Services
The manual work between your data and your decisions
Most of it is one of four things. Which one you need usually becomes obvious within half an hour of talking.
Automation
Stop doing it by hand
Anything that happens on a cycle and eats a person’s week. Month-end close, reconciliations, accruals and journals, the report someone rebuilds every Monday morning, invoice coding, variance commentary drafted from the numbers instead of typed from scratch.
Time
Days of the close returned to the team every month
Risk
Fewer manual touches, fewer errors, a real audit trail
Speed
Results land while they can still change a decision
AI agents
Let people ask for themselves
An agent that knows your data, your structure and your definitions, so an operations manager can ask what they spent last month from a phone and get a straight answer — instead of emailing finance and waiting three days. Not a chatbot bolted on the side: built on your own numbers, with the boundaries and permissions your business actually needs.
Time
Finance stops being a help desk for its own numbers
Reach
Everyone who owns a budget can see it, wherever they are
Adoption
People use it because asking is easier than a dashboard
Advanced modelling
Know where the money goes
Unit economics by product, customer, site or job. Scenario and sensitivity modelling. Forecasting that updates off actuals rather than being rebuilt each quarter. Separate models joined into one, so this month’s results feed straight into what you’re planning next.
This is the half most technical contractors can’t do. Building the model needs someone who can read the accounts — how overhead gets allocated, what margin should even mean for your business, where the accounting and the operational truth diverge.
Data & reporting
One version of the truth
Getting numbers out of the systems they’re trapped in, matched across sources that were never designed to talk, and into reporting people trust enough to act on. Including the unglamorous part: deciding what a number means before five teams each define it differently.
- Python
- SQL
- Power BI
- Snowflake
- SAP
- TypeScript
- PostgreSQL
Governance
Built to survive an audit
The hesitation about handing finance systems to someone outside the business is rarely whether the thing will work. It’s what happens to the data, and whether anyone can explain the system once the person who built it has gone. Both are answerable.
Your data
- Work happens inside your environment. Your tenant, your systems, your region — nothing extracted onto a personal machine.
- Least-privilege access, scoped to the job and revoked at handover. Read-only wherever the work allows it.
- Your data is never used for anything beyond your engagement. Not for training, not for anyone else’s build.
- As a Chartered Accountant I’m bound by the CA ANZ code of ethics and its confidentiality obligations, with a professional body behind it. NDA signed before anything changes hands.
Your systems
- An audit trail on everything automated, because your auditors will ask how it works and someone has to answer.
- Segregation of duties preserved. Automation has a habit of quietly collapsing an approval step that existed for a reason.
- Role-based access on the agents, so people see their own numbers rather than everyone’s.
- Documented lineage: where a number comes from, and who owns its definition.
- Handover documentation written for whoever inherits it, so you aren’t dependent on me being reachable.
Most people building automation learn these constraints when something breaks. I started inside them.
Ways to work together
Three ways in
Discovery
One to two weeks. I map the process, find what’s automatable, and come back with options and a recommendation. You keep the findings either way.
Build
Four to twelve weeks depending on scope. Built alongside your IT and data teams and handed over documented and tested.
Training
A programme for finance teams who want to build this themselves — from working in an IDE through to running their own agents. Interactive, on your own data.
Not sure which one you need?
Most people aren’t. Describe the problem and I’ll tell you what it actually is — including if the answer is that it isn’t worth fixing.