What this automation does
Through the week, the newsletter fills itself with ingredients. On issue day, an editor gets a nearly finished draft instead of a blank template.
- Collect. New posts from your blog RSS feed, changelog entries and links that teammates save (from a browser bookmarklet, a Slack shortcut or a Notion form) land in one newsletter inbox with the saver’s note.
- Fetch and clean. For each external link, the workflow fetches the page, extracts the article text and records the title, publisher and date. Anything it cannot fetch is marked for manual summary.
- Rank. On issue day, a model scores the inbox items against your newsletter’s stated purpose and picks the top items per section, explaining each choice in a sentence.
- Summarize. For each selected item, the model writes a two to four sentence summary from the fetched text, plus a line on why it matters to your readers.
- Assemble. Sections are slotted into your email template with links, the subject line and preview text options, and an intro paragraph drafted from the issue’s themes.
- Draft campaign. The workflow creates a draft campaign in your email platform, sends a test to the editor, and posts the preview link in Slack.
When it is worth building
This pays off when a newsletter is valuable to the business but its production keeps slipping, or when one person spends four or more hours per issue gathering links, writing blurbs and fighting the template. It works best for curated and roundup formats where much of the content already exists somewhere.
It is not the right tool for a personal essay newsletter where the writing is the product. In that case AI might help edit, but automating assembly saves little. And if your newsletter only reposts your latest blog post, an RSS-to-email feature, which many email platforms include, already does that.
Architecture
| Layer | What we use | Why |
|---|---|---|
| Inbox | Notion or Airtable database | Easy for anyone to drop a link and note; status fields drive the workflow |
| Sources | RSS for your blog, changelog API or sheet | New items arrive without anyone remembering to add them |
| Orchestration | n8n | Built-in Notion, Airtable and Slack nodes; HTTP Request for campaign creation in Mailchimp, Kit or Klaviyo |
| Article fetch | HTTP fetch with readability-style extraction | Summaries must come from the real text, not the headline |
| Ranking and summaries | Claude Sonnet 5.5 or OpenAI’s GPT models | Long enough context to read several articles and follow a style guide |
| Assembly | Your email platform’s template with merge regions | Keeps your design, unsubscribe footer and deliverability setup intact |
| Review | Test send plus Slack notification | Editor reads it in a real inbox before sending |
How to build it
- Define the format. Sections, items per section, summary length, tone, and what makes an item worth including. Write this down; it becomes the ranking and summary prompt.
- Build the inbox. Fields for URL, note, who saved it, section suggestion, fetched text, status and the issue it was used in.
- Automate collection. Add RSS and changelog sources, plus a quick save path from Slack or the browser so teammates actually contribute.
- Fetch and extract on save. Do the fetch when the link is saved, not on issue day, so failures surface early and articles that later go offline are still captured.
- Rank and summarize. Pass the format guide, the candidate items and past issues’ item list. Tell the model to skip anything already covered in the last few issues.
- Build the template mapping. Map sections to editable regions or merge fields in your email template, and render one test in the main email clients you care about.
- Create the draft and notify. Create the campaign as a draft, send a test to the editor, and mark used items in the inbox so they are not reused.
The downloadable starter template below covers the Notion inbox, article fetching, summarizing and a Mailchimp draft campaign.
What breaks in production
- Summaries of the wrong article. Cookie walls, login pages and redirects return text that is not the article. Check that the fetched text is long enough and mentions the title before summarizing.
- Broken links at send time. Links die between saving and sending. Re-check every link on issue day and flag failures in the draft.
- Template rendering. Generated text with stray characters or very long URLs can break layouts in some email clients. Sanitize text and shorten display URLs.
- Repeated stories. Two teammates save the same news from different publishers. Deduplicate by topic, not only by URL.
- Misleading open rates. Some mail clients preload images, which inflates opens. Judge issues by clicks and replies more than opens.
Timeline and cost
A newsletter pipeline with one inbox, RSS sources, AI summaries and a draft campaign in one email platform usually takes one to three weeks. Lumeor builds this in the range of $4,000 to $9,000, depending on the number of sources, sections and segment versions, and how complex your email template is. Model usage per issue is small. If your long-form content also needs social posts, the content repurposing pipeline shares most of the same building blocks. For triggered emails to new customers rather than a broadcast, see customer onboarding emails.
How to measure success
- Editor hours per issue, before and after
- Issues sent on schedule versus planned
- Share of AI summaries published with no factual correction
- Click-through rate per item, which also tells the ranking step what readers want
- Unsubscribes and replies per issue