Agents and automations running in production.
Real systems our team built, across support, finance, production, e-commerce, product content and reporting. Each case starts with the work a team was doing by hand, then shows what we built, how it runs and where a person stays in control. Client names stay out because of NDAs.
Case studies
Customer support
An AI support desk where a person still decides what gets sent
Drafts for every ticket across email, WhatsApp and social comments, a one-click review step, and a narrow lane where short social replies post on their own inside hard limits.
- n8n
- Gorgias
- Supabase
- Slack
Finance
A finance back office that reads its own post and supplier mail
Scanned letters and supplier emails read by AI and filed into Odoo as drafts, marketplace credit notes booked on schedule, contractor approvals carried to payroll, and an alert the hour a bank detail changes.
- n8n
- Odoo
- Claude
- Google Drive
- Slack
Operations
Production planning that is done before the team arrives
Reorder points turned into linked purchase and work orders, operators assigned at 06:00, packaging and batch records checked against the warehouse every morning, and suppliers chased until they reply.
- n8n
- Manufacturing ERP
- 3PL warehouse
- Odoo
- OpenAI
- Slack
E-commerce operations
Shop, marketplace and warehouse kept in step without anyone watching
Every Shopify stock change pushed to a marketplace, fragile orders moved to a safer carrier, return labels created from the help desk, and wholesale accounts welcomed after approval.
- n8n
- Shopify
- Gorgias
- Marketplace API
- 3PL warehouse
- Slack
Product and content
Shopify product pages built from one approved record
An approved product record becomes a complete Shopify product with SKUs, tax, ingredients, warnings, shelf life, SEO and metafields, one product at a time, with a reset and an alert when anything fails.
- n8n
- Shopify
- OpenAI
- Postgres
- Slack
Leadership and engineering
Reports that draft themselves and alarms that fire before things break
The weekly all-hands slides drafted by AI from the week's numbers, ad performance pulled into the data warehouse every morning, and every AI and API key tested every 15 minutes.
- n8n
- Claude
- Google Slides
- Meta Ads
- Slack
How we write these
Clients give us access to their help desk, their inbox and their customer data. These pages keep that private.
No names or logos
Most clients sign an NDA. We describe the type of business and the team, and leave out anything that points to who they are.
Diagrams with sample data
Every diagram is redrawn. No screenshots of a client's help desk, inbox or workflow canvas.
Numbers only when measured
A figure goes on a case once it comes from at least 30 days of real runs. Until then, we describe what the system does.
Have work your team still does by hand?
Tell us where the hours go. We'll tell you what an agent or an automation can take over, and what should stay with your team.
Book a 15-min call