Canada’s AI adoption is accelerating, but accountability is struggling to keep pace

Artificial intelligence is moving deeper into Canadian businesses, but the structures intended to govern it are not always developing at the same speed. The resulting problem is technical and cultural. Organisations must decide who can approve AI tools, who monitors their performance and who answers when an automated system produces a damaging or discriminatory outcome. These questions become more complicated when individual departments buy or build AI systems outside the control of central information technology teams.

Verified Canadian evidence indicates that adoption is accelerating. Statistics Canada data reported by Digital Journal showed that 12.2 per cent of Canadian businesses used AI to produce goods or deliver services in 2025, compared with 6.1 per cent in 2024. A further 14.5 per cent expected to adopt the technology during the following 12 months. The figures suggest that business use doubled within a year. While the proportion remains relatively modest across the economy as a whole, much higher adoption rates have been reported among larger businesses and organisations actively pursuing digital transformation. The difference illustrates an important feature of the Canadian market: AI adoption is uneven, with some organisations scaling multiple projects while others are only beginning to experiment.

AI is going around central IT

One of the most immediate governance risks is the use of AI tools without formal organisational approval. At the CIO Association of Canada’s 2026 Peer Forum in Vancouver, Canadian data presented by Lenovo indicated that only 18 per cent of organisations had a formal AI policy, while 67 per cent of employees were using unsanctioned AI tools. The practice is commonly described as “shadow AI”, reflecting situations in which employees introduce AI services without the knowledge or approval of the technology function. The findings were discussed with Canadian technology leaders at the conference.

Shadow AI creates an awkward separation between authority and accountability. A marketing, finance or customer-service team may select a tool because it solves an immediate problem, but the consequences of that decision can extend across the enterprise. Employees could, for example, enter commercial, customer or personnel information into services that have not been assessed for privacy and security. AI-generated content may also be used without sufficient checking, creating the possibility of inaccurate information entering business decisions or customer communications.

The CIO can therefore remain responsible for the security and resilience of the organisation’s systems without controlling every AI purchasing or deployment decision. This is one reason that governance cannot be confined to an IT policy document. It must influence procurement, data handling, workforce training and operational decision-making throughout the organisation.

The governance challenge is also connected to Canada’s continuing difficulty in converting AI experimentation into measurable results. Research from BDO Canada, based on an Angus Reid survey of 520 Canadian business leaders, found that 46 per cent of organisations were experimenting with AI without achieving meaningful return on investment. A further 18 per cent were actively embedding it into operations and workflows, while 26 per cent remained focused on modernising core systems. The study suggests that many organisations are measuring activity rather than business outcomes.

Other research presents a similar picture. An RSM survey covering 203 Canadian respondents at companies with annual revenues of between $30 million and $1 billion found that 69 per cent had partly or fully integrated AI. However, only 15 per cent described their approach as transformational across the enterprise. Among businesses reporting limited success with pilot projects, 53 per cent identified data quality as a barrier to scaling, while 47 per cent pointed to integration difficulties. Security concerns and unclear returns were each reported by 33 per cent. These findings matter because governance is sometimes presented as an obstacle to innovation. In reality, inadequate governance can prevent organisations from moving beyond scattered pilots. If ownership is unclear, data is unreliable or different functions use incompatible systems, even a technically successful pilot may be difficult to scale. Good governance should therefore enable adoption rather than merely restrict it. Clear approval routes, reliable data, performance measures and named accountable owners can help organisations distinguish valuable applications from projects driven primarily by novelty.

AI is also changing expectations of the chief information officer. Canadian technology leaders are increasingly expected to connect infrastructure, cybersecurity, data, productivity and organisational change. This places the CIO between two competing demands: maintaining stable and secure systems while supporting a technology that could substantially alter established processes.

Also at the CIO Peer Forum, technology leaders described this tension directly. Discussions focused on how CIOs could scale AI without undermining privacy, trust or accountability. Kate Burnett-Isaacs, then acting chief data officer at Housing, Infrastructure and Communities Canada, described an approach based on identifying the problem, considering privacy and ethical implications, and placing projects within a defined lifecycle. The discussions positioned governance as a framework that must evolve with operational practice. This broader mandate inevitably crosses into the work of human resources, legal, compliance, risk and business operations. AI can change how tasks are divided, which skills are required and how decisions are reviewed. Workforce redesign is consequently not an issue that the CIO can resolve alone.

The emergence of the Chief AI Officer adds another layer to the ownership debate. A dedicated AI executive can provide strategic focus, but creating a new title does not automatically clarify responsibility. The central question is whether the role possesses genuine authority over budgets, data, risk acceptance and business-unit deployments. A global IBM study of 2,000 chief executives, reported by Digital Journal, found that 76 per cent of organisations had a Chief AI Officer. The same coverage noted that 77 per cent of CEOs believed talent and technology leadership roles were converging. However, Gartner advisory director Jonathan Tabah questioned whether the CAIO position would become a standard feature of the corporate hierarchy. The evidence points to continuing uncertainty about whether AI requires a permanent executive office.

The Canadian picture reinforces the need for caution. PwC Canada’s 2026 CEO research found that 94 per cent of Canadian CEOs used AI to some extent, but only 29 per cent had scaled it throughout the business, compared with 43 per cent globally. This leaves many Canadian CAIOs, where the role exists, governing portfolios that may still be dominated by pilots. Installing a CAIO without defining the relationship with the CIO, chief data officer, chief information security officer and legal function could fragment accountability further. The more useful question is therefore not whether an organisation has a CAIO, but whether it has a coherent operating model for AI.

Shared governance, identifiable accountability

Canadian organisations need distributed participation in AI governance but should avoid distributed ambiguity. Business functions should retain responsibility for the outcomes they seek from AI. Technology teams should oversee architecture, integration and security. Privacy, legal and compliance specialists should examine data and regulatory risks, while human resources should address workforce and training implications. However, every AI system also needs an identifiable senior owner with enough authority to stop deployment, require corrective action or withdraw the system when its performance falls outside agreed limits.

The evidence from Canada does not support the idea that AI governance belongs exclusively to either the CIO or a newly appointed CAIO. It points instead towards cross-functional oversight backed by clear individual accountability. AI adoption is already moving beyond controlled experiments. The organisations most likely to benefit will not necessarily be those that deploy the greatest number of tools. They will be those that can explain what each system is for, what risks it introduces, how its performance is tested and who is answerable for the result.