n8n emails a morning ops briefing with Claude
Faraz Ahmed runs hosting and SaaS tools for hundreds of clients, and he was still missing problems that sat in plain sight across dashboards. He built an n8n workflow that pulls server disk and backup state, analytics, subscriptions, CRM movement, inbox volume and refund notes every morning. Code computes the deltas and flags outliers before Claude sees anything. Claude then writes a three-sentence summary, labelled hypotheses and three actions for the day. The briefing lands on WhatsApp and email. In the first week it surfaced an unenforced quota and broken member access. He puts the marginal model cost under a dollar a month at his volume, with the host and APIs already paid for.
Copy this prompt
Build an n8n workflow that runs every morning. Pull [DISK / BACKUPS / ANALYTICS / SUBSCRIPTIONS / CRM / INBOX COUNTS]. In code, compute change versus yesterday, versus last week and versus a four-week average, and flag anything past my thresholds. Send only those prepared numbers to Claude with this brief: three-sentence summary, anomalies labelled as hypotheses, three actions for today, say when data is missing. Deliver the briefing to [WHATSAPP / EMAIL]. Never let the model do the arithmetic.
Our own prompt, written from what the source describes. We haven't run it ourselves. Text in [BRACKETS] is what you replace.
Watch out for
- Compute deltas in code first. Language models are weak at arithmetic you cannot audit.
- Add a data-quality gate. Bad input produces confident nonsense.
- Deliver into a channel you already open. A new dashboard is another thing to ignore.
Where this comes from
Original article by Faraz Ahmed on Blog, Nov 5, 2026. We rewrote it in our own words. The result is what the article reports.
Read the originalWant to do this yourself?
Our guide compares which agents can do this job, with steps and a prompt.
Turn a messy inbox into a 5-minute daily summary
Paste in today's emails and get back what needs a reply today, what can wait, what to ignore, and a one-line reply for each. It also flags emails that look like scams.