a pattern can be spread across too many weeks for one tired human to hold all at once. Intuita keeps eligible session material connected. You stay the clinician and decide whether the connection matters.
What it sees
Choose a suitable session source For supported workflows, Intuita can use a live session, uploaded recording or therapist-authored recap. A recap is not a transcript and cannot recover omitted dialogue. Speaker separation and transcription can be wrong; the therapist verifies the source.
Engagement, read against the patient's own baseline Intuita keeps two views separate. A fixed per-session map shows elaboration and depth from the first eligible analyzed session. A longitudinal layer compares up to six channel families — reply pattern, depth, pauses, homework, attendance rhythm and talk share — only with this patient's own eligible earlier sessions. The first three eligible sessions do not produce longitudinal channel readings, and each channel has its own readiness. One quiet session can contain a moment worth inspecting; the overall trend moves only when the same channel shifts across two consecutive eligible sessions.
And it distinguishes some protest from withdrawal The engine explicitly separates bounded forms of connection-seeking protest from rejection of the therapeutic work. Feeling aimed at you — "why didn't you tell me sooner?" — can be movement toward the relationship rather than resistance. This is a narrow Dutch-and-English safeguard, not a complete attachment model: explicit rejection of task or frame still takes precedence. A patient who usually answers briefly is not flagged for being consistent; a quiet run must break that patient's own timed pattern and pass corroboration gates. It describes behaviour in the session, never the person.
The relational pattern that keeps returning (CCRT) What your patient sought, how the other person was described as responding, and how your patient felt or acted in the exchange — Intuita codes sufficiently evidenced, specific narrated moments through the Luborsky CCRT structure. Distinct recurring patterns remain separate across eligible individual sessions. The model points to a patient turn and code retrieves the complete supporting turn from the transcript; the quote was said, while the CCRT classification and explanatory reason remain AI-generated observations.
Therapist moves and the client's immediate response Intuita classifies the dominant move in a therapist turn — an open question, reflection, psychoeducation and nine other families — and the patient's response in the immediately following turn, bound to the topic where possible. It keeps raw counts, not averages: a technique followed sometimes by opening and sometimes by pushback remains visible in both places. This is immediate conversational adjacency, not effectiveness, causality or a recommendation.
Guidance for your next session After a qualifying analyzed individual session: one headline and a short queue of observable loose ends — explicit commitments, deferred subjects, possible friction, goals and between-session work. Transcript-backed categories are mechanically matched to session evidence; goal and homework references use a different, lighter gate. You review and edit every item, and can set aside non-safety threads. Qualifying explicit safety-language threads receive protected review handling; the feature does not perform risk assessment.
The long line: topics, course, and gaps Intuita follows topics as a tree that grows with every session, reads the course against therapist-set goals, and separately detects declining, reactive or collectively declining topic patterns from topic-level interaction, elaboration and response data. Topic gaps are a companion lens, not a consequence of the quiet-run detector. They require their own history and evidence; an unavailable or failed analysis must not be read as an absent gap.
The people in the story Who keeps surfacing in your patient's story? Intuita tracks who comes up in each session and lays out the line per person across sessions.
Notes that meet you halfway A draft note in one of nine progress formats — and then a separate check in the other direction. "Did I miss something?" first builds a checklist from the transcript without seeing the note, source-matches each excerpt, and then asks whether the saved note already carries the same substance. Low-confidence items are removed; verified medium- and high-confidence safety-category flags receive protected review handling but do not constitute risk assessment. In a session with multiple saved formats, the current checker cannot yet guarantee it read the requested format, so every result requires direct note review. EMDR measurements such as SUD and VOC are excluded from the AI draft schema. You review, edit and separately decide what to save.
How the hour actually moved How much did you talk, and how much did they? Open questions or closed ones? Intuita lays out the shape of the conversation as plain observation — no score, no verdict. A quiet mirror for the part of the work you can never see while you are inside it.
An assistant over eligible session material Where the assistant is enabled, it answers from the bounded case material available to it. An answer can omit or misread context and is not a colleague's clinical judgment; the therapist verifies it against the source.
The guardrails
One boundary runs through the whole system: these features are designed to produce observations, not diagnoses, risk assessments or treatment directives. Prompts, constrained schemas, mechanical evidence checks and output guards reduce error; they do not turn generated text into clinical truth. Every result requires therapist review, and saving a note is a separate therapist action. Core extraction supports Dutch and English; some optional generated narration is currently English-only. The detailed pages above state the evidence, readiness and limitations of each capability separately.
How to read the evidence
- Published construct: the clinical or research concept exists independently of Intuita.
- Product implementation: the page describes how Intuita operationalises that concept in software.
- Internally tested behavior: engineering checks support the stated behavior, but do not establish clinical validity.
- Unvalidated heuristic: a threshold or label helps organise review and should not be read as a clinical scale.
This page describes product method, not proof of improved care or independent clinical validation. Each capability page states its own source data, eligibility, limitations and therapist-review boundary.
Current implementation and review status
The methods and limitations were checked against the repository on 25 July 2026. Intuita is the publisher. No independent clinical reviewer, clinical-validity study or improved-outcome claim is presented.
Sources and questions
Are these capabilities clinically validated? Not as a combined Intuita system.
Does a source link make an AI interpretation correct? No.
Who decides clinical meaning and record content? The therapist.
Ask about the current implementation or review the therapy-note workflow.