A risk case tracks how your business alerts, investigates, and acts on
suspicious or risky activity related to AML. A single risk case may consist of
many events and may include multiple parties, but each party must have its own
investigation process tracked through a risk case. In particular, each party in
a risk case should have distinct entries for events listed in the following
section.

## Event types

To express these events, AML AI uses the following types in the
[RiskCaseEvent](https://docs.cloud.google.com/financial-services/anti-money-laundering/docs/reference/schemas/aml-input-data-model.md.txt#riskcaseevent) table.

> [!IMPORTANT]
> **Important:** Only include risk cases and events relevant to money laundering risk in the risk case events table. Exclude risk cases related to other risks.

Initial alerts trigger the following event types:

- `AML_ALERT_GOOGLE`: triggered using Google AML [risk score](https://docs.cloud.google.com/financial-services/anti-money-laundering/docs/concepts/glossary.md.txt#risk-score)
- `AML_ALERT_LEGACY`: triggered by another automatic risk assessment system
- `AML_ALERT_ADHOC`: manually triggered, for example, by the relationship manager
- `AML_ALERT_EXPLORATORY`: triggered based on considerations other than pure risk, for example, data exploration to improve the quality of AML models
- `AML_ALERT_EXTERNAL`: triggered based on an external request

Events for the start and the end of an investigation process trigger the
following event types:

- `AML_PROCESS_START`: investigation process starts
- `AML_PROCESS_END`: The investigation process is complete and no further outcomes are expected. If there is no `AML_SAR` or `AML_EXIT` by this point, then AML AI safely concludes there will be none.

Actions taken based on the investigation trigger the following event types:

- `AML_SAR`: suspicious activity report (SAR) is filed against the party for suspicious behavior
- `AML_EXIT`: A customer exit is triggered. This event must have a corresponding `AML_PROCESS_START` event for the same party and risk case ID.

Optionally, if the investigation identifies a period of suspicious activity, you
may add the following event types:

- `AML_SUSPICIOUS_ACTIVITY_START`: this event marks the beginning of suspicious activity identified during the investigation
- `AML_SUSPICIOUS_ACTIVITY_END`: This event marks the end of suspicious activity identified during the investigation.

> [!IMPORTANT]
> **Important:** It is recommended that you define a suspicious activity period whenever the time duration of the activities under investigation is known. Make sure to specify a single `AML_SUSPICIOUS_ACTIVITY_START`event and a single `AML_SUSPICIOUS_ACTIVITY_END` event for a given party and risk case ID, or to not specify either. In the absence of `AML_SUSPICIOUS_ACTIVITY_START` and `AML_SUSPICIOUS_ACTIVITY_END`, AML AI selects a period prior to the start of the investigation process.

## Risk case stages and outcomes

Risk case event data helps AML AI understand the lifecycle of
risk cases and how to classify risk cases with positive outcomes (leading to an
exit), ongoing investigations (not finished yet), or negative outcomes (finished
and not leading to an exit).

![AML positive case diagram](https://docs.cloud.google.com/static/financial-services/anti-money-laundering/images/lifecycle-case-positive.svg)
**Figure 1.** Positive case example.

![AML negative case diagram](https://docs.cloud.google.com/static/financial-services/anti-money-laundering/images/lifecycle-case-ongoing.svg)
**Figure 2.** Negative case example.

The stages illustrated in the preceding diagrams can be expressed with risk case
events as in the following steps:

1. **An alert is created for a given party.**

   Alerts are triggered in a number of different ways:
   - An ad hoc alert triggered manually somewhere within your business, for example, by a relationship manager (`type: AML_ALERT_ADHOC`)
   - An automated alert triggered using a Google AML [risk score](https://docs.cloud.google.com/financial-services/anti-money-laundering/docs/concepts/glossary.md.txt#risk-score) as input (`type: AML_ALERT_GOOGLE`)
   - An automated alert triggered by a rules-based or AI system other than AML AI (`type: AML_ALERT_LEGACY`)
   - An alert triggered based on considerations other than pure risk, for example, data exploration to improve quality of AML models (`type: AML_ALERT_EXPLORATORY`)
   - An external request triggered an investigation process (`type: AML_ALERT_EXTERNAL`)
   - An existing investigation process identifies risk with an additional party and starts an investigation process into that party (type: `AML_ALERT_ADHOC`).

   If this alert is related to investigations of other parties and worked as
   part of the same case, it can reuse the same risk case ID. Otherwise, a new
   risk case ID should be used.
2. **Following an alert, an AML investigation process starts
   (`type: AML_PROCESS_START`).**

   While your process may consist of multiple investigation stages, this event
   marks the start of the first stage.
3. **During some investigations, risk is acted upon, or other alerts or
   investigations for other parties are triggered**.

   - A SAR is filed relating to AML risk of this customer (`type: AML_SAR`).
   - The investigation process triggers the process to exit the customer (`type: AML_EXIT`).
   - The investigation process triggers an investigation (`AML_PROCESS_START`) or an alert (`AML_ALERT_ADHOC`) for another party.
4. **The investigation process concludes and no further investigation activity
   for this party is expected as part of this risk case
   (`type: AML_PROCESS_END`).**

   In particular, if there has not been an `AML_SAR` or `AML_EXIT` event prior
   to the end of the process, then AML AI concludes that there
   was not sufficient risk identified to justify these outcomes for this party.

**Suspicious activity period:** Risk case events are also used to specify the
period during which suspicious activity, if any, was identified. To identify
this period, you may specify a single `AML_SUSPICIOUS_ACTIVITY_START` and
`AML_SUSPICIOUS_ACTIVITY_END` pair for a given party and risk case ID. This
input helps models identify risk at the correct time, for example, during or
shortly after the period when the suspicious activity occurred.