Advanced analytics and automation for processing applications
Canada is a top choice for people who want to visit, study and work, or to start a new life with their families. Newcomers strengthen our communities, help us fill jobs and contribute to our economy. We’re using new technologies to modernize processing and help reduce the number of waiting applications so we can focus on more complex cases.
We use computer-based tools, sometimes called models or systems, which help officers by automating different tasks in processing applications.
On this page
- Making routine processing decisions
- Helping assign applications to officers
- How the model “rules” are developed
Making routine processing decisions
Some tools are designed to help officers by making some decisions in the application review process for routine cases. These are decisions to approve all or part of the application. Our tools don’t refuse or recommend refusing applications.
The types of decisions the tools make can be different based on the needs of the program and the clients applying for it. The overall goal is for the tools to take on the most straightforward cases and parts of the decision-making process, so officers can focus their attention on more complex work.
For example, we use a tool to help officers process visitor record applications in the following way:
- The tool reviews and approves eligibility for the most routine cases, which would be very likely to be approved by an officer. Any cases that aren’t considered routine are sent to officers to review for eligibility. The tool doesn’t refuse or recommend refusing these applications and the majority will still be approved by officers after they review them. The tool uses rules both developed by officers and generated through machine learning based on data from previous Immigration, Refugees and Citizenship Canada (IRCC) files.
- An officer
- does all security and criminal background checks to determine if a person is admissible to Canada
- makes the final decision on granting the visitor record
Text version: How the eligibility and approval process works
- The tool reviews applications to determine whether a full officer review is required.
- For applications that may not require a full officer review, the tool sorts them by how complex they are.
- Applications that are identified as “complex” are sent directly to an officer for a full manual review. Applications that are identified as “routine” have their eligibility approved automatically.
- The application is then sent to an officer to review for admissibility and to make a final decision. The officer has the ability to overturn the tool’s eligibility decision during this review.
Note: For some types of applications, we are able to create processes to fully automate all approval decisions on an application (including eligibility, admissibility to Canada and the final decision) within our existing case management system, rather than using a separate tool. These processes use rules developed by IRCC officers.
How it helps
When the tool is able to make some decisions on routine cases, officers can focus on reviewing and making decisions on more complex cases. It also helps clients get a decision on their application sooner.
Helping assign applications to officers
Some tools are designed to triage applications. This means they sort them into bins or groups where they can be more easily assigned to officers based on their expertise, without making any decisions or recommendations on the application. They do this by
- assessing the expertise needed to review an application
- separating the applications into bins
- creating case annotations or notes for officers from information in our case management system
Then the applications will be assigned from the bins to officers who have the right skills and experience to review and make a decision on them.
For example, we use a tool to help assign study permit applications based on how complex they are and the type of expertise needed to review them. A less complex study permit application could be one for an applicant under 21 who has included all required documents (confirmed by an officer) and has no adverse information on their file.
How it helps
This process helps ensure cases are efficiently assigned to officers who are able to assess them, without influencing their decisions. The technology automates a lot of the clerical and repetitive tasks involved in sorting applications, so officers can spend their time reviewing them and making decisions. This helps us make the most of our processing networks, and reduces the time needed to process all applications so clients can get decisions more quickly. We may move applications between our offices to make sure they are processed as efficiently as possible. This means applications may not be processed at the office closest to where the applicant lives or where they submitted their application.
How the model “rules” are developed
The tools make decisions or sort files based on pre-determined rules or instructions. These rules can be developed in different ways:
-
Rules developed by officers
Rules are developed by people based on their expertise in the type of application being reviewed.
-
Rules based on data and advanced analytics
Rules are developed through advanced analytics, which use machine learning technology to find patterns in data.
For example, a tool can scan through a large number of previously reviewed applications to determine how likely it is that a similar type of file would be approved by an officer, or what type of expertise is needed to review the file.
Even though these rules aren’t developed by officers, they are carefully reviewed before implementation, and the performance of the model is monitored often so improvements can be made if necessary. Any updates to the rules are overseen and controlled by IRCC—the system doesn’t create or use new rules on its own.
Notes
- The rules that the tools use are kept internal to IRCC to help prevent fraud and protect the integrity of our immigration programs, as well as the safety and security of Canadians.
- The tools are reviewed routinely to make sure they are working as intended and that the results are consistent with applications that receive a full human review.
Any rules used by tools, whether developed by officers or through advanced analytics, go through an extensive review process involving
- legal experts who assess the rules for potential implications under the Canadian Charter of Rights and Freedoms and consideration of legislative criteria
- policy experts who verify that rules follow best practices on using automated decision support systems, as determined by the Treasury Board of Canada Secretariat
- migration officers who assess how rules are relevant to program requirements
- data analysts who verify that rules are grouping applications in a way that is consistent and makes sense
What are “black-box” technologies and do we use them?
Black-box is a sub-category of artificial intelligence, in which tools use internal processes and logic that aren’t visible or easily understood by people. These tools use many layers of complex analysis to assess information to make a prediction or classification, or to take appropriate actions as needed. The actions a black-box tool takes and its performance can be measured and monitored, but the logic it uses to take those actions isn’t well understood, even by data experts.
The rules we use to help sort and process applications use an if-then logic to make decisions, so they are easy for those reviewing the tools to review and understand. In a very simple example, if an application needs additional documents, then it should be transferred to an officer for review. These aren’t black-box technologies.
We do use some black-box technologies outside of decision making on applications to help us work more efficiently to serve clients. For example, we use a tool that helps overseas migration offices sort and prioritize messages sent by clients.
While the inner workings of these tools may not be fully explainable, their key functions and operations are well understood, and their performance can be reliably measured.
We don’t use black-box technology to make decisions on applications.
Page details
- Date modified: