Skip to content
Okko Oulasvirta
Financial Services

AML Investigation Workspace

Give investigators the transaction network, the customer history and the prior cases in one place, so an alert becomes a defensible decision instead of a week of screen-scraping.

The decision it supports

Is this alert suspicious enough to escalate, and what is the evidence either way?

What changes

Faster alert clearance, fewer genuine cases lost in the backlog, and a case file that stands up to a regulator. The investigator decides to file or dismiss — the workspace assembles the evidence.

Data domains

  • Customers
  • Transactions and payments
  • Compliance and financial crime
  • Risk and exposure
  • Channels and interactions

Interactive preview

AML Investigation Workspace is described here as an opportunity rather than shown as a working demonstration. The interactive build for this industry is Customer & Risk Decision Workspace — it uses the same primitives, the same governed data model and the same decision-to-action pattern this application would.

We build one flagship per industry properly rather than four shallow ones. The pattern transfers.

The workflow

  1. 01An alert arrives from monitoring or screening
  2. 02The customer, their related parties and their transaction network are assembled
  3. 03Behaviour is compared against the customer's own history and expected pattern
  4. 04Similar prior cases and how they were resolved are shown
  5. 05The evidence is drafted into a narrative with every fact cited
  6. 06The investigator confirms, adds to or rejects each point and makes the call
  7. 07A suspicious activity report is prepared or the alert is closed with a reason — both recorded

Platform capabilities most relevant here

  • OneLake

    One governed copy of enterprise data that analytics, AI and applications can all build on.

  • Curated business entities

    Agree once what an order, an asset or a customer means — then reuse it everywhere.

  • Graph relationships

    Answer questions that depend on how things connect, not just what they contain.

  • AI functions

    Apply language models to business data as a normal part of a data pipeline.

  • Fabric Apps

    Build the application where the governed data already lives, instead of copying it out.

  • Application state

    Remember decisions, assignments and approvals — the things a report cannot hold.

  • Workflows and approvals

    Keep accountability visible when a decision needs a second pair of eyes.

Not every application needs every capability. This is the subset that this decision actually depends on.

Talk about building this