Diverse Landscapes - A Job Costing and Stock Platform Built Beside Fergus

Client Overview

Diverse Landscapes is a landscaping business based in Wainuiomata, running crews across several sites at once. Quoting and customer invoicing happen in Fergus, the job management software the office already knew well. The trouble was everything in between.

The Challenge

Fergus was being asked to do every job: hold the price books, receive the day sheets, carry the supplier costs and produce the numbers at the end. Most of that was done by people typing.

• Supplier price books were updated by hand, line by line, from supplier PDFs.
• Supplier invoices were checked against agreed prices by eye, when there was time.
• Day sheets were typed as free text into Fergus, with paperwork kept in the ute.
• Material bought in bulk was allocated to jobs afterwards by reading the day sheets.
• Each supplier invoice was keyed into Fergus, then keyed again into Xero.
• Back costing a job meant assembling the numbers manually, and quote follow-ups lived in heads and notebooks.

The obvious fix, building it all inside Fergus, was not possible. The Fergus API lets another system read nearly everything and write nearly nothing. There is no way to put a supplier invoice, a purchase order or a price book update into Fergus from outside.

The Approach

Fergus stays, and keeps doing the two things it does well: quotes are still built there from the business's own recipes, and customer invoices still go out from there. Everything between those two points moved into a new platform that sits beside it.

The platform reads jobs, quotes, price books, plans and photos from Fergus on a schedule and serves them from its own database, so no screen ever waits on Fergus. It writes back exactly one thing: charge-up time and material, pushed to the job as a draft quote for the office to review and send.

Under the Hood

It is a web app that saves to a phone or tablet home screen and behaves like an app, so it works on phones, iPads and office PCs from one build with no app store in the way. Behind it sits a serverless stack on AWS in Sydney: Lambda functions behind an authenticated API, DynamoDB, S3 for files and docket photos, SES for email, and Step Functions and scheduled jobs for the Fergus sync and the daily and weekly summaries.

Less of it is AI than the name suggests, and that was deliberate. Reading a supplier PDF and matching a description to a product code is AI, using Claude on Amazon Bedrock through an Australian endpoint so supplier documents stay in Australia. Everything else is conventional software, because job costing has to be predictable and auditable.

Built to be trusted with the numbers

• Costs stay with the people who need them: roles are enforced by the API, and site managers never receive costs, prices, margins or reports.
• Every change is on the record: each create, update, approval and allocation writes an audit event that is never edited or deleted.
• Stock that adds up: stock movements are only ever added, never overwritten, and balances are worked out from them. Yard stock and job stock are kept apart.
• A light touch on Fergus: every request to Fergus shares one rate budget, so the platform stays well inside its limits.
• Switched on a module at a time: each module sits behind its own switch, so the whole platform could be deployed at once and turned on as each phase was ready.

How it was delivered

The work was scoped up front in a written specification where every requirement has its own id, and business rules such as price rise thresholds and contingency live in settings rather than in code, so a change is configuration. Delivery was split into phases that each stand on their own: price books and the quote pipeline first, then supplier invoices, then orders, stock and day sheets, then reports. Every change was reviewed against the requirements it claimed before the next one started.