One architecture creates another representation of the institution that must continuously remain aligned with it. DataFab resolves the authoritative estate in place and operates over that resolved context.

Acquire in place and from open sources · discover and catalogue · resolve with evidence · derive the model · extract the knowledge · encode it in the Studio · govern execution — then it is Monday.

Eleven resolve on evidence. The twelfth is similar, not the same — it stays in the queue with its reason. Never forced. Always reversible.

Thirty-one were handled by the platform itself — updated, re-screened, review date untouched. Six arrive as cards: what changed, where from, what is proposed, what happens if nobody acts. One is restricted.

Marcus sees the book. Tomasz sees Helios without its sanctioned parent. Sophie sees her client and one sentence. No count of what was withheld; the exclusion appears only in the authorised audit.

The cheapest investigation is the one the institution already performed.

Pushed before the case was opened: the proposed holder shares an address and a director with a restricted case. The trace runs over nodes that already existed.

Not a workflow somebody drew: a plan from the rules and their dependencies, re-planned when the world answers back. Every version kept. The case replays exactly as it ran.

Four hundred and twelve alerts resolve to ninety-six parties. Nothing is cleared on a name alone; nothing decided is decided twice; three the platform refused to clear because the evidence changed.

An event on one customer reaches its counterparties, their counterparties, and the other customer that shares them. Continuous risk should reduce reviews, not trigger them earlier.

1,471 on primary evidence already in the estate · 212 on evidence found on a related node · 729 specific requests on the regulator’s cadence · one refusal, to a person. £210 → £9 a file, illustrative — ask for their number.

The cost of a platform is not only the cost of running today’s rules. It is the cost of changing tomorrow’s.

The join paths the systems already share become the entity types and their relationships. Processes are generated from the model. When the next source arrives it proposes its own mapping — a connector and an approval, not a programme.

A domain expert builds in language; a platform engineer builds with every node. The same governed artefact. What they build is agentic: agents, agencies with a person at the gate, chains and graphs of agencies, and published utilities that run on the resolved estate the moment they exist.

Autonomy is a dial per step, recorded in a versioned register your leadership owns. Every agency runs sandboxed and tool-authorised, holds at the human gate, and writes a replayable audit — and what it builds cascades back into the Fabric with provenance, version and authority.

Four times today the platform knew something nobody entered: a path over a resolved graph, followed. Marked, confirmed, walled, replayable.

Decision record with six elements, evidence pack with every claim to source, factor trace reconciling to the arithmetic — on every one of the 2,041.

Time to value is not simply how quickly software can be installed. It is how much engineering, modelling, configuration and specialist translation has to happen before the first operational outcome. The implementation organisation is part of the product’s economics.

Software, cloud, integrator, professional services, the people who exist to keep the platform working, every regulatory change, every new source, every next utility. The marginal cost of the next outcome falls — it does not reset.

Six of thirty-seven events reach a person. Four judgements in a complex onboarding. Seventy percent of a cohort closed before a letter. Platform priced per client deployment, not per analyst.

What it produced is an asset for the second client, not a sunk cost trapped in the first: connector patterns, agency designs, the quality method — your method, in executable form.

Each bank lights up faster than the last — the fifth mobilises in the time the first took to sign — because the method carries and the data never does. You do not standardise the bank’s risk appetite; you standardise how your service operates around it.

A method change made Monday can be replayed across every client by Tuesday — each bank sees its own impact, follows its own approval route and releases its own version.

Seven programme lines a financial-crime platform usually requires — none of them ran on this Monday.