Home / Security & data
Security & data

What we do with your data. Exactly.And what we don't have yet.

Your tickets hold your customers' worst days: a charge they didn't expect, an account they can't open. We treat that like it matters. Here's the full list of what we do, what we won't do, and where we're not there yet.

This is the least funny page on our site, on purpose. We asked the jokes to wait in the lobby. Everything below is plain and literal.

Book a call → Request the DPA
Our commitments

Eight promises.In writing, before we see a single ticket.

01

NDA and DPA first

We sign both before we get access to anything. No "quick look" at your data before the paperwork.

02

Never used for training

Your data and your customers' data are never used to train models. Not ours, not anyone's.

03

GDPR-compliant processing.

GDPR is the law we work under, and our DPA reflects it. The DPA is available before you share anything.

04

Role-based access

People see what their role needs, and nothing more. You decide who on your side can change flows.

05

Humans approve risky actions

Large refunds, account changes, anything you flag: you can require a person to say yes first.

06

A full audit log Live

Every step the agent takes is logged: what it checked, what it did, what the system answered.

07

Deletion on request

Ask us to delete your data and we delete it. You don't have to explain why.

08

Only to deliver the service

We process your data to run Pronomy for you. Not to sell, not to profile, not for anything else.

What we don't have yet

We're not SOC 2 certified.We'll say so the day we are.

Some vendors put a badge on the page and hope you don't ask. We'd rather tell you straight. If your security team requires SOC 2 today, we may not be a fit yet, and that's a fair call to make.

Not certified
  • NDA and DPA, signed before access
  • No training on your data
  • Audit log of every agent step
  • Your security questionnaire, answered line by line, including the "no" answers
  • SOC 2 report
  • ISO 27001 certificate

We don't list certifications we don't hold. When that changes, this box changes.

AI transparency

It says it's AI.Your customers deserve to know.

The agent is labelled as AI by default. In many places the law says customers must be told, and we think honesty is part of good support anyway. You choose how it introduces itself: the name, the tone, the words.

What makes support good isn't hiding the bot. It's fixing the problem, and handing off to a person the moment that's the right move.

Intro you chooseExample
Option AHi, I'm Pronomy, an AI assistant for Acme. I can check your account and fix most things right here. Want a person instead? Just say so.
Option BHey! I'm Sam, Acme's AI helper. Tell me what's going on and I'll sort it out.

Illustrative wording. "Acme" and "Sam" are placeholders.

Guardrails

You set the limits.The agent stays inside them.

Autonomy is set per action, per amount, per case type and per risk. Most teams start at the bottom of the ladder and climb when the scores earn it.

The autonomy ladder

  1. Analyze and recommendIt reads the case and suggests what to do. It touches nothing.
  2. Draft the reply and proposed actionsYour team reviews, edits and sends. This is shadow mode Live.
  3. Act after a human approvesIt prepares the action and waits for a person to say yes.
  4. Act on its own, inside an allowed flowOnly within the steps, amounts and rules you set.
  5. Mandatory handoffUnder conditions you define, it always goes to a person, with a summary.
Guardrails example flowRefunds

Amount limits

  • ✓Refund up to $50 → agent acts alone
  • ◷$50 to $200 → waits for human approval
  • ✕Over $200 → never; hand off to billing lead

Mandatory handoff when

  • →The customer mentions a chargeback or legal action
  • →The customer asks for a person
  • →The agent isn't sure what policy applies
  • →A system call fails or returns something unexpected

Illustrative limits. You choose your own numbers and conditions.

One more rule we never bend: a promise is not an action. The agent's reply reflects what the target system actually confirmed. If the refund call didn't succeed, it won't tell the customer it did.

Sub-processors

Listed in the DPA. Yours on request.

The full list of sub-processors that handle data for Pronomy is part of our DPA. Ask and we'll send it, along with the DPA itself, before you share anything.

Ask for the list
Responsible disclosure

Found a problem? Tell us first.

If you think you've found a security issue in Pronomy, email us directly. Give us a fair chance to fix it before you share it publicly, and we'll keep you posted on what we do.

oleksii@pronomy.io

Please don't access other people's data, disrupt the service or test against customers' live systems.

Straight answers

Security questions.Answered without the fog.

Do you train AI models on our data?
No. Your data and your customers' data are never used to train models. We use your data only to deliver the service to you.
Are you SOC 2 certified?
No, not yet. We'd rather say that plainly than hint at it. We'll say so the day we are. Until then, we sign an NDA and DPA before any access and answer your security questionnaire honestly.
When do you sign the NDA and DPA?
Before we look at anything: your tickets, your docs, your systems. Mapping your flows starts after the paperwork, not before.
Can the agent issue a refund without anyone checking?
Only if you allow it. You set limits per action, per amount and per case type. You can require human approval for anything risky, and define cases that always go to a person.
Will our customers know they're talking to AI?
Yes. The agent is labelled as AI by default, and in many places the law requires it. You choose how it introduces itself.
Can we see what the agent did on a specific ticket?
Yes. The audit log records every step: what it checked, which action it took and what the system answered. It's how Scorecards check that promised actions actually happened.
How do we get our data deleted?
Email oleksii@pronomy.io and ask. We'll delete it and confirm when it's done. No lock-in and no minimum term, so you can leave whenever you want.
Who are your sub-processors?
They're listed in our DPA, which we share on request. Ask and you'll get the list before you share any data.
P.S.

If your security team sends a questionnaire with three hundred rows, send it straight to me. I’ll answer every row myself, including the ones where the honest answer is “not yet”.

OleksiiSay hi on LinkedIn →

Bring your security team.We'll answer every question, even the awkward ones.

Book a call → Write to us instead