Key takeaways

The automation has four parts: a trigger, sources, a prompt, and a destination.
n8n and Make can both do it. Pick the one you already have access to.
I used Claude, n8n and Make to cut weekly research from four hours to one. This lesson shows the shape of that set-up. It doesn't copy my scenarios.
A person reads every brief and strikes any line with no source.

An automation is the written steps plus a machine that runs them on time. If the steps are unclear, the machine only makes the mess faster. That’s why I wrote the steps down first and only then built anything.

I used Claude, n8n and Make to cut weekly research from four hours to one. I haven’t published my exact scenarios, so this lesson gives the shape. Any version of it has four parts.

The four parts

A trigger that starts it every Monday morning.
Sources: a fixed list of pages or feeds, such as competitor blogs, pricing pages and news about your accounts.
A prompt: what to look for, and the format. One page, what changed since last week first, a source link on every line.
A destination where the team already looks: email, Slack or a shared doc.

In n8n

Add a Schedule Trigger set to weekly.
Add one HTTP Request or RSS node per source.
Add an AI node, such as the Anthropic (Claude) node, with your prompt and the fetched text.
Add an Email or Slack node that sends the result.

In Make

Start with a scenario scheduled weekly.
Add an HTTP module (or an RSS module) for each source.
Add an AI module, such as the Anthropic (Claude) app, with your prompt.
Add Gmail, Slack or a Google Docs module to send the result.

Check it before anyone acts on it

A person reads the brief each week and strikes every line with no source. It stays at one page, because my first version was too long to read. The source list gets a review once a month.

This week:

Write the four parts for one job of yours.
Build the trigger and one source in n8n or Make.
Run it once by hand and read the result.

Part of AI automation in marketing. Next: AI transformation for a social team.

Sources

DemandNexus case study: the tools used and the 4 hours to 1 figure.

Frequently asked questions

Should I use n8n or Make?

Use the one you can already get into. Both run on a schedule, fetch web pages or feeds, call an AI model and send the result to email, Slack or a doc. Neither is better for this job.

How long does it take to build?

I haven't timed building mine, so I won't give you a number. Budget a morning for a first version, and a few weeks of weekly use to get the brief right.

Three-minute lesson on this Build a weekly research brief Marketing course, advanced. Free, no sign-up.

More from the blog

All articles →