AI phone assistant for psychotherapy practices
A months-long waiting list and callers who need a person, not a script. Tel-Agent asks one question first and hands over on any answer but no.
- Crisis to a person
- Waiting list position
- Self-pay or insured
Why a psychotherapy practice cannot leave the phone ringing
Between sessions there is nobody at the desk, and a therapist in a session cannot be interrupted. So the phone rings out - and the person calling has often spent a week working up to making the call at all. An AI phone assistant for a psychotherapy practice is only defensible if it is faster at reaching a human than the alternative, which is nothing.
How the assistant handles a call to a therapy practice
It answers and asks one question before any other: is this an emergency.
On any answer other than a clear no, the script ends - the call goes to a person, or the caller is given the crisis number.
Only then does it take a name, whether they are already a patient, and self-pay or insured.
It asks when they could attend and for a safe time to call back.
The enquiry goes into one dated list, which is more than a notebook by the phone.
What the agent is told to ask
Written by you, in your own words, as the assistant’s instructions.
What a therapy practice assistant must never do
What it does
- Answers between sessions and after hours
- Asks the emergency question first, every time
- Takes contact details and availability only
- Records the enquiry with its date for the waiting list
- Keeps everything on the practice's own machine
What it never does
- Ask what the problem is - that is the appointment, not an intake field
- Assess, reassure or respond to a description of distress
- Promise a place or a date on the waiting list
- Continue the script when the emergency answer is anything but no
- Hold a recording anywhere outside the practice
Questions
Should an assistant answer this phone at all?
Only with a hard rule in front of it. The first question is whether this is an emergency, and any answer other than a clear no stops the script and reaches a person or gives the crisis number. Handled that way it is better than a phone that rings out, which is what most practices offer between sessions.
What must it never do?
It must not ask what the problem is, assess anything, or offer reassurance. It takes contact details and availability. A machine that responds to a description of a mental health crisis with a generated sentence is the failure mode this whole page exists to prevent.
Is what a caller says here health data?
Yes, and among the most sensitive there is. Article 9 of the GDPR covers it, and it is why the recording and the transcript staying on a machine the practice controls is not a preference here.
Can it manage the waiting list?
It can take the enquiry and write it into one searchable list with the date, which is more than a notebook by the phone does. It does not promise a position or a date - only the practice knows what is actually free.
Does the EU AI Act allow an AI to answer a first call from somebody in distress?
Yes, under Article 50 of Regulation 2024/1689, which asks that the caller be told they are speaking to a machine - that line belongs in the greeting you configure. Annex III names emergency healthcare triage, so anything that ranked callers by how ill they sound would be a high-risk system. The distinction the Act draws is between assisting and deciding, and this agent assists: it never judges how urgent a symptom is. Austria's KI-Servicestelle at RTR publishes the same reading of the transparency duty.
Who is the data controller for what a caller to a therapy practice says on the phone?
You are. A first call from somebody in distress is health data, which Article 9 treats as a special category, and Tel-Agent is self-hosted with your own model keys, so the recording, the transcript and the record it writes stay on your server and never reach Dpro GmbH. There is no processor agreement to sign with us because we never receive the data, and the source is AGPL-3.0, so you can check that rather than take our word for it.
Can a therapy practice run this in the United States?
There is no single federal AI statute to comply with, but a phone line carries older rules. A US practice is under HIPAA, and HIPAA is where self-hosting pays: there is no business associate agreement to sign with us, because no call reaches us. Consent to record is set state by state, and California and Illinois among others want every party to agree. Running the software yourself makes you the caller of record, so the disclosure and the recording switch are yours to set.
Not usable yet. The day it answers a real call will be a release, not a newsletter.
Watch the repository
