Working with Penpal, Letterhead's AI assistant

Penpal is Letterhead’s assistant — a panel that sits beside whatever you’re working on, in every account. This article covers how Penpal works in general (opening it, starting over, chat history, handing off to a person) and everything it can do with a newsletter draft in conversation. For what Penpal can do in other parts of Letterhead — promotions, transactional email performance, or (from a connected AI client like Claude Desktop) goals, reports, and automations — see Related below.

You tell Penpal what you want in plain language — “shorten the subject line”, “make this intro punchier”, “swap in this story” — and it changes the newsletter for you. When it touches a draft, a preview of the finished email opens next to the conversation and re-renders as the work happens, so you can see the result without leaving the chat or reloading anything.

Open the assistant

Select the speech-bubble button in the top bar. The assistant slides out on the right; the page you were on stays visible beside it, so you can keep working while the conversation is open. On a narrow screen it opens as a full-width panel instead.

The button is in the top bar on every page, so you can reach the assistant from your newsletter list, the composer, your reports, or your settings — wherever the question comes up. Select it again to close the panel; your conversation is still there when you come back.

On a very wide screen, Penpal may already be open when you arrive — it’s meant to be noticed rather than hunted for. That only happens when there’s genuinely room for it beside your work: a large external monitor, roughly 1,500 pixels across or wider, rather than a typical laptop screen. Everywhere else the panel waits until you open it, so a smaller display isn’t given over to a conversation you didn’t ask for. If you’d rather it stayed out of the way on the wide screen too, close it once and it won’t open itself again, on this visit or any later one. The speech-bubble button still opens it whenever you want it, and closing it never loses your conversation.

That choice is remembered across sessions, so closing the panel keeps it closed when you return to Letterhead later.

Ask for a change

Type what you want in the message box and send it. The assistant works with your real newsletters, so you can refer to them the way you’d say it out loud, and it knows which page you’re on — “this draft” is enough when you’re looking at one.

Things it can do with a newsletter:

• Read a draft back to you — its title, subject line, preview text, status (draft, scheduled, published), creation and last-modified times, and what’s in it.
• Change the subject line or preview text — including asking for a few options to choose from. This works on any newsletter.
• Rewrite the words of a draft it wrote — tighten the introduction, change the tone of a section, cut a paragraph down. See the note below on which drafts this covers.
• Pull in content — add a story you point it at, or something from your Curations library.
• Answer questions about how a send performed — opens, clicks, and the numbers behind your last newsletter.

Which drafts the assistant can rewrite

If you asked the assistant to draft a newsletter, it owns that draft’s wording and can keep editing it for you. A newsletter you built yourself in the composer is made of blocks, and its body is edited there — so if you ask the assistant to rewrite a section of one, it will tell you that and leave the body alone. Its subject line and preview text are still yours to change by asking.

Some example asks:

• “Make the opening two sentences punchier.”
• “Give me three subject lines for this issue.”
• “Shorten the second section — it’s twice as long as the others.”
• “Add the piece about the new library branch after the events roundup.”
• “How did last Thursday’s newsletter do?”

The assistant only changes something when you ask it to. It won’t edit or send a newsletter on its own.

Say yes with one click

Sometimes the assistant ends its turn by offering to do one specific thing — “Want me to write a subject line from that draft?” When it does, a confirmation button appears beneath that message. Click it and you’ve approved the work, without typing anything.

The button sends your yes as an ordinary message from you, phrased to fit the question it’s answering, so the conversation reads as a conversation rather than a run of one-word confirmations. Nothing else about it is special — you can ignore the button and type whatever you like instead, including a question of your own, and the assistant still only acts once you’ve said so.

The button is only ever on the assistant’s latest message, and only while its offer is still open. Answer it — by clicking or by typing — and it goes, so there’s nothing to click by accident on a question the conversation has already moved past.

Ask it to draft a whole newsletter

You can also start from nothing: describe the newsletter you want in conversational terms — “Draft a newsletter about this week’s tech policy updates” or “Create an issue with our top curated stories about sustainable design.” This works from the in-app panel and from a connected AI client such as Claude Desktop.

To build the draft, the assistant will:

• Pull relevant content from your curations
• Source fresh articles from the web when needed
• Write the newsletter body in your channel’s established voice, derived from your top-performing letters
• Render it into email-safe HTML with proper formatting
• Append a compliance footer with the correct unsubscribe link and mailing address
• Save it as a draft in your Letterhead account

Drafts arrive in your channel’s visual identity

A drafted newsletter is rendered in your channel’s brand colors, fonts, and logo, sourced from your channel’s settings — so it looks like your actual newsletter from the moment it’s created. The response also includes a direct editor link that opens the draft straight in the composer for review.

If a draft isn’t what you wanted, ask the assistant to delete it — “Delete that draft” — and request a new one, without manually removing the previous one first.

The assistant never schedules or sends anything: every draft waits for you to review it. When you’re happy with one, open it in the composer to make final edits, select your audience, choose a publication date, and send.

Ask it to take you somewhere

As well as answering, the assistant can move you around the app. Ask for a screen and it opens it, instead of telling you which menu to look under — “take me to my drafts”, “open the audience page”, “show me the analytics for the morning briefing”. It tells you where it’s taking you, then takes you there, and the panel stays open beside the new screen so you can keep talking.

It can reach the screens you’d normally click to: your newsletters, templates, automations, and curations; your audience; your analytics, reports, and goals; your promotions; your brand kits and image gallery; your settings, integrations, and team; your notifications and your own account. Name a newsletter and it opens that newsletter’s own editions, automations, audience, analytics, reports, or settings.

If you work in more than one Letterhead workspace, you can ask to move between them — “switch me over to Riverside” — using whatever you normally call it. A workspace is a separate place to work, so this one reloads the page, and you’ll start a fresh conversation on the other side.

It can only take you where you can already go. Ask for a newsletter or a workspace you don’t have access to and it says so instead of moving you — the same permissions apply as when you navigate yourself. If it can’t find what you named, it’ll ask you to name it the way it appears in your menus.

Watch the draft update beside the conversation

As soon as the assistant works on a draft, a Live preview pane opens to the left of the conversation showing the newsletter as a reader would receive it, with the draft’s title above it. Each time the assistant makes another change, the preview re-renders — so you can ask for a rewrite, read it in place, and ask for another pass without opening the composer.

Close the preview with the × in its header when you want the room back. The draft itself keeps every change the assistant made; closing the preview only hides the view.

The preview appears on desktop-width screens, where there’s space for it beside the conversation.

If the newsletter is open