AI-Assisted Triage for Employment Insurance Claim Reviews

Eligibility & Public Benefits

What it collects that can identify you

Sensitive personal information

Identifiable data

EI production data about individual claimants, including benefit rate and entitlement information derived from Records of Employment. Data is classified Protected A and is controlled by the federal government. A primary key is used for de-identification at certain lifecycle stages.

AIA — EI ML Workload

Also collects operational data, which is anonymized data.

Run by: Employment and Social Development Canada (ESDC)
Where: No fixed location
Kept: Retained Not specified in source documents
Shared with: Accountable organization
Your copy: You cannot see the data it holds about you. What you can do

What it is for

Eligibility & Public Benefits

This system uses a machine learning model to classify Employment Insurance claim recalculations into three categories — benefit increase, benefit decrease, or no change — so that officers can focus their attention on cases most likely to affect a claimant's payment. Cases predicted to result in no change may be closed without full officer review, though clients can request human reconsideration at any time. The system is used internally by Government of Canada employees and does not make final benefit decisions on its own.

What it collects and what happens to it

Data taken in

Sensitive personal information

Identifiable data

EI production data about individual claimants, including benefit rate and entitlement information derived from Records of Employment. Data is classified Protected A and is controlled by the federal government. A primary key is used for de-identification at certain lifecycle stages.

AIA — EI ML Workload

Operational data

Anonymized data

EI benefits data from the EI production systems, including claim status (terminated or dormant), benefit rate, and weeks of entitlement. Data originates from multiple internal sources and interfaces with other IT systems.

Processing

Classification & Prediction

A Random Forest model trained on EI production data classifies each recalculation work item into three outcome categories: increase in benefit rate, decrease in benefit rate, or no change. The model achieves approximately 90% accuracy in identifying no-change cases.

What it does

Deciding (Analytical AI)

Human decides

The Random Forest model classifies each EI claim recalculation into one of three outcome categories. Officers retain authority over final decisions; the model supports — and does not replace — human judgment. Randomized manual spot checks are conducted on cases predicted as no-change.

Outputs

A recommendation or prediction

Anonymized data

The system produces a triage classification for each EI claim recalculation: increase, decrease, or no change in benefit rate. Cases classified as no-change may be closed; cases predicted to result in a change are prioritized for officer review. This is an advisory output — officers and the Integrity Service Branch retain oversight.

Run by

Employment and Social Development Canada (ESDC)

The federal department responsible for deploying and operating the EI Machine Learning Workload system within the Employment Insurance Program, Benefits and Integrated Services Branch.

Government of Canada AI Register — EI ML Workload

Built by

Government of Canada

The model was developed internally by the EI Program Performance team in consultation with stakeholders within the EI program; no external vendor was involved.

AIA — EI ML Workload

Kept for

Retained Not specified in source documents

An audit trail recording all recommendations and decisions made by the system is maintained. All key decision points and system version information are logged. The retention duration for these records is not specified in the available source documents.
Duration: Not specified in source documents

Shared with

Available to the accountable organization

Output classifications and audit trails are available to ESDC officers and the Integrity Service Branch. The system operates within a closed system with no connections to the Internet, Intranet, or external systems.

Not available to me

Individual claimants do not have direct access to the system's classification output or audit trail for their own claim. Claimants interact with EI officers who mediate access to information about decisions affecting them.

Stored

Stored locally

The system operates within a closed system (no Internet, Intranet, or external connections). Data is controlled by the federal government of Canada and stored within federal infrastructure. The system uses Protected A classified data.
Duration: Not specified in source documents

How to read the colours

Can it identify you?

Anonymized data: Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
Pseudonymous data: Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
Identifiable data: The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.

Who completes the loop?

Human decides: This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
Human executes: This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
Autonomous: This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.

Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.

What you can do

Ask about this system

Questions go to the Helpful Places, not the vendor.

Your rights

Right to Be Informed of AI UseThe AIA is publicly available on the Government of Canada Open Data portal. A plain-language explanation of the system's role in the decision-making process, including input data, criteria, and outputs, must be published and discoverable on the departmental website per the Directive on Automated Decision-Making requirements.
Right to ContestClients who wish to challenge a decision supported by this system may use the established recourse process for EI decisions. If a client makes an inquiry about their file, officers will proceed with a full manual recalculation regardless of the system's classification.
Right to a Human ReviewHuman override of system decisions is enabled. Officers can override any classification produced by the model, and all override instances are logged. The Integrity Service Branch independently reviews undeclared contentious issues to ensure accuracy and fairness.
Right to Algorithmic TransparencyPer the Directive on Automated Decision-Making (Impact Level 1), a meaningful explanation of the system's role, input data and sources, evaluation criteria, and outputs must be made publicly available in plain language through the AIA and a departmental website. The algorithm is not a trade secret and is described as interpretable.

Risks and safeguards

Financial & business harmThe model may misclassify a claim entitled to higher benefits as no-change, resulting in financial shortfall for the claimant (estimated worst-case: 10% error rate on no-change predictions).Safeguard: Randomized manual spot checks are conducted by agents on no-change cases; clients who inquire about their file trigger a full recalculation; a recourse process allows clients to challenge decisions; human override of system decisions is enabled and logged; the Integrity Service Branch reviews contentious issues.