Technology
AI used the way it should be used.
On the work, not on the judgement. We operate a regulated business on it ourselves, and execute with it for customers who carry obligations of their own. The machine carries most of a matter end to end: the gathering, the chasing, the reading, the preparation. People handle the exceptions, and people make the call.
Most attempts stop at the pilot
Regulated work is where applied AI usually goes to die. The demo convinces, the pilot runs, and then someone asks which requirement the output satisfied and on what evidence, and there is no answer that survives the question.
Ours is in production, inside a live regulatory regime, on commercial revenue, with a record behind every matter built to be examined. That makes iDeed a working example of AI applied to the labour, in a domain that resists it, with accountability intact.
Guidance is derived, not generated
Point a general-purpose model at a regulated file and you get an answer with no standing.
01
The rules define the required state
The regime, codified as deterministic logic. What must be true for a matter, and why.
02
The engine observes the actual state
Models read the documents, registries and responses. That is what they are good at.
03
The gap is the next action
We can always show what was required, what was seen, and what remains. Nothing rests on a model asserting something.
The operator acts on the gap and owns the decision. Some judgements, establishing who really controls an entity and making the final determination, stay with a person permanently.
What the capability consists of
Five assets. Every case worked makes each one better.
The contract engine
Runs any contract that has a legal structure: parties, obligations, states. A new contract type is configuration, not a rebuild.
The codified framework
The regulatory regime turned into deterministic rules, kept current as guidance changes.
The resolved case corpus
Every hard case we have worked: what made it hard and how it was decided, held as structured data rather than in someone’s head.
The guidance engine
For each matter it works out the options in real time: which rule applies, what evidence is missing, and how cases like it were resolved before.
The operating method
How the work is run to one standard, written down so it moves to a new market with us.
Built this way from the start
The difficulty for established providers is not that they lack AI. It is that their architecture predates it, which leaves two bad options: attach something to the side, or rewrite underneath a live install base.
Not tied to a model generation
The regulatory logic is deterministic and lives in its own layer. The model occupies a defined role rather than being threaded through the product, so what sits behind it can be replaced without touching the business logic. A platform weaving a language model through its features today is accruing exactly the debt this avoids.
One architecture, two regimes
The same structure carries multiparty financial instruments and a regulatory regime. Not a claim about flexibility: something already done, which makes the cost of a third known rather than estimated.
Designed in, not retrofitted
Audit trail and version history as properties of the system. Access control and delegated approval enforced by the platform, not by procedure. All expensive to add later, cheap to design in.
Secure by design
The data these processes handle is among the most sensitive an organisation holds. The platform runs on serverless cloud infrastructure with role-based access, delegated approvals, single sign-on and detailed logging, so governance is enforced by the system rather than by procedure.
Residency by design
Intellectual property, client data and processing sit in Australian cloud regions under Australian law, with customer-managed keys held here. Client data is never exposed to public models, and nothing moves offshore without a documented legal basis.
Residency is tagged on every record at creation against the client’s own jurisdiction. A market that requires data to remain within its borders is therefore a region we add, not an architecture we rebuild, which is the difference between entering that market and declining it.