Work

Three systems, in production.

Each system below was scoped, built and deployed by Range AI and is in daily use. Client names stay private; the detail of what was built is real. With the client's consent, a reference conversation can be arranged on request.

Delivery dockets, read and filed into the warehouse system.

The job

Suppliers and carriers send delivery and dispatch dockets by email, as PDFs, scans and Word files. Each docket is keyed into the warehouse management system: owner, direction, product code, batch, quantity, customer, reference and transport. Staff keyed every one by hand.

What we built

  • A dedicated mailbox that collects every docket, converts Word files and straightens rotated or upside-down scans.
  • An SI transcription step that copies down everything the docket prints.
  • A matching step that checks customers, products and freight terms against the client's own lookup tables and accepts exact matches only. Anything uncertain is raised as a flagged suggestion.
  • A review screen behind the company's single sign-on, showing the docket beside the completed form. The reviewer fixes any flag and presses Accept. Outgoing dockets that pass every check are accepted automatically, with stock and duplicate-number safeguards.
  • Secure filing into the warehouse system, followed by a check that the docket has appeared there, so every filing is confirmed.
  • A learning loop: reviewer corrections become suggested standing rules, and each accepted rule feeds straight back into the SI.

The result

  • 97.8%of fields correct before human review
  • 500–600dockets filed a month
  • About 30cSI cost per docket

Accuracy measured across 2,566 fields on an audited set of 922 historical docket emails. Volume from live filings, August to September 2026.

Next phase, in prototype: a 3D digital twin of the warehouse covering six buildings and 2,564 pallet positions, with iPad putaway screens and barcoded pallet labels, fed directly by the docket pipeline.

A private SI workspace for the whole agency.

The job

The agency runs social content, campaigns, client reporting and proposals for many brands at once. Time reports were assembled by hand in Excel each week, meeting follow-up was written up manually, and the team's SI history sat across individual accounts.

What we built

  • A private SI workspace for the team on a dedicated server, behind one secure login. Every past SI conversation was imported and sorted into a workspace for each client, so the history sits with the right brand.
  • Weekly and monthly time reports built automatically from live Everhour data: hours per client against budget, billable targets per person, and a flag on any project running over. The weekly report reaches the agency's leadership early every Tuesday.
  • An Hours page inside the workspace with live charts, custom date ranges, a people view and the billing pipeline by financial year.
  • Meeting recordings turned into draft packs: minutes, content plan, shoot brief, next agenda and campaign plan as Word documents, sent to the staff who attended and filed in the client's folder.
  • A client content portal: a staff calendar with true-to-platform post previews, and a private review page where each client approves, comments or asks for changes.
  • Image, video and voice generation from the chat box, with a shared media gallery.
  • A library of agency skills, from brand voice to campaign plans, client reports, proposals and meeting minutes, kept in sync with the team's ChatGPT and Claude.

The result

  • 1,266past SI conversations moved into 49 client workspaces
  • 900 of 900report cells matched the hand-built version on first audit
  • 23agency skills in the shared library

Figures from the rollout, August to September 2026.

Supplier paperwork, worked from the inbox into Xero, ServiceM8 and Dropbox.

The job

Supplier bills, credit notes, statements, remittances and customer purchase orders arrive in a shared accounts mailbox. Each one is entered into Xero, coded and tagged to the right job, added to the job's costs in ServiceM8 and filed in the job's Dropbox folder.

What we built

  • An agent that watches the accounts mailbox and reads every attachment, including emails that carry several documents at once.
  • Classification into supplier bills, credits, purchase orders, receipts, statements and remittances, each with its own workflow: 18 in total.
  • A plan for every email: find or create the supplier in Xero, check for duplicates, find the ServiceM8 job by number or address, then apply the job's tracking tag and account-coding rules drawn from 621 past bill lines.
  • A review screen where the office sees every planned action and approves it with one click, or replies in plain English. The agent revises the plan and keeps the correction as a standing rule.
  • Execution across all three systems: the draft Xero bill with the PDF attached, credit allocations and payments, the cost line on the ServiceM8 job, the PDF filed in the right Dropbox folder, and the email moved to Processed.
  • A ServiceM8 connector so the owner's own Claude can look up jobs and clients directly.

The result

Every supplier document reaches Xero coded, attached to the right job and filed, ready for the office to approve.

  • 18document workflows
  • 621past bill lines behind the coding rules
  • 1 clickto approve each plan

Each system proposes, a person decides, and every entry is recorded. They run on the client's own accounts and connect to the software the team already uses. Your first system can follow the same pattern.

Contact

Talk to Tom about your first system.

Tell him what your business runs on and which work takes the most time. He will tell you plainly whether a system like these fits.

[email protected] 0468 920 484