This page is for insurance agencies. It explains how Ventar's renewal assistant handles your clients' information. The same commitments are in the agreement we sign with you.
The short version
- The assistant runs in your own accounts (Google Workspace and n8n), which you open in your agency's name. Your policy list, activity log and emails stay there. The AI runs on Ventar's Anthropic account, in a workspace used only for your agency.
- Your agency is in charge of your clients' information. Ventar acts only on your instructions, as your service provider, under a written agreement.
- The AI sees the minimum it needs. Premium figures from your policy list never go to it: premium changes are worked out by plain rules in your sheet. Social Security, license, bank and card numbers in client replies are masked before it reads them.
- Licensed people stay in charge: they answer every client question and approve every notice with premium figures.
- If something goes wrong, we tell you within 48 hours. When you leave, we remove our workflows and our access, and your data stays with you.
Your role and ours
Your agency collects and controls its clients' information. Ventar is your service provider. We use your clients' information only to provide the renewal assistant to you. We do not sell it, share it for anyone's marketing, combine it with other data, or use it to train AI models. This follows the service-provider rules of the Gramm-Leach-Bliley Act and the state insurance privacy laws based on it. We support your obligations under state insurance data security laws, such as those based on the NAIC Insurance Data Security Model Law and New York's cybersecurity regulation (23 NYCRR Part 500), where they apply to you.
What information the assistant uses
- From your policy list: client name, email, phone, line of business, carrier, renewal date, expiring and renewal premium, producer and CSR, and dates such as when the client joined.
- From client replies to the assistant's emails: the text of the reply, with the quoted history removed.
- What it creates: status, notes, summaries of replies (labelled "AI summary"), and an activity log of every action.
The assistant does not need Social Security numbers, driver's license numbers, dates of birth, bank or card details, or health information. Please keep them out of the policy list.
Where it lives
| Service | Account holder | What it holds |
|---|---|---|
| Google Workspace (sheet and mailbox) | Your agency | The policy list, activity log, and all emails to and from clients |
| n8n Cloud | Your agency | The workflows, the stored sign-ins they use (encrypted by n8n), and a short run history. From go-live it keeps full details only for runs that fail. |
| Anthropic (Claude API) | Ventar, in a workspace with its own key and monthly limit, used only for your agency | The limited information below, while it writes or reads an email. Under Anthropic's commercial terms, API data is not used to train its models, and retention follows those terms. |
| Ventar's own systems (Google Workspace) | Ventar | No copy of your policy list. Failure alerts and support emails sent to Ventar can contain a small amount of client information, such as a name or email address in an error message. They stay in Ventar's mailbox, protected by two-step sign-in, and are deleted within 30 days after the problem is fixed. |
What the AI sees
- When writing an email: the client's name, line of business, carrier, renewal date, your agency's name and phone, and the CSR's first name. It writes only the greeting and closing; the questions, disclosures and signature are fixed text.
- When reading a reply: the reply text, with Social Security, driver's license, bank account and card numbers masked first, plus the line of business, carrier and renewal date.
- Never sent to it from your records: email addresses, phone numbers, premium amounts or policy numbers. A client's reply is passed on as written, apart from the masked numbers.
Safeguards built into the assistant
- A code check holds any AI-written draft that mentions coverage, promises anything, contains amounts or percentages, mentions discounts, asks for sensitive information, mentions binding, or describes a premium change in the wrong direction. Held drafts go to a person.
- Renewal notices with premium figures are sent only after a person approves them.
- Every client email says an AI assistant helps your office. Email only, never texts or calls.
- A trial week in which every email goes to your own staff first, a switch that stops everything, a limit on emails per day, and an email to you whenever a scheduled run fails. Nothing is retried automatically.
Who at Ventar can access it
Only named Ventar staff who set up and support your assistant, through the access you give them in your own accounts, with two-step sign-in. We access your accounts only to set up, support, fix or update the assistant, or when you ask us to. You can see and remove our access at any time.
If something goes wrong
If we become aware of a security incident affecting your data, we tell you without undue delay and within 48 hours. We tell you what happened, what information may be affected, and what we are doing about it, and we help you meet your own notice obligations to clients and regulators.
When you leave
Within 10 business days of the agreement ending, we delete our workflows from your n8n account, remove our access to all your accounts, delete any of your data we hold (we aim to hold none), and confirm this to you in writing. Your policy list, activity log and emails stay in your accounts.
Providers involved
Google (Workspace) and n8n (automation): your agency holds the account and contract with each. Anthropic (AI): Ventar holds the account and contract, and the cost is included in your package. Ventar's own Google Workspace mailbox receives failure alerts and support emails, which can include the small amount of client information described above. These two are the only providers Ventar uses for your clients' information. We will ask you before we add any other.
Questions
info@ventarsolutions.com · We are glad to complete your security questionnaire once a year.