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-Assisted Classification of Employment Record Comments

Eligibility & Public Benefits

What it collects that can identify you

Sensitive personal information
Identifiable data
  • The system ingests the free-text Block 18 comment field from Record of Employment (ROE) forms, together with the Reason for Separation (RFS) and RFS-Detail codes. ROE data is classified Protected B and contains employment information about individual Canadians, including the circumstances of their job separation. Personal identifiers such as name and Social Insurance Number are present in the source ROE data but are removed before being passed to the AI model.

AIA — Q46: Protected B / Protected C classification; Q53: Input data is ROE form Block 18 comments

Run by
Employment and Social Development Canada (ESDC)
Where
No fixed location
Kept
Retained Minimum 12 years (aligned with existing ROE retention period; deleted only when the ROE itself is deleted)
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 automatically reads the free-text comment field on Records of Employment submitted by employers and determines whether those comments contain information about the reason an employee left their job. When it finds relevant information, it may update the Reason for Separation code used in Employment Insurance (EI) benefit processing. The system handles straightforward cases to reduce the workload of Service Canada agents, while more complex cases are still reviewed by a human. Importantly, users are not informed that AI is involved in processing their information.

What it collects and what happens to it

Data taken in

Sensitive personal information
Identifiable data
  • The system ingests the free-text Block 18 comment field from Record of Employment (ROE) forms, together with the Reason for Separation (RFS) and RFS-Detail codes. ROE data is classified Protected B and contains employment information about individual Canadians, including the circumstances of their job separation. Personal identifiers such as name and Social Insurance Number are present in the source ROE data but are removed before being passed to the AI model.

AIA — Q46: Protected B / Protected C classification; Q53: Input data is ROE form Block 18 comments

Processing

Classification & Prediction
  • A machine learning model trained on historical ROE comment data classifies whether a Block 18 comment contains Reason for Separation information and predicts the appropriate RFS code. Additional business logic then determines whether the RFS code on the form should be adjusted. The model was trained to replicate the judgement of human agents on simple cases.

AIA — Q24: System evaluates text to classify RFS information; Q25: System produces prediction about RFS information in comment

What it does

Deciding (Analytical AI)
Human decides
  • The model classifies whether Block 18 text comments contain Reason for Separation information and whether the existing RFS code is consistent with that information. For simple cases it may adjust the RFS code automatically; for complex or uncertain cases, it passes the item to a human agent who makes the final determination.

AIA — Q21: Partial automation; Q19: Decisions involve determining if comment contains RFS information

Understanding (Semantic AI)
Autonomous
  • The system uses natural language processing to extract meaning from free-text employer comments in Block 18 of the ROE form, matching them to structured Reason for Separation categories without human review of each comment.

GC AI Register — Capabilities: Natural language processing · AIA — Q8: Text and speech analysis capability selected

Outputs

A decision about you
Anonymized data
  • For simple cases, the system outputs an adjusted Reason for Separation (RFS) code that is written to the interpretative notes used in EI claim processing. The original ROE document is not modified. Outputs are stored in a de-identified manner and retained for a minimum of 12 years. For cases the system cannot confidently resolve, no output action is taken and the case is passed to a human agent.

AIA — Q7: Actions saved in interpretative notes; Q25: System produces prediction about RFS; Q37: Outputs stored in de-identified manner

A recommendation or prediction
Anonymized data
  • For cases outside the scope of simple automated action, the system's classification serves as a recommendation to the human agent reviewing the file, surfacing whether the Block 18 comment may contain relevant RFS information that warrants attention.

AIA — Q22: In cases where AI does not take action, text comment left for human agent to inspect

Run by

Employment and Social Development Canada (ESDC)
  • Employment and Social Development Canada deploys this AI system within Service Canada's Employment Insurance benefits delivery process. The Chief Data Officer Branch (CDOB), EI Benefits Delivery Services (BDS), and the Innovation, Information and Technology Branch (IITB) are jointly responsible for its implementation and operation.

GC AI Register — Department: Employment and Social Development Canada

Built by

Government of Canada
  • The system was developed internally by the Government of Canada. The Chief Data Officer Branch led the development of the machine learning model, trained on past ROE data held by the institution.

GC AI Register — Developed by: Government of Canada · AIA — Q51: Who collected the data used for training? Your institution

Kept for

Retained Minimum 12 years (aligned with existing ROE retention period; deleted only when the ROE itself is deleted)
  • AI system outputs (RFS predictions and interpretative notes) are retained for a minimum of 12 years to align with the existing ROE retention period under PIB ESDC PPU 385. Data is de-identified before storage.
  • Duration: Minimum 12 years (aligned with existing ROE retention period; deleted only when the ROE itself is deleted)

AIA — Q37: Outputs retained for minimum 12 years to align with ROE retention period

Shared with

Available to the accountable organization
  • AI system outputs and audit logs are available to ESDC / Service Canada staff involved in EI benefit processing. Access is managed through a formal process to grant, monitor, and revoke permissions. The system interfaces with internal IT systems used in EI administration.

AIA — Q24 (Mitigation): Process in place to grant, monitor, and revoke access; Q50: System interfaces with other IT systems

Not available to me
  • Individual claimants and employers do not have direct access to the AI model's outputs or the interpretative notes it generates. The original ROE document submitted by the employer is not modified by the system. Claimants are not informed that AI processed their ROE comments.

GC AI Register — AI use disclosed to users: N · AIA — Q7: Original ROE document will not be modified

Stored

Stored locally
  • The system operates within a closed Government of Canada internal network with no connections to the Internet or external systems. Data is controlled by the federal government and stored within Canadian federal government infrastructure. PIB: ESDC PPU 385.
  • Duration: Minimum 12 years

AIA — Q34: Closed system with no Internet connections; Q47: Federal government controls the data

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 UseIndividuals whose Records of Employment are processed by this system are currently not informed that AI is involved (the register entry states AI use is not disclosed to users). The Directive on Automated Decision-Making requires ESDC to publish a plain-language explanation of the system's role in decision-making. Individuals can access this via the Algorithmic Impact Assessment published on the Open Government Portal at https://open.canada.ca/data/en/dataset/daa9ca66-566f-4c2e-a285-d2e217c2a00f.
  • Right to a Human ReviewA recourse process is established for clients who wish to challenge decisions related to their EI claim. Human agents retain the ability to override AI-generated RFS adjustments at later stages of EI processing. Complex cases and cases where the AI lacks sufficient confidence are automatically routed to human agents. For queries or recourse, contact Service Canada at 1-800-206-7218 or visit a Service Canada Centre.
  • Right to Algorithmic TransparencyThe Algorithmic Impact Assessment for this system is publicly available on the Government of Canada Open Government Portal. It describes the system's purpose, the data used, the decision logic, and how the AI fits into the EI benefits delivery workflow. Under the Directive on Automated Decision-Making, ESDC is required to publish a plain-language explanation of the system covering its role, input data, criteria, outputs, and principal factors behind decisions. Access the AIA at https://open.canada.ca/data/en/dataset/daa9ca66-566f-4c2e-a285-d2e217c2a00f.
  • Right to ContestA recourse process is in place for clients who wish to challenge decisions made during EI claim processing. Because the system's RFS adjustments feed into the broader EI eligibility decision made by human agents, individuals may contest the overall EI determination through existing Service Canada appeal mechanisms. The impacts of any AI decision are reversible during the claim lifecycle. Contact Service Canada at 1-800-206-7218 or visit a Service Canada Centre to initiate a review.

Risks and safeguards

  • Financial & business harmAn incorrect RFS code assignment by the AI could affect the speed or accuracy of an EI eligibility determination, potentially delaying or mis-characterising a claimant's benefit entitlement.Safeguard: The system is designed so that errors are reversible — administration and fact-finding steps by Service Canada agents later in the process can correct any AI-assigned RFS. The system cannot cause benefits to be erroneously denied without subsequent human intervention. Confidence thresholds restrict the model to acting only on cases it can handle with sufficient certainty. A fairness assessment was conducted to evaluate model equity across sociodemographic groups.
  • Societal & cultural harmAI use is not disclosed to claimants (AI use disclosed to users: N), creating a transparency gap that may erode public trust in EI administration. Members of the public cannot know that an automated system contributed to handling their ROE information.Safeguard: The AIA is publicly available on the Open Government Portal. The Directive on Automated Decision-Making requires a meaningful explanation to be published for common decision results, covering the system's role, input data, criteria, and output. ESDC is required to make this information discoverable via a departmental website.