AI Use Policy

Last updated: October 16, 2026

This policy is part of the Terms of Service. It applies when a workspace on Pro Agency uses AI Sales Manager or AI Ads Manager. If you do not use those features, this policy does not turn them on.

1. What the features are

AI Sales Manager sends or drafts replies from instructions the workspace saved. It can act across SMS, email, WhatsApp, Messenger, and Instagram. It can reply, escalate to a human, or offer a booking. A human outbound message pauses it on that conversation. An admin can turn it off for a conversation or for the workspace.

AI Ads Manager advises, drafts Meta campaigns, and can run an optional daily optimizer inside caps you set. Drafts are not published until you publish them. The optimizer does not move money to Pipecove. Ad spend is paid to Meta.

Neither feature is included on Free Starter, Solo Sales, or Growth Team.

2. Your instructions

You write the instructions. They should cover the offer, the price rules, the tone, the topics the assistant must refuse, and when a human takes over. Instructions that tell the model to deceive, to ignore an opt-out, to impersonate a real individual, to invent credentials, or to evade a platform rule are a breach of the Terms.

We store instructions to run the feature. They are Customer Data. See the Privacy Policy.

3. Billing

Each AI Sales Manager reply is billed from the usage wallet by the tokens it uses, at the rate shown in the workspace. The $99 plan fee is not a pool of tokens. If the wallet cannot cover the call, the reply should not spend. Auto-recharge follows the setting you turned on.

AI Ads Manager drafting is part of Pro Agency except where the product shows a separate wallet charge. Any such charge is shown before it happens. Media spend is never a Pipecove charge.

Token use that produced a reply is not refunded because you dislike the wording. A charge for a reply our fault never produced is a billing error. Write to billing@pipecove.com.

4. Accuracy and authority

Models are wrong sometimes. They can omit a condition, misstate a price, or sound more certain than your instructions allow. You will review instructions and sample replies before you let the assistant send unsupervised. You will not tell a customer that a reply is a regulated professional opinion unless a qualified person is actually behind it.

We do not warrant that a reply is accurate, complete, or on-brand. We do not warrant leads, appointments, or return on ad spend. The product is written to avoid "guaranteed results" language in ad advice. Do not add it back.

5. People on the other end

Tell the truth if a person asks whether they are talking to a person. Do not design the assistant to insist it is a named employee. An opt-out, a stop word, or a platform block ends autonomous contact the same way it ends human contact. The Acceptable Use & Messaging Policy applies to every send, including a send the model composed.

Do not paste sensitive information into instructions that the reply does not need. Do not use the assistant to decide whether a person gets credit, housing, a job, or insurance in a way that would be an automated legal decision you have not separately justified.

6. Ads

Special ad categories remain special. The assistant must not be told to target housing, employment, or credit ads by age or gender when the platform forbids it. You review the draft before it spends money. Publishing is your act.

7. Data sent to a model

To produce a reply or a draft we send the model the instructions, the relevant conversation or campaign context, and the settings you chose. We do not sell that content. We do not allow it to be used to advertise to your contacts. Providers are subprocessors for that request, listed through the Data Processing Addendum.

8. Logs

The workspace can see what the assistant did. Logs exist so you can audit a reply and so the wallet ledger matches a real call. Retention follows the Privacy Policy. You may ask us to delete instruction text and logs we are not required to keep.

9. Suspension

We may disable AI for a workspace that uses it to spam, to deceive, or to break a platform rule, without refunding tokens already used. The rest of the plan remains, unless the conduct also justifies suspending the workspace under the Terms.

10. Contact

support@pipecove.com privacy@pipecove.com