Work

The kind of systems I build

Four patterns I’ve built and run in production — from a straightforward lead pipeline to document processing at volume. Here’s what goes into each one, and what it replaces.

01 — Lead handling

From enquiry to first reply, with nobody touching it

Enquiries arrive from a website form, ads and DMs. Instead of landing in three different places and waiting for someone to notice, they get validated, logged and answered.

  • Every source feeds one pipeline
  • AI classifies the enquiry and drafts a first reply
  • Deal created in the CRM with the right source attached
  • Follow-up fires automatically if nobody responds

Replaces: manual inbox triage, spreadsheet lead lists, forgotten follow-ups.

Lead pipeline
Website formSource: organic
IN
Ad campaign leadSource: paid
IN
AI qualifies & drafts replyClaude API
AUTO
Deal createdPipedrive · stage “New”
AUTO
Follow-up after 48hIf no human reply
SCHEDULED
Order operations
New order placedShopify webhook
TRIGGER
Stock updatedInventory sheet
AUTO
Invoice generated & sentPDF + email
AUTO
Shipping notificationCustomer + warehouse
AUTO
Review request7 days after delivery
SCHEDULED

02 — E-commerce operations

One order triggers everything that has to happen next

An order shouldn’t need a person to push it along. Stock, invoicing, notifications and post-purchase follow-up all run from the same trigger.

  • Inventory updated the moment an order lands
  • Invoices created and sent without anyone opening a template
  • Customer and warehouse both notified automatically
  • Review requests timed to actual delivery, not to a guess

03 — Agency operations

Client onboarding that doesn’t eat a day

A signed contract kicks off the whole setup — folders, tasks, accesses, kickoff email — instead of a checklist someone works through by hand.

Proposal signed The only manual step
Workspace created Drive + project board
Tasks generated From your template
Welcome pack sent Email + intake form
Reports scheduled Every Monday, 08:00

Replaces: a 30-step onboarding checklist and Monday-morning report building.

04 — Document processing

The heavy end: hundreds of documents, no typing

Incoming invoices, contracts and supplier files in every format imaginable. AI reads them, the data gets validated against your records, and only the genuine exceptions reach a person.

Batch run · 340 documents
340 documents collectedEmail attachments, shared drive, upload form
IN
AI extracts the fieldsSupplier, number, date, line items, totals, VAT
AUTO
Checked against your recordsMatched to purchase orders and contracts
AUTO
11 exceptions routed to a humanTotals that don’t reconcile — everything else passes
REVIEW
329 posted automaticallyInto accounting, with the file attached
DONE

Replaces: a person typing invoice data into a system for two days every month.

Under the hood

Built on tools you can hire for

No lock-in to anything exotic. If you ever want someone else to maintain this, they’ll recognise every piece of it.

Make.comHubSpotPipedrive OpenAIClaude WebhooksREST APIsGoogle WorkspaceGoogle Sheets SlackTelegramGmailShopifyStripe AirtableNotionWordPressZapierTwilio

Your process is probably close to one of these

Tell me how it works today and I’ll show you what the automated version looks like — before you commit to anything.