A construction business
Client
From paper and spreadsheets to a live view of every build
The problem
The numbers lived in Xero, spreadsheets, email threads and paper files, and none of it was connected. Every time someone needed to know whether a job was on budget, or how many labour hours it had taken, they waited for the numbers to be pulled together.
What I did
Brought it all into one place, then built on top of it.
- Moved Xero, the spreadsheets, the emails and the paper records into one central, cloud-based database, structured so reporting and AI can be built straight off it.
- Built a platform on that database that pulls up from anywhere and answers the questions that used to wait: “are we tracking on budget?”, “how many labour hours did this take?”, and a brief on every build.
- Budget against actual, materials and supplies, and timesheet analysis, all running off the same data.
What changed
They stopped waiting for numbers. There is now a live view of every build, served from one database instead of spreadsheets and paper files, and it’s in the cloud, so it’s there wherever they are.
T&G Global · Horticulture
In-house
Finance transformation across the apple business
- Python
- SQL
- Power BI
- Snowflake
- SAP
- AI agents
- Governance
The problem
Month-end close ran for days on manual consolidation. Operating units couldn’t see their own spend or budget without asking finance and waiting. Profitability and planning lived in separate models that had never been joined, so results never fed forward into what came next.
What I did
Defined the programme, scoped it, and built it.
- Automated the month-end close, taking it from days to hours.
- Combined development modelling with orchard profitability, so what a block actually returns now feeds the decisions made about it.
- Built an accommodation allocation tool for seasonal workers that sorts against preset rules, so placements are made in minutes and stay inside the accreditation tiers employers are held to.
- Wrote the governance process that decides which projects go through the PMO, and on what criteria.
- Built AI agents that let operating units ask their own questions about spend, budget and performance from a phone or a laptop.
What changed
Close went from days to hours and recurring reporting from hours to seconds. Operations stopped queuing behind finance for answers. Planning now runs off the same numbers as the results.
Retail & hospitality
Own product
Turning transaction data into operator decisions
- TypeScript
- React
- PostgreSQL
- Python
The problem
Most point-of-sale systems record what happened and stop there. Operators end up making ranging, pricing and floor decisions on instinct, sitting on top of data that could answer the question.
What I did
- Software that reads the layer underneath the transaction: basket affinity — what sells together and what never does; table and cover profitability for restaurants; repeat ordering patterns; where the margin actually sits rather than where it’s assumed to be.