AI Transparency

How we label AI in our email

Every email from this company ends with a line that says how much AI was involved. Three labels cover every case. This page explains what each one means, and how the fully automated tier works.

The Labels

Three labels, one per email

Label 1 · Written and sent by AI

An AI drafted this email and an AI sent it. No person read it before it left. The mail comes from the AI's own mailbox, so the sender line tells you before the signature does. A human is cc'd on every send, visibly, and a human reads every reply. Nothing auto-responds. The identity block names no model or vendor on purpose: the AI behind the mailbox can change, and the mailbox stays the same.

Label 2 · Drafted with AI, reviewed and sent by a human

AI produced a substantial part of the draft. A person read it, edited what needed editing, and pressed send from their own mailbox. That person is accountable for what got sent.

Label 3 · Written by a human

A person wrote this start to finish. Spellcheck and grammar tools do not count as AI drafting.

The Rule

Default to disclosure

AI disclosure means labeling content when AI substantially generated it. The threshold is substantial generation: AI produced sentences that survived into what you received. Fixing a typo does not meet that bar. When a label is debatable, we use the more disclosive one.

This is the same rule we write into client AI policies, next to risk tiering, approved tool lists, and audit logging. We run our own email under it because we ask clients to run theirs the same way.

Read the company-wide AI policy guide →

FAQ

Frequently asked questions

Why label email for AI involvement at all?

Most AI-written email pretends a person wrote it. We think you should know who, or what, wrote to you, and whether a person approved it before it was sent. The label costs us one line per email.

What counts as drafted with AI?

AI produced sentences that survived into the sent email. Spellcheck, grammar fixes, and autocomplete do not count. When the call is close, we use the more disclosive label.

Can my company adopt this?

Yes. Disclosure rules like these are part of the AI policy work AutomationTactics does with clients. Book an intro call or start with the AI policy guide.