The Audienceful Agent

Audienceful has an AI agent built into the product. It isn't a writing widget bolted onto the side — it can actually do things: write and restyle an email in the editor, build an audience out of a sentence, draft a whole automation, generate an image, look up how a send performed, and schedule the next one.

It can also go and read the web for you — paste the blog post you just published and get the newsletter version back, or ask it to research this week's news and write from what it finds.

You talk to it like you'd talk to a teammate. It does the clicking.

Where you'll find it

• The agent bar — the chat box that follows you around the app. Ask for anything from anywhere.
• The Agent page — open Agent in the sidebar for a full-screen workspace with your conversation history beside it, plus a Context tab for teaching it about your brand.
• In the email editor — the same agent, with the email you're looking at as its subject. This is where the writing and design tools live.

The agent knows what you're looking at. Open an email and "make this shorter" means this email. Open an automation and "add a follow-up two days later" means that automation.

Conversations are shared with your workspace, so a teammate can pick up a thread you started, and everyone sees what the agent changed.

What it can do

Write and design emails

• Write a complete email from a one-line brief — or edit just one paragraph, headline or button
• Restyle your theme: colors, fonts, spacing, dark mode
• Drop in preset blocks and saved components, multi-column layouts, callout boxes and product lists
• Set your shared header and footer
• Add merge tags and event properties for personalization
• Create a new draft, duplicate one, or start from a template
• Write in the voice of your own top-performing emails
• Undo its last edit — the newest body edit it made to the open email carries a Revert this edit button

→ Writing and editing emails with the agent

Write from the latest content — on the web or on your own site

This is one of the biggest things the agent does, and the easiest to miss. It doesn't only write from what a model happens to remember. It can go and read the real thing:

• Paste a link — your newest blog post, your changelog, your updated pricing page, a press mention — and it reads the live page and writes the email version of it. Up to three links at once, which makes a link roundup a single request.
• Ask it to search — "search for what's happened in AI this week" — and it runs a real web search, opens the top results, reads them, and writes from what it actually found, with the sources listed under its reply.

Publish a post, paste the URL, get the newsletter. That's the whole loop.

→ Writing from the web and your own site

Generate images

Ask for an image and it makes one, matched to the imagery already on your website. It can also edit an image that's already in your email — change the background, swap the subject, adjust the mood.

→ Generating images with the agent

Build audiences, automations and sends

• Turn plain English into an audience — including "everyone who didn't open the Black Friday email"
• Build a full automation — trigger, waits, tags and the email bodies, drafted for you
• Edit, summarize or delete an automation you already have
• Add or tag a contact, or run a bulk tag or list clean-up
• Schedule a send, with a live recipient count before you confirm

→ Audiences, automations and sending

Answer questions

• "How did my last newsletter do?" — opens, clicks, top links, revenue, deliverability
• "What's this contact been up to?"
• "How do I set up a custom domain?" — answered from these help docs, with a link to support when it can't
• "What did our competitors announce this quarter?" — researched on the web, with sources

→ Asking the agent about your data

How it works: proposals vs. actions

Some things the agent just does. Others it proposes, and waits.

• It does it — editing an email, changing your theme, creating a draft, generating an image. These are low-stakes and reversible, so it applies them right away. The most recent body edit for the email you have open carries a Revert this edit button on its message (which flips to Re-apply edit afterwards).
• It proposes it — audiences, automations, contact changes, bulk actions and scheduled sends. You get a card showing exactly what will happen (rules, steps, recipient counts) with Confirm and Cancel. Nothing is created until you confirm, and confirming doesn't cost you extra credits.
• Multi-step plans — for a request with several stages ("create a draft, write it, and schedule it for Tuesday"), it lays out an ordered plan and works through it, pausing whenever a step needs your confirmation.

Brand context and custom rules

The difference between generic AI copy and something that sounds like you is context. When you add your company website, Audienceful scans it and builds a picture of your business — what you sell, who you sell to, how you talk, and what your imagery looks like. Everything the agent writes draws on it.

You can edit all of it, add a tone of voice, and set standing rules it follows on every generation — all on the Context tab of the Agent page, on every plan. It comes in three parts:

• Brand — the business summary scanned from your site, plus a tone of voice in your own words ("warm and conversational", "B2B-formal, no fluff"). Edit the summary to correct anything the scan got wrong or add what your site doesn't say.
• Custom rules — standing instructions the agent follows on every generation, and which take priority over everything else it knows. One per line: "Always sign off from Sam." "Never use exclamation marks." "Always end with a single call to action." This is the right home for the corrections you find yourself repeating in chat.
• Images — a description of how AI-generated imagery should look, plus the brand images pulled from your site that steer it. See generating images.

Once you edit a field, an automatic re-scan won't overwrite it; the Re-scan website button is an explicit "start fresh from my site" and says so — it's limited to once a day per workspace, since it re-reads and re-summarizes your whole site. The scan also runs again on its own if you change your workspace's website. If your site sits behind bot protection that refuses us, the scan tells you rather than failing quietly, and you can fill the fields in by hand.

Context is scoped to your workspace and shared with your team.

One thing worth separating: the scan teaches the agent who you are, once. It isn't a feed of your latest posts. To write from something you just published, paste the link or ask the agent to search — see above.

Credits and plans

The agent is available on every plan, including Free. Usage is metered in AI credits: your allowance resets at the start of each calendar month and unused credits don't roll over. The meter in the agent panel shows what's left, and paid plans get substantially larger allowances. Image generation costs more than a text request and is available on paid plans only.

Credits are charged for the work the model actually does, so a reply you stop partway through can still cost the tokens it had already used. Confirming or cancelling a proposal is free — it isn't a new generation.

Your data and privacy

• Nothing you send the agent is retained by the AI models we route it through, and none of it is used to train them. Every request is sent with data collection denied, so it only ever reaches providers whose published policy is not to store or train on what they receive.
• Your emails, contacts and analytics are never shared across workspaces — the agent only ever sees the workspace you're working in.
• The live content of the email you're editing is passed along for that one request and isn't stored.
• Files you attach are stored privately in your workspace and reach the model through short-lived links that expire.
• Everyone in your workspace can see everything in the agent. Conversations, the files people attach and the contact lookups they run are shared with your team, the same as any other workspace data — and a conversation stays until the workspace is deleted.

Example prompts to try

The fastest way to get a feel for the agent is to throw real work at it. A sampler across everything it can do:

Writing and design

• "Write a newsletter announcing our summer sale — 20% off everything through Sunday."
• "Make the third paragraph punchier and cut the last section entirely."
• "Make this dark mode and put the three features side by side in columns."
• "Generate a photo of a ceramic coffee mug on a wooden table for the header."

From the web and your own site

• "Read https://ourblog.com/whats-new and turn it into a product update email."
• "Write our monthly newsletter from these three posts: [link] [link] [link]."
• "Search the web for what's happened in AI this week and write a roundup newsletter."
• "Research what our competitors shipped this quarter and draft an email about how we compare."
• "Read our launch post at [link], search for what people are saying about it, and write it like my best emails."

Audiences, contacts and automations

• "Build an audience of everyone who didn't open the Black Friday email."
• "Suppress all the hard bounces."
• "Build a three-email welcome series for new subscribers, two days apart."
• "What does my onboarding automation actually say in each email?"

Reporting and sending

• "How did my last newsletter do?"
• "How much revenue did email drive in the last 90 days?"
• "What's the story with sam@example.com?"
• "Schedule this to my subscribers audience for Tuesday at 9am."

Bigger asks

• "Create four drafts for next month's newsletter and write the first one."
• "Write a win-back email, build an audience of people with no activity in 60 days, and schedule it for Friday morning."