Trust
The answers, before you have to ask twice.
If you run a law firm, a dental practice or an agency, someone is eventually going to ask where your clients’ information goes. This page exists so you can answer that without calling us first. It is written to be forwarded.
The short version
Five things true of every engagement.
One
Your data is yours
We process it to do the work you have asked us to do, and for nothing else. We do not sell it, we do not share it, and we do not mine it to build something for someone else.
Two
It stays inside your account
Every account is logically isolated in the United States: every read and write is filtered to your account, and a request for another account’s records returns not found. Your data is never reachable through another client’s account.
Three
The language models don’t train on your data
The language models the agents reason with run inside AWS Bedrock under our agreement with AWS. Bedrock does not use customer prompts or outputs to train the underlying models. Phone calls use separate speech vendors under their own terms, and your data never goes through a consumer AI service.
Four
A person is accountable
You get a named specialist, not a support queue. Their access is role-based, logged, and revocable by you at any time.
Five
We will put it in writing
A Business Associate Agreement where protected health information is involved. A confidentiality agreement and data-processing addendum for anyone else who wants one. Ask on the first call.
The processing chain
Every vendor that touches the work.
Short on purpose. If you are running vendor diligence, this is the list your reviewer wants, and we tell you before it changes.
Core
Amazon Web Services
Compute, storage, encryption keys, and the language models the agents reason with, through AWS Bedrock. United States, us-east-1.
Core
Neon
The managed PostgreSQL database behind the application. United States, us-east-1.
Core
Vercel
Hosting for the web application. Handles data in transit; nothing of yours is stored there. United States.
Core
Cloudflare
DNS, edge protection and identity gating in front of the service. Traffic in transit, encrypted.
Core
Doppler
Secrets management. Credentials and keys only, never client data. United States.
Voice, text messaging and physical mail bring in a small number of specialist vendors, and a dental engagement adds a clearinghouse and a practice-management connector. We name every one of them, and put the necessary agreements in place, before that part of the work goes live. Ask for the current list and we will send it.
Where the boundary sits
What the AI does not decide.
Boundary
It does not give professional advice
The agents do not give legal, medical, dental or insurance-coverage advice. When a caller asks for it, the agent says so and routes to a person. That boundary is written into the agent’s instructions, checked in our testing, and visible to you in the transcript.
Boundary
It is a screen, not a clearance
Where an agent checks something — a conflict, an eligibility, a coverage gap — it surfaces what it found in the systems it can see. The determination stays with your professional, and your existing process stays the record.
Boundary
It reminds; it is not the system of record
Deadline and follow-up agents sit on top of your calendar or docket and raise things early. They do not replace the system you are already responsible for keeping.
Boundary
It stops when it is unsure
Low-confidence work is queued for a person rather than filed confidently. We would rather cost you ten seconds of review than put one wrong value in a file.
Diligence questions
Asked and answered.
Do you train AI models on our data?
The language models that read, draft and reason over your documents run inside AWS Bedrock under our agreement with AWS, and Bedrock does not use customer prompts or outputs to train the underlying models. Phone calls use separate speech-to-text and voice-synthesis vendors under their own terms. We do not route your data through a consumer AI service.
Where does our data actually live?
In our service, hosted in the United States, in AWS’s us-east-1 region. Every account is logically isolated: every read and write is filtered to your account, and a request for another account’s record returns not found. Your data is not reachable through another client’s account.
Who can see our information?
The specialist assigned to your account, and the small technical team at our software provider, CapeFear.ai. Access is role-based and logged, and you can revoke it at any time, the same way you would revoke an employee’s.
Will you sign something about confidentiality?
Yes. Where protected health information is involved we sign a Business Associate Agreement before go-live. For law firms and anyone else carrying their own confidentiality duty, we sign a confidentiality agreement and a data-processing addendum. Ask on the first call and we will send both.
What happens to our data if we stop working together?
You can export it, and we delete our copies on request. Nothing is retained to build a product with.
Is the AI making decisions on its own?
No. The agents draft, extract, sort and remind, and a voice agent answers routine calls live. Drafted client communications such as letters, updates and documents are queued for a person to approve before they send. The two exceptions are deliberate: the voice agent speaks to callers directly, and scheduled follow-ups go out on rules you set. Anything the system is not confident about stops and waits rather than guessing.
Next step
Send this to whoever asks the hard questions.
If your reviewer, your carrier or your managing partner needs something this page does not cover, raise it on the call and we will put it in writing.