AI × Supply Chain build breakdowns

The build library

The builds behind the products, written up with the actual code, a leverage rating out of 10, and where each one failed.

  1. oac's brain, part 1: the data layer

    Every agent I run sits on the same data layer: one Google Sheet workbook as the single source of truth, fed by scheduled Python syncs with health checks. Here's how I wired Shopify, Klaviyo, subscriptions, inventory and finance into it without paying for five APIs, and what failed first.

    Read the build →

    data foundationsautomationagents

    Leverage 6/10
  2. Free replacements were quietly wrecking my numbers

    A lost parcel means a free replacement. Process it wrong and it lies to every system: pads orders, tanks AOV, tells your demand model you sold what you gave away. Here's the tagging standard and the skill that does it right.

    Read the build →

    customer opsfinanceautomation

    Leverage 3/10
  3. My daily briefing, and how it's changed three times

    Every morning an agent writes me a briefing before I'm at my desk: what needs action, what's slipping, what to prep. It started as a glorified summary. Here's how it became something I actually run my day on.

    Read the build →

    reportingautomationagents

    Leverage 4/10
  4. The scenario-planning agent that couldn't live in the grey

    I tried to teach Claude the judgement call every planner makes: do I let one flavour dip below minimum for two weeks, given what's landing next month and a retail deal that might close? It couldn't do it. Here's how far I got.

    Read the build →

    supply planningagents

    Leverage 4/10
  5. Three inboxes, one source of truth: a dispatch tracker for a 4-person team

    Dispatch issues, delivery problems and carrier complaints were landing across three different inboxes. Built in a couple of hours: claude scans the inboxes twice a day, classifies, posts a single summary into slack, archives the structured data into a sheet. Built once, compounds daily.

    Read the build →

    automationcustomer opsreporting

    Leverage 3/10
  6. SOPs nobody had time to write, written in minutes

    In a small team the 'process' is whatever's in one person's head. I use AI to turn how I actually do a task into a clean SOP the team can follow, in minutes, instead of never.

    Read the build →

    teamautomation

    Leverage 3/10
  7. An afternoon of invoice reconciliation, down to minutes

    Every month the wholesale invoices in Xero have to match the orders in Shopify. By hand it's an afternoon squinting at two exports. This skill finds the files itself and hands back only the mismatches that actually matter.

    Read the build →

    financeautomation

    Leverage 5/10
  8. The month-end stock count, rebuilt from the data

    My accountant needs a valued stock count every month: product, packaging, merch, free stock. It used to be CSV exports and manual valuation. Now it's built straight from the databases, with the double-count traps designed out.

    Read the build →

    inventoryfinancereporting

    Leverage 5/10
  9. The pouch is the constraint nobody plans for

    Most plans stop at finished product. But you can't run a batch without the pouch, the film, the box, each with its own lead time and minimum. I put packaging in the plan and it immediately found the constraint nobody was watching.

    Read the build →

    supply planningpackaginginventory

    Leverage 4/10
  10. Perpetual inventory on a Shopify budget

    A small brand can't run an ERP for perpetual inventory, and 'just check Shopify' isn't good enough. Here's how I used a cheap inventory app as the perpetual layer, and the column that quietly breaks everyone who tries it.

    Read the build →

    inventorydata foundations

    Leverage 3/10