Skip to content
This is NOT an official site of the Government of Canada. Click here for the official AI registry.

Helpful Places

Government of Canada – DTPR AI Register

AI Triage for Employment Insurance Claim Recalculations

Eligibility & Public Benefits

What it collects that can identify you

Sensitive personal information
Identifiable data
  • Personal information about EI claimants classified at Protected A, collected by ESDC and used directly in the model's classification decision. The data is drawn from multiple internal sources and the use is verified as consistent with existing Personal Information Banks and Privacy Impact Assessments.

AIA — Section 3.1, Questions 24–34

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 the AIA
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 a backlog of older Employment Insurance claim recalculations into three outcome categories: benefit increase, benefit decrease, or no change. Claims predicted to result in no change are closed automatically, while those likely to result in a benefit change are prioritized for human processing. The model operates on Protected A personal information held by Employment and Social Development Canada and replaces decisions that would otherwise be made by a human agent.

What it collects and what happens to it

Data taken in

Sensitive personal information
Identifiable data
  • Personal information about EI claimants classified at Protected A, collected by ESDC and used directly in the model's classification decision. The data is drawn from multiple internal sources and the use is verified as consistent with existing Personal Information Banks and Privacy Impact Assessments.

AIA — Section 3.1, Questions 24–34

Operational data
Anonymized data
  • Historical EI claim recalculation records from multiple internal ESDC data sources, used as both training data and runtime input. Data is held within a closed system with no internet or external system connections.

AIA — Section 3.1, Questions 30, 32–34

Processing

Classification & Prediction
  • A machine learning classification model (one-shot approach) assigns each EI recalculation claim to one of three outcome categories: Increase in benefit rate, Decrease in benefit rate, or No Change in benefit rate. The model identifies 'No Change' cases with approximately 90% accuracy.

AIA — Project Description (Section 9) and Section 3.1, Question 21

What it does

Deciding (Analytical AI)
Human executes
  • A one-shot classification model predicts whether each EI recalculation will result in an increase, decrease, or no change in benefit rate. For 'No Change' claims, the system renders a decision without direct human involvement; for the remaining claims, the system's prioritization output is acted on by human agents.

AIA — Section 3.1, Questions 9–10; Section 2 Human-in-the-loop

Outputs

A decision about you
Identifiable data
  • For claims classified as 'No Change', the system renders a binding determination to close the recalculation without further review — a decision that directly affects whether a claimant receives a revised benefit. Claims classified as likely to change are prioritized for human agent action.

AIA — Section 2 Human-in-the-loop; Section 3.1, Questions 9–10

A recommendation or prediction
Anonymized data
  • For claims classified as likely to result in a benefit change (increase or decrease), the system produces a prioritization ranking that guides human agents on which recalculations to process first, rather than rendering a final determination.

AIA — Project Description (Section 9)

Run by

Employment and Social Development Canada (ESDC)
  • The federal department responsible for the Employment Insurance Program, which developed and deployed this machine learning system through its Benefits and Integrated Services Branch to reduce a backlog of older EI claim recalculations.

AIA — Project Details, Questions 3–5

Built by

Not stated by the Helpful Places.

Kept for

Retained not specified in the AIA
  • The AIA confirms that an audit trail is maintained recording all decisions and model versions, but does not specify a retention period for system outputs or personal data. A Privacy Impact Assessment has been conducted; refer to ESDC's applicable Personal Information Banks for retention schedules.
  • Duration: not specified in the AIA

AIA — Section 3.2, Questions 15–18, 31

Shared with

Not available to me
  • The data produced by the system — including classification outputs and audit trail records — is not directly accessible to individual claimants. The system operates in a closed environment with no internet connectivity, and outputs are internal to ESDC.

AIA — Section 3.2, Question 33

Available to the accountable organization
  • ESDC maintains a full audit trail of all recommendations and decisions made by the system, including which version of the system was used, all key decision points, and a log of all model changes. The data is controlled by the federal government.

AIA — Section 3.2, Questions 14–19, 29; Section 3.1, Question 29

Stored

Stored locally
  • The system operates within a closed environment controlled by the federal government of Canada, with no connections to the internet, intranet, or external systems. Data remains under federal jurisdiction.
  • Duration: not specified in the AIA

AIA — Section 3.2, Question 33; Section 3.1, Questions 29, 32

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 UseClaimants are entitled to be informed that an automated decision system is being used to process their EI recalculation. A plain language notice is required to be posted through all service delivery channels in use, including internet, in person, mail, or telephone.
  • Right to ContestClaimants whose EI recalculation was closed by the system have access to a recourse process to challenge the decision. The AIA confirms that a recourse process is established for clients who wish to contest the outcome. Contact the Employment Insurance program through Service Canada for further information.
  • Right to an Explanation of a DecisionThe AIA requires that a meaningful explanation be provided with any decision that results in the denial of a benefit or service. However, the system is currently not able to produce reasons for its decisions when required (AIA Section 3.2, Q23). Claimants should contact Employment and Social Development Canada if they require an explanation of a decision affecting their claim.
  • Right to a Human ReviewHuman override of system decisions is enabled. A process is in place to log instances when overrides are performed. Claimants may initiate a recourse process to have their case reviewed by a human agent. Contact Employment and Social Development Canada through Service Canada for more information.

Risks and safeguards

  • Financial & business harmA claimant entitled to a higher benefit rate could be incorrectly classified as 'No Change' and have their recalculation closed without receiving the owed amount. The AIA estimates worst-case losses of $300 (median) or $900 (average) per affected claim, with approximately 10% probability of any benefit change.Safeguard: The model identifies 'No Change' with ~90% accuracy; a recourse process is established for claimants to challenge decisions; human override is enabled and logged; an audit trail records all system decisions with version tracking; documented processes test datasets against biases; and a Privacy Impact Assessment has been completed.
  • Civil liberties harmThe system renders decisions without direct human involvement for 'No Change' claims, affecting claimants' access to a potential benefit entitlement without the procedural safeguard of human judgement. The system cannot produce reasons for its decisions when required, and not all decision points are linked to relevant legislation. No Gender-Based Analysis Plus of the data was undertaken, raising potential concerns about disparate impact on protected groups.Safeguard: A recourse process exists for clients to challenge decisions; human override is possible and logged; an audit trail records all decisions with version information; internal stakeholder consultations included ATIP and Legal Services; a Privacy Impact Assessment was conducted; and the system operates in a closed environment. Plain language notice is required across all service delivery channels.