The Canada School of Public Service (CSPS), the federal government’s in-house training division, has issued a request for proposals (RFP) seeking up to three artificial intelligence (AI) companies to build a tool that will be used for regulatory review.
The RFP is the first to use a list of 74 companies approved to bid on government AI projects posted in January. The government wants to use the AI platform to keep track of all its regulations so it can eliminate overlapping, inconsistent or outdated rules.
Talking Point
The federal government is looking to use artificial intelligence to manage its 2,600 regulations. It’s hiring up to three companies to build technology that will help regulators find rules to cut or change. The new system comes as the government rolls out a wide-ranging plan to cut the red tape that businesses complain is holding them back.
Senior government officials—like former Treasury Board presidents Scott Brison and Jane Philpott and chief innovation officer Alex Benay—have talked up AI’s potential to improve service delivery to citizens and decision-making within the civil service. But adoption has been slow. Departments and agencies spent a combined $8.4 million on AI tools between November 2015 and September 2018, according to a government report.
Proposals for the platform were due Monday, but the government extended the deadline to May 22. Selected firms—up to three in the first round—are expected to build a prototype in 65 business days and, if they pass, must deliver the finished product with access for up to 200 users in another 100 business days. Companies cannot charge more than $86,975 for the prototype and $217,438 for the fully-functioning system.
The new AI-powered platform could be used to identify “overlapping, obsolete or outdated regulations or requirements in the regulatory stock as well as opportunities to reduce regulatory burden on stakeholders,” according to bid documents. Regulators may also use it to assess how specific economic sectors or small businesses are impacted by all levels of regulation that apply to them.
The platform’s machine-learning algorithms will be trained using publicly available databases of federal legislation and regulation, as well as those for the provinces and territories, the United States and the European Union.
Ten of the 65 eligible firms on the sources list have been chosen to bid, including global consulting firms Accenture and KPMG, as well as smaller Canadian software companies like Nuvoola and SageTea.
Another company, Lixar IT, has already worked on a major AI project for the government. In 2017, the Ottawa-headquartered firm built a search and recommendation system for the online portal through which citizens can submit Access to Information and Privacy requests.
The bid document for the regulatory platform lays out several test scenarios that suggest the finished product could help cut red tape.
One case involves helping a civil servant identify the 139 regulations for which Innovation, Science and Economic Development Canada (ISED) is responsible, and ranking them by how urgently they need to be modernized based on how old, prescriptive, contradictory and duplicative they are. The system will also need to generate graphs and reports that show the links between those rules and their relationship to the government’s 2,461 other regulations.
Another trial involves finding all instances of a specific word like “environment” in federal regulations and generating a graph of how those rules are linked, to show how changing the term’s definition in one regulation could impact impact others. That’s currently a “tedious and time consuming task, subject to human error, and requires that individuals have some knowledge or expertise on how different regulations are interrelated,” according to the bid document.
The federal government has made significant promises on regulatory reform in the last six months. Finance Minister Bill Morneau’s Fall Economic Statement in November 2018 revealed that federal regulators will soon be required to consider “efficiency and economic growth” as part of their mandates.
The economic statement also outlined a number of measures to modernize Ottawa’s rulebook, which contains 2,600 regulations. They include issuing a rule modernization bill every year to remove obsolete or overlapping requirements, and setting up a new committee to advise on regulatory competitiveness.
Asked whether the new AI system would be used to put together the bill or inform the work of the committee, Vanessa Vermette, director general of communications and engagement at CSPS, said the platform “is not being planned or designed with any one particular legislative initiative in mind.” The technology will be available to 34 departments and agencies—including key rulemakers, like the national food inspection and public health agencies, and the Treasury Board Secretariat, the government’s in-house policy shop. The CSPS will set up demonstration projects with those organizations to figure out how they might use the system.
Martin Potvin, a Treasury Board spokesperson, said the department “will consider using the platform once it is selected and we have further details on its functionality.”
