sofiaby FINdustries

Security your IT lead can review before the call

Your firm's data isn't used to train AI models. Nothing connects until your IT lead agrees the access. Every credential sits in a scoped vault, and every use is audited.

What happens to your firm's data

The questions your IT lead and your clients ask first, answered in a line each.

Never used for training
Your data isn't used to train AI models in our systems or by our model providers.
Your vendors' terms still apply
Systems you connect stay governed by your firm's own agreements with those vendors.
IT agrees every connection
Nothing connects until your IT lead agrees the access.
A never-touch list
Name the systems and folders Sofia must never reach. They stay out.
What we build is yours
The skills, workflows and knowledge built for your firm are yours to keep.

How a system gets connected

Your IT lead sees and approves every connection before Sofia can reach anything.

  1. Before anything connects

    We map your systems

    We list the systems your firm runs on, and the ones that go on your never-touch list.

  2. At sign-off

    Your IT lead agrees the access

    Nothing connects until your IT lead agrees the access. Anything on the never-touch list stays out.

  3. Every time she works

    Each person's own access

    Where a person has their own login, Sofia uses it and sees only what they can.

Keys and access, handled the way a security review expects

The access problems a reviewer looks for in an AI tool, and how Yuxibu, the platform behind Sofia, handles each one.

The usual riskHow Yuxibu handles it
One shared admin login for everythingEach person's own access, where they have one
Keys pasted into prompts and config filesKeys in a vault; agents never see platform secrets
No record of which key was usedEvery credential use is audited
Secrets echoed back in answers and logsSecrets stripped from output and logs
Anyone can connect a new systemNothing connects until your IT lead agrees

Who can do what, and the record it leaves

Roles and single sign-on decide who gets in. Audit logs show what happened after.

Four roles
Owner, admin, editor and viewer, set for each person in your firm's workspace.
Single sign-on
Your people sign in to the Command Center, your firm's AI workspace, with single sign-on.
Audit logs for admins
Tool use, credential use, file access and memory access are logged. Only admins see them.
Approvals where you want them
Put approval steps on any workflow that writes to a client system or sends messages.
Questions, not guesses
When a decision belongs to a person, Sofia stops, asks and waits for the answer.

What stops Sofia claiming work she didn't do

Every turn passes a rule-based review before it reaches you, and a person at your firm keeps the final say.

  1. Before she reports back

    A rule-based review

    Every turn passes a deterministic review before delivery. If Sofia says she filed a document, the record must show it.

  2. Before it reaches a client

    Your approval step

    Workflows that write to client systems or send messages can wait for a person at your firm to approve.

  3. Before it leaves the firm

    A professional signs off

    A person at your firm does the final review.

The company and the models behind Sofia

Three names you'll see, one company accountable for all of them, and the AI models that do the work.

Who builds it

One company behind the names

Sofia is the AI agent. Yuxibu is the platform she runs on. FINdustries, LLC builds and operates both. When you connect Google, Google's consent screen shows Yuxibu.

Explore
AI models

The best model for each task

Sofia uses the best model available for each task, from OpenAI or Anthropic's Claude. New models are admitted after evaluation, and none train on your data.

Explore
Releases

Tested before it reaches you

Every release is validated in staging before it reaches your firm, so your team isn't the test environment.

Explore

Questions firms ask

Is our data used to train AI models?

No. Your firm's data isn't used to train AI models in our systems or by the model providers we use. Systems you connect, such as your document or accounting software, stay governed by your firm's own agreements with those vendors.

Do you hold a security certification or attestation?

No. Your data is protected by roles, single sign-on, audit logs and scoped credentials, and our technical team walks your IT lead through each in the security review.

Where is our data hosted?

We don't publish hosting or data-residency details on this site. Your IT lead gets those answers from our technical team during the security review.

Which AI models see our data?

Sofia uses the best model available for each task, whether that's from OpenAI or Anthropic's Claude. Each model is evaluated before it's admitted, and none of them train on your firm's data.

Can Sofia send something to a client without anyone seeing it?

Only where your firm hasn't asked for an approval. Approval steps can go into any workflow that writes to a client system or sends a message, and we help you place them.

What do we keep if we leave?

You own the end product: the skills, workflows and knowledge built for your firm are yours to keep. Skills are plain files, and we deliver them to you on request.

Why Google shows Yuxibu

When you connect Google Calendar or Google Drive for use with Sofia, Google’s authorization screen identifies the application as Yuxibu, the platform that provides these integrations. See our Privacy Policy and Terms of Service.

Bring your IT lead to the call

We'll walk your IT lead or vCISO through the access model, the credential vault and the audit record, and answer your questionnaire in writing.

One real workflow, every month. The before-and-after time, what still needs review, and which systems it touched.

Know a firm that needs this? Partners and clients who refer firms to us are part of our partner network.

Refer a firm