Standards that don't move for a deadline.
Our products carry people's money, tax position and client relationships. That raises the cost of being confidently wrong, so every business in the group is held to the same operating rules.
Six rules every product is built against.
Unknown beats incorrect
When a system doesn't have reliable information, it says so. It does not fill the gap with a plausible number. A blank field you can trust is worth more than a figure you can't. Unknown values are never quietly converted to zero.
One source of truth
Every important calculation lives in exactly one place. No screen, report or export gets to quietly disagree with another about the same number.
AI explains, systems calculate
We use AI to classify, extract, summarise and explain. It never becomes the source of a financial value, a tax figure or an eligibility result — those come from verified data and deterministic logic.
Show the working
Customers can see where a number came from, what fed into it, whether it is actual or estimated, and when it was last updated.
Start with the problem
Features exist because a customer is struggling, not because they would be interesting to build. If we can't name the customer and the workaround they are stuck with, we don't start.
Complexity is ours to carry
Simple is not simplistic. The complexity in these industries is real — our job is to absorb it inside the system rather than hand it to the customer.
How we look after customer information.
Our products hold financial records, client data and business performance information. We treat the handling of it as part of the product, not as an afterthought.
Access and isolation
- Each customer's data is isolated from every other customer's.
- Permissions are enforced on the server, not just hidden in the interface.
- Administrative actions are restricted and recorded in an audit log.
- Access follows the least a person needs to do their job.
Data handling
- Source documents and records are preserved, not overwritten without history.
- Important values carry their origin and the date they were last updated.
- Actual, estimated and projected figures are distinguished throughout.
- Where an external service fails, we say so rather than showing a substitute figure.
How we use AI — and where we stop.
AI is genuinely useful in our products. It reads documents, sorts information and explains complicated things in plain language. It is not allowed anywhere near the numbers themselves.
What AI does
- Classifies and sorts incoming documents.
- Extracts information for a person to review and correct.
- Summarises and explains what the data shows.
- Drafts content and suggests next steps.
What AI never does
- Determine a tax liability or a financial balance.
- Decide eligibility for anything.
- Produce an interest rate or a property value.
- Reach a compliance conclusion.
AI may explain the truth. It does not get to invent the truth.
What has to be true before we build something.
A small group can only do a few things well. Every significant initiative has to answer the same questions before it gets any of our attention.
- Is this a real problem, with evidence that it is real?
- Is the customer willing to pay to have it solved?
- Does it strengthen a business we already care about?
- Can it become an asset rather than ongoing manual work?
- Can we build it to the standard we have set?
- Is it the best use of our attention right now?