How the Brauni AI understands your way of working as a psychologist

"And AI understands that I work from the cognitive-behavioral?" - is one of the most frequent questions we receive from psychologists testing Brauni for the first time. The short answer is yes. The long answer is much more interesting.

The Brauni AI Assistant is not a generic chatbot. It is a system specifically designed for clinical psychologists that adapts to your therapeutic orientation, knows your patients and learns from your way of documenting. In this article we explain, with transparency, how it works inside.

Your professional profile: the starting point

When you set up your account in Brauni, you can complete three fields that define your professional identity:

• Theoretical orientation: your theoretical approach (TCC, psychoanalytic, systemic, humanist, integrative, etc.)
• Working Methodology: how to structure your sessions (duration, frequency, techniques you use)
• Professional profile: your experience, specializations and areas of interest

These three fields are the first level of context that AI uses to adapt each response. They are not rigid forms: they are free text fields where you describe your practice with your own words.

What does AI do with this information?

Every time you interact with the assistant, your professional profile is injected into the conversation as context. This means that AI always knows who you are and how you work, without you having to repeat it in each query.

If your orientation is cognitive-behavioral, AI will:

• Suggest evaluation tools related to your approach (BDI-II, BAI, STAI)
• Structure session notes in the format you would use
• Prioritizing evidence-based interventions
• Talk in terms of cognitions, schemes and restructuring techniques

If you work from the psychoanalytic, the suggestions change completely:

• Instruments such as TAT, Rorschach or SWAP-200
• Focus on transfer, resistance and unconscious processes
• Interpretation-oriented notes and associative material

Patient context: AI knows the story

The second level of context is the specific patient you are working with. When you open the AI assistant in a patient’s profile, the system automatically gives you access to:

• Basic data: name, age, gender, treatment status, starting date
• medical records: history, diagnosis, therapeutic objectives
• Contacts: tutors, emergency contacts (relevant to minor patients)

This information is injected directly into the wizard's instructions, so from the first message the AI already knows who you're working with.

Access to pre-session notes

This is where AI becomes really useful. The wizard has the ability to read your previous session notes when you need it. It can:

• List sessions by date, with their metadata (session number, mode, duration)
• Read full content of specific notes you need to answer your query
• Find patterns semantically throughout many sessions

The latter is key: if you ask him "how did this patient's anxiety evolve in the last 3 months?", AI does not look for the word "anxiety" literally. Use semantic search to find related mentions even if you use different terms such as "excessive concern," "rumiation," "panic attacks," or "anticipatory malfeasance."

Semantic search: how you find patterns in your notes

Brauni uses a vector embeddings to understand the meaning of your notes, not just the words. It works like this:

1. Each session note you save is transformed into a number representation (a vector of 3.072 dimensions) that captures its semantic meaning
2. These vectors are stored in the database using pgvector, a specialized postgreSQL extension
3. When AI needs to search for relevant information, turn your question into a similar vector and search for notes with closer meaning

What does this mean in practice?

Imagine that you have a patient with 40 sessions. You ask AI: "When did family conflicts begin?" The system:

1. Turns that question into a semantic vector
2. Compare that vector against the vectors of 40 notes
3. Find the most relevant 5 notes, even if you have never written the phrase "family conflicts" textually
4. Presents that information to the AI to respond to your query

This is radically different to search by keywords. The system understands concepts, not just text.

AI tools: what you can do (and what not)

The AI wizard has access to a specific set of tools that allow you to interact with your patients' information. Each tool is designed for a specific purpose:

Patient information

Meetings and notes

Documentation

What can AI create?

The AI can create documents in your clinical notebook: session preparations, reflections, progress summaries or any note you need. These documents are saved as drafts that you review and edit before you finish.

In addition, from a chat-separated interface, you can ask AI to help you complete or edit existing session notes. Chat itself does not directly modify documents: it is always an action that you start from the editor.

How do you know which tool to use?

AI does not need to be told what tool to use. When you ask a question, decide autonomously what information you need to consult. Some examples:

"What was the original reason for consultation?" → AI calls the medical records tool and looks for history.

"How did the mood evolve in the last 5 sessions?" → The AI lists the last sessions, reads their notes and generates a summary of evolution.

"I need to prepare tomorrow's session, what issues remained pending?" → AI semantically searches for unfinished topics in the last notes and generates a preparation guide.

"Arm me a draft report for health insurance" → The AI reads the medical records, recent notes and generates a document in your notebook with a draft that you review.

Smart Form Completed

In addition to chatting, AI helps complete clinical forms such as session notes, patient profiles, and medical records.

When recording or writing a session summary, AI can automatically extract:

• State of mind at the beginning and end of the session
• Main topics discussed
• Techniques or interventions used
• Therapeutic approach applied
• Relevant clinical observations

And here again your therapeutic orientation matters: if you work with CCT, AI will look for automatic thought records, intersession tasks, and cognitive restructuring. If you work with a systemic approach, it will identify family dynamics, alliances, and relational patterns.

Context security: how your information is protected

This is the part that matters most to us. Having an AI that accesses clinical information requires an exceptional level of security. This is how we implement:

Insulation by professional

The AI can only access data from your patients. Each tool validates that the patient belongs to the professional doing the consultation. It is impossible for the AI of one psychologist to access data from another.

Separation of professional context

Your professional profile (orientation, methodology, bio) is injected as a separate message, not as part of the basic instructions of the system. Why? For safety. By keeping it separate, the risk of malicious content in a text field is mitigated.

Encrypted Embeddings

The semantic vectors of your notes are encrypted with Fernet before they are stored. They are only deciphered at the time of search. Even the Brauni technical team cannot read the stored embeddings.

No training with your data

The Brauni AI uses third-party language models (Google Gemini via Vertex AI) but never send your data to train models. Clinical data is sent as a temporary context to generate a response and is not stored on AI provider servers.

Access audit

Every interaction with AI is recorded in the audit log: what queries you did, what tools the AI used, what data you accessed. Full traceability.

AI as a co-pilot, not as a replacement

It is important to clarify something: Brauni AI is a assistant, not a diagnoser. Its role is:

• Help you quickly access relevant information
• Identify patterns that could go unnoticed in 40+ sessions
• Reduce administrative documentation time
• Suggest (never impose) instruments or approaches compatible with your guidance

You are still the professional who makes clinical decisions. AI saves time, organizes information and gives you perspective, but clinical evaluation and professional judgment are always yours.

A system that improves with your use

The more documents in Brauni, the more context the AI has to help you:

• More session notes = best semantic search and pattern detection over time
• Full medical records = More precise answers about history and evolution
• Active Clinical Notebook = more material to prepare sessions and reflect

You don't need to "teach" anything to AI. Your way of working is already in your professional profile and your clinical documents. AI simply reads, understands and adapts.

---

The Brauni AI Assistant uses advanced language models via Google Vertex AI. Clinical data is processed under strict security protocols and is never used to train artificial intelligence models. Clinical fields are encrypted with Fernet (AES-128) and also operate on secure infrastructure with rest encryption and TLS 1.3 in transit. Brauni complies with the principles of Personal Data Protection Act 25.326.