Reports that draft themselves and alarms that fire before things break
The team runs a weekly all-hands, reports on its ad spend, and depends on dozens of automations that use AI and data API keys. We built the n8n workflows that draft the Monday slides, keep ad data in the warehouse and warn the team before a broken key breaks everything that uses it.
Three schedules, each with one job and one place where people see the result.
- Team
- Leadership and engineering
- Built with
- n8n
- Claude
- Google Slides
- Meta Ads
- Slack
The problem
Mondays started with copy and paste.
Before every all-hands, someone collected the week's sales, metrics and team updates from several tools and built the slides by hand.
Ad performance lived in the ad platform. Any report that needed it started with an export.
Many automations shared the same AI and API keys. When one expired or ran out of credit, several workflows failed at once, and the first sign was a pile of confusing errors.
What we built
All-hands slides drafted on Monday morning
Every Monday at 06:00, the workflow gathers the week's sales, metrics and team updates, and Claude writes them into the all-hands slide deck. A Slack message says the draft is ready for review.
Ad performance in the warehouse every morning
Early every morning, ad performance reports are pulled into the data warehouse, each one starting where its last run stopped. Reports read from one place instead of from exports.
Every AI key tested every 15 minutes
Every 15 minutes, a workflow tests 34 AI and data API keys. A failed key raises an alert in the infrastructure channel before the automations that use it start to fail, a reminder every 6 hours while it stays broken, and a message when it works again.
Guardrails
How it stays trustworthy
Reports and alarms only help if people believe them.
- The slides are a draft that the team reviews before the meeting
- A dry-run switch tests changes without touching the real deck
- Each ad report resumes from where it stopped, so a failed run is caught up
- Failed ad syncs are logged, not skipped
- Key checks alert before dependent automations fail
Why we built it this way
Draft, not deck
Claude writes the first version. The person presenting still owns the story.
Resume, not reload
Each ad report remembers where it stopped, so a failed run is caught up on the next one instead of reloaded from scratch.
Test what everything depends on
One expired key can break many workflows at once. Checking the keys directly turns a pile of confusing failures into one clear alert.
Measured
What the team can see
Each schedule reports in the channel that owns it.
- A Slack message when the weekly draft is ready
- Ad sync failures, logged per run
- Which key failed, when, and when it worked again
More case studies
Other systems our team built.
- Customer supportAn AI support desk where a person still decides what gets sentRead the case →
- FinanceA finance back office that reads its own post and supplier mailRead the case →
- OperationsProduction planning that is done before the team arrivesRead the case →
- E-commerce operationsShop, marketplace and warehouse kept in step without anyone watchingRead the case →
- Product and contentShopify product pages built from one approved recordRead the case →
Related
Who builds your Monday numbers?
Tell us where your numbers live. We'll show you what can be drafted before your team logs on.
Book a 15-min call