Review & Audit Solutions
Solutions overview
Fourthline’s Case Review Portal is a secure, user-friendly web environment for handling cases.
We offer the following functionalities in the portal:
Operating model
Agree your operating model with your Fourthline delivery manager:
| Operating model | Description |
|---|---|
| Fourthline operated | Fourthline agents process and audit cases |
| Fourthline powered | Your agents process and audit cases |
Case Processing
Every case is first processed by our AI agent and, if it can't assign an automated outcome, the case passes to a human agent. Human agents manually review cases flagged by the AI agent to perform additional anti-fraud and anti-financial crime checks. They may send complex, poor quality, or incomplete cases back to you to request more information. Each case is then assigned an outcome and we generate a detailed, fully compliant, evidenced CDD Report.
Case Processing
Case Processing includes the following features:
| Feature | Description |
|---|---|
| Checks library | Configure checks to validate more than 200 data points. |
| Automatic case assignment | The portal automatically assigns cases to human agents based on a predefined set of filters, e.g. product/solution, hit type, SLA time, and business partner. |
| Case categorization | To streamline workflows, cases are split into unassigned, assigned, and pending second opinion categories. |
| Time tracking | Human agents can set their work status to track their activity and productivity. |
| Outcome management | Human agents can approve, reject, and update check outcomes. |
| Real-time AI agent double-check | The AI agent rechecks any data point updated by a human agent. |
| Responsive video viewing | Human agents can automatically play and pause selfie and document videos, change the playback speed, and enlarge the video frame. |
| Responsive image viewing | Human agents can compare a series of images next to each other, adjust the view of images taken at difficult angles to make them easier to read, and zoom in up to 100 times. |
| Google Maps and Street View integration | Human agents can perform additional checks on locations. |
| Data correction | If we detect any errors in information entered by the client, we correct it in line with their ID document. The CDD Report contains the both original and the corrected values. |
| Escalations | Human agents can escalate complex cases to supervisors. The portal creates an audit history of when, how often, and why cases are escalated. |
| Four-eyes review | A second human agent performs an extra check on cases with: • A high-risk geolocation to ensure the client isn't applying from a sanctioned country. • Low scores for facial recognition and document image sharpness |
| Human review (optional) | If recommended by your regulator (e.g. BaFin) you can configure for a human agent to review every case, even those successfully processed by the AI agent. |
Our default service-level agreement (SLA) states that we must process 80% of cases per month within 5 minutes.
So far, we have never failed to meet this.
We have proven we can handle extreme growth while upholding the SLA.
Case Auditing
Human agents audit completed cases, like a live, interactive CDD Report. They can't change the original outcomes.
Case Auditing
Case Auditing includes the following features:
| Feature | Description |
|---|---|
| Case filter | Filter completed cases to audit by human agent, case identifier, client identifier, product/solution, and time frame. |
| Checks library | Validate 200+ data points to understand how specific checks or whole cases were assigned their final outcome. |
| Responsive video viewing | Auditors can automatically play and pause selfie and document videos, change the playback speed, and enlarge the video frame. |
| Image comparison | Auditors can compare a series of images next to each other, adjust the view of images taken at difficult angles to make them easier to read, and zoom in up to 100 times. |
| Google Maps and Street View integration | Auditors can perform additional checks on locations. |
| Time tracking | Auditors can set their work status to track their activity and productivity. |
| Check correction | Auditors can correct incorrect checks. |
Case Guidance
Case Guidance features detailed, specific reference information and procedures on how best to process all checks, covering more than 500 ID document types. The guidance is dynamically displayed exactly when and where needed to assist human agents.
Case Guidance
Quality Assurance
Performing quality assurance (QA) reviews on completed Identity Verification cases helps eliminate or minimize processing errors, improve criteria and procedures, and enhance human agents' performance and training.
Quality Assurance
You can leverage this based on the following operating models:
| Operating model | Description |
|---|---|
| Fourthline operated | We can share the findings from our internal QA reviews with you, as an additional service. |
| Fourthline powered | You can enable the QA section in your Case Review Portal to perform your own reviews. |
Fourthline QA reviews
Fourthline's highly qualified Assurance Team performs independent, unbiased QA reviews on a daily basis.
The team forms part of the second line of defense within Fourthline.
The QA reviewer re-checks the entire case, including the AI agent checks and any AML hits. Reviews are then subject to further oversight by senior Assurance Team members or our Anti-Financial Crime (AFC) Team. We also periodically validate case data for accuracy, privacy, and correct logging.
All our processes and evidence comply with the requirements of Fourthline's external auditors, as part of our ISAE 3000 certification. We use the information from reviews to improve our offering and continually calibrate our AI agent.
Sampling methodology
Sampling methodology
QA reviews are conducted on 2 sample types:
- Random samples from the entire population of cases to obtain a statistically relevant assessment of quality, following industry best practices
- Targeted samples to actively monitor and search for errors in specific sets of cases, e.g.:
- Cases with increased fraud risk
- Cases for new business partners, products, markets, or document types
Categories and Priorities
Categories and Priorities
The following categories and priorities are available for reviewers to log processing errors:
| Category | Description | Scenarios |
|---|---|---|
| Missed fraud | The status code should have been in the 5xxx range. | Priority level: High • Fraud detected after Identity Verification case successfully completed |
| Missed rejection | The status code should have been in the 4xxx range. | Priority level: Medium • Another person assisted in capturing the selfie, selfie video, document, or document video • ID document data doesn't match the personal data entered by the client Priority level: High • Increased risk data points, such as an unsupported residency country • Underage client • Other verifications incorrectly approved |
| Missed sendback | The status code should have been in the 2xxx or 3xxx range. | Priority level: Low • Unacceptable document condition • Poor document photo quality • Poor selfie quality • Insufficient photos uploaded • Insufficient security features • Inadequate client input • Non-residential address • GDPR breach Priority level: Medium • Document photo is of a photocopy • Missing device metadata Priority level: High • Expired document • Unsupported document • Other verifications incorrectly approved |
| Incorrect data | Applicable for errors related to ID document data. | Priority level: Low • Inaccurate client input, such as nationality, document type, other Priority level: Medium • Errors in the client's name |
| Incorrect screening outcome | Applicable for errors related to AML Screening . | Priority level: Medium • Potential AML hit accepted (only for AML Monitoring) Priority level: High • Confirmed AML hit accepted |
Reports
Reports
We share findings with you on a monthly basis, unless agreed otherwise. On request, we can email high-priority cases to you immediately.
Reports don't contain sensitive case data, only case identifiers.
Updated 17 days ago