The Bank of Canada recently found that more than two-thirds of Canadian business leaders use AI personally during a typical workweek. Yet only 8% of businesses report using AI significantly in their core operations.
Still, almost every conversation about technology eventually becomes a conversation about AI.
- Which model are you using?
- Have you rolled out Copilot?
- Are you experimenting with agents?
- What’s your AI strategy?
For many credit unions, the more useful question might be:
Where, exactly, does AI fit into the way the organization actually gets work done?
Access to AI is now the easy part. We’ve become remarkably good at putting AI into individuals’ hands, but we haven’t yet become nearly as good at putting AI into the way organizations actually work.
The Experimentation Phase Is Already Well Underway
The Bank of Canada’s 2026 Financial System Survey provides an interesting snapshot. Nearly all respondents reported using AI. Banks, broker-dealers and credit unions expect to expand its use across areas including operational processes, financial-crime prevention, risk management and stress testing.
But unsurprisingly, 58% cited difficulty integrating AI into their existing infrastructure and workflows.
That matters, because it suggests the next challenge isn’t getting AI into the organization. It’s getting AI into the operation.
There’s an important difference.
A Copilot Can Make a Person Faster. A Workflow Can Make the Organization Better.
Consider a commercial credit analyst. Give that analyst an AI assistant, and it might help summarize financial statements, analyze information or draft a credit narrative. Useful? Absolutely, but the analyst still needs to manually run tasks:
- find the right documents;
- determine whether anything is missing;
- retrieve information from different systems;
- compare current and previous financials;
- check lending policy;
- identify exceptions;
- send the file for the appropriate approval;
- record the decision;
- make sure the relevant systems are updated.
AI might make one person considerably faster while leaving the underlying process almost completely unchanged.
That’s why the next phase of AI adoption in financial services will be less about AI tools and more about AI-enabled workflows.
Put AI Where It Helps. Keep Judgment Where It Matters.
The Bank of Canada’s findings reinforce another important point:
Financial institutions generally aren’t using AI to replace human judgment or fully automate critical decisions because of the financial, legal and reputational consequences involved. Instead, they’re largely using it to complete existing tasks faster.
That’s exactly the right starting point. Imagine a commercial loan review.
AI could:
- extract information from financial statements;
- compare it with previous periods;
- identify missing documentation;
- summarize material changes;
- surface relevant policy;
- prepare a first-pass briefing for the analyst.
But should AI independently decide whether that member receives the loan? That’s a very different question.
The opportunity isn’t to automate the judgment. It’s to remove the work surrounding the judgment.
And financial institutions do a lot of that kind of work.
- Commercial lending
- AML investigations
- Fraud cases
- Member onboarding
- Regulatory reporting
- Mortgage renewals
- Complaints
- Treasury exceptions
In each case, experienced people spend time gathering information, moving between systems, reading documents, chasing inputs and preparing themselves to make a decision.
Those are precisely the places where AI becomes valuable.
AI Shouldn’t Sit Beside the Process
There’s another risk worth considering. If we simply give employees increasingly powerful AI tools without redesigning the process around them, we may accidentally create another disconnected layer of technology.
- An employee receives a document.
- Downloads it.
- Uploads it somewhere else.
- Asks AI to analyze it.
- Checks the answer.
- Copies the result.
- Pastes it into another system.
- Emails somebody.
- Updates the original system.
We’ve added AI. But have we actually modernized anything?
The better model is to make AI a controlled participant inside the workflow.
- The process determines what information AI receives.
- AI performs the task it’s suited to.
- Rules determine what happens with the output.
- People remain involved where judgment matters.
- And the organization retains an audit trail of what happened.
That’s much closer to how AI creates meaningful operational leverage in a regulated environment.
The Productivity Question
Again, because it’s hugely significant:
The Bank of Canada recently found that more than two-thirds of Canadian business leaders personally use AI during a typical work week. Yet only 8% of businesses report using AI significantly in their core operations.
Individual AI use can produce productivity improvements, but the larger economic effects are more likely to emerge as businesses integrate AI more deeply into how they actually operate.
That might be the bigger opportunity for credit unions too.
Not another chatbot.
Not another Copilot.
Not AI everywhere simply because we can.
But carefully identifying the workflows where AI can remove administrative effort, accelerate analysis and give employees better information while leaving accountability and important judgment exactly where they belong.
Where to Start: Finding Inefficient Workflows
Which workflows would become materially better if AI participated in them?
Then take one. Map it. Identify where people spend time gathering, reading, comparing, summarizing and preparing information. Identify where rules should remain deterministic. Identify where human judgment must remain. Put AI into the gaps where it genuinely improves the process.
That might be a less exciting AI strategy than announcing an army of autonomous agents.
But in financial services, it may prove much more valuable.
Sources & further reading
Bank of Canada, Financial System Survey Highlights 2026, May 28, 2026.
Bank of Canada, Canadian Businesses’ Use of AI: What the Evidence Shows, August 24, 2026.
Bank of Canada, Canadian Survey of Consumer Expectations, Q1 2026, April 2026.
Author Profile:
Moray Hickes is an Enterprise Technology Solutions leader with over two decades of experience helping organizations navigate complex technology systems and deliver high-value solutions across regulated industries.