AI Terms

1.Where Behavly uses AI

The AI Features are the following, and no others:

• Writing follow-up emails — one email per recipient, written from that person’s behaviour profile, your brand settings, your campaign brief and the earlier emails they received. In template mode the model rewrites only the text of your template.
• Adaptive campaign decisions — for each recipient in an adaptive campaign, deciding whether to send the next email now, wait, or stop, within the limits you set (maximum emails, minimum gap, quiet hours, exit on reply or login).
• Reply handling — classifying the intent of replies that arrive in your Connected Mailbox (interested, not interested, out of office, unsubscribe, referral, question, other) and, where you use it, drafting a reply for your review.
• Behaviour summaries — a one-sentence summary of an End User’s recent activity, used as context for the writing step.
• Segments from plain language — translating a description such as “people who looked at pricing twice” into rules you confirm before they are applied.
• Design and brand tools — generating and editing email templates, generating images from a description, suggesting subject lines, rewriting text you select, building an email signature, extracting a brand profile from your website, and guessing mail-server settings from a domain.
• The in-app assistant — answering questions about how to use Behavly.

2.Providers

Text features are powered by Anthropic’s Claude models. Image generation is powered by OpenAI. Both are used through their business APIs, under terms that prohibit the provider from using your prompts or the responses to train its models. We may change models or providers; if we add a provider that processes End User data we will tell you first, and the Privacy Policy will always list the current providers.

3.What reaches the model

Every prompt passes through Behavly’s privacy boundary before it leaves our systems:

• Email addresses, phone numbers and similar direct identifiers are replaced with consistent pseudonyms. Only a recipient’s first name is retained, and only where a greeting needs it.
• Fields whose names indicate sensitive content (for example passwords, tokens, card numbers) are removed.
• In strict mode, which a Workspace owner can enable, all free-text values are removed as well and only structured signals remain.
• Reply text is truncated and scrubbed of identifiers before classification.

The behaviour data that informs an email is the End User’s activity on your own property, as described in the Privacy Policy. Behavly does not enrich prompts with data from any other source.

4.Controls you have

• Off switch. A Workspace owner can disable all AI Features. Campaigns that need them pause until they are re-enabled; nothing is sent in the meantime.
• PII mode. Choose between pseudonymised prompts (the default) and strict mode.
• Approval. A campaign can require your approval of each generated email before it is sent. Adaptive campaigns with approval enabled hold every decision for you.
• Limits. Adaptive campaigns never exceed the maximum number of emails, the minimum gap, or the quiet hours you set, whatever the model suggests.
• Access log. Every AI use is recorded with its purpose, provider and model, and the log can be reviewed and exported in the Workspace settings.
• Retention. The activity data behind the profiles is deleted on the schedule described in the Privacy Policy.

5.Your responsibility for the output

Models make mistakes. They can misread behaviour, state something untrue about your product, use a tone you would not, or make a poor follow-up decision. Behavly adds safeguards such as URL checks, suppression of unsubscribed addresses and the limits above, but it cannot guarantee that any output is accurate, appropriate or effective.

You are the sender of every email and the author of every design that goes out under your name. You agree to review output where review is available, to configure limits that fit your audience, and to take responsibility for what is sent, whether or not you reviewed it. Do not rely on the AI Features to give legal, medical, financial or other professional advice to your recipients.

6.Ownership of output

As between you and Behavly, everything the AI Features produce for you — emails, subject lines, designs, images, summaries and drafts — is yours. We claim no rights in it beyond what we need to deliver it to you and to your recipients. Because models can produce similar output for different customers, we cannot promise that output is unique or protectable.

7.Prohibited uses

You may not use the AI Features to:

• generate deceptive, misleading or manipulative content, or content that impersonates a person or organisation;
• generate content that is unlawful, harassing, discriminatory or harmful, or that infringes someone else’s rights;
• attempt to extract the models’ instructions, to bypass the privacy boundary, or to make the models act outside the features described in section 1;
• process special categories of personal data, payment card data, or the data of children.

8.Fair use

AI Features cost us money per use. We may apply reasonable usage limits to protect the Service, and where we introduce paid plans, the included usage will be stated on the plan before you subscribe.

9.Changes and contact

We will update these AI Terms when the features or providers change, with notice as described in the Terms of Service. Questions: hello@behavly.io.