Engineering AI & Automation That Ships
The build practices behind automation that survives contact with production.
A working demo and a system that runs every morning, unattended, are different engineering problems. The second needs dedup, retries, monitoring, and a way to fail loudly instead of silently.
We build automations as real software - tested, documented, and handed over with the source - so they can be maintained and extended rather than rebuilt.
This cluster is the technical side: how we structure builds, handle edge cases, and leave behind systems a team can own.
In this guide
15 posts- Automating sensitive data without creating a compliance problem
- Google Apps Script: the most underrated automation tool
- The security review every AI integration needs before launch
- Monitoring AI systems like you monitor infrastructure
- CRM integrations break in the same five places
- Evals are the part nobody wants to build
- What retrieval-augmented generation actually gets wrong
- Monitoring automation you can't watch manually
- Idempotency: the boring fix that prevents the worst automation bugs
- Cost control for AI pipelines before the bill surprises you
- Before you build a CRM integration, answer these five questions
- The real cost of a broken integration
- 128k context windows: what actually changes when the model can read more
- Prompt injection is the SQL injection of this generation, and most teams aren't ready
- What we learned putting frontier models in production this year
Related recipes
2 recipesReconcile Salesforce records against an external source nightly
Salesforce and a second system (billing, ERP, a partner feed) drift apart over time, and reconciling them is a manual export-and-compare job.
Scheduled job runs nightly → Salesforce records compared field-by-field against the external source → Mismatches logged with the specific field and record → Summary sent for review
Export Stripe invoices to accounting nightly
Invoices generated in Stripe get re-entered into the accounting system by hand at the end of each month, a backlog that grows all month.
Scheduled job runs nightly → New Stripe invoices pulled since the last run → Invoices written into the accounting system
Where it applies
For your team
Common in
Continues · Want this built
Want this built, not just read about?
Tell us the workflow. We'll scope it and build it as production code you own.
We typically respond within 24 hours