- Over 40% of agentic AI projects will be canceled by the end of 2027, notably due to inadequate risk controls, according to Gartner (2025).
- Gartner predicts at least 15% of day-to-day work decisions will be made autonomously by agentic AI in 2028, up from 0% in 2024.
- According to Statistics Canada (Q2 2026), 19.2% of Canadian businesses use AI, and the rate reaches 40.4% in finance and insurance.
- PlanAxion frames every AI agent in finance with four controls: a written scope of action, human approval thresholds, an audit trail and a review of exceptions.
Month-end Friday at a distributor. Overnight, an AI agent matched 312 payments, posted the entries and closed the batches. Monday morning, the controller asks a simple question: who approved all of this?
Nobody has a clear answer. The agent works, the numbers balance, but no document states what it is allowed to do on its own. That is exactly the gap AI agent governance exists to close.
“Over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls.” Source: Gartner, June 2025
What is AI agent governance?
AI agent governance is the set of rules that defines what an agent may do on its own, what requires human approval, and how every action stays traceable. An AI agent no longer just suggests: it acts. It reads a remittance advice, matches a payment, posts an entry.
The difference from classic automation is autonomy. A bot follows a script. An agent pursues a goal and picks its own steps. Quadient describes this agentic automation as systems acting independently within defined rules. The whole discipline of governance lives in those two last words: defined rules.
Gartner predicts that at least 15% of day-to-day work decisions will be made autonomously by agentic AI in 2028, up from 0% in 2024. The question is no longer whether your systems will decide on their own, but within what frame.
Why are your finance processes on the front line?
Because finance is already the most AI-advanced sector in the country, and agents there write directly into the ERP. According to Statistics Canada, 19.2% of Canadian businesses used AI in the second quarter of 2026, a rate that has tripled in two years. Finance and insurance reach 40.4%.
Concretely, vendors are embedding agents in the financial modules: cash application, reconciliations, close. Our analysis of agentic AI in the ERP details what SAP, Microsoft and Oracle are shipping in 2026.
An agent that drafts text is easy to correct. An agent that posts to the general ledger commits your financial statements. The bar cannot be the same.
What controls should you require before an AI agent acts in your ERP?
Four controls are enough to start: a written scope of action, human approval thresholds, a complete audit trail and a periodic review of exceptions.
Scope first. One page naming the permitted transactions, the maximum amounts and the systems touched. What the page does not name, the agent does not do.
Thresholds next. A payment matched with high confidence under a defined amount posts on its own; above the threshold, or in case of doubt, the agent prepares and a human approves. It is the same principle as a signing authority.
The audit trail, finally, answers the Monday morning question: who did what, when, and on what basis. Without it, neither your auditor nor your controller can trust the numbers.
The benchmarks to keep in mind for 2026:
- Over 40% of agentic AI projects will be canceled by the end of 2027, notably for inadequate risk controls (Gartner, 2025)
- At least 15% of day-to-day work decisions will be made autonomously by agentic AI in 2028, up from 0% in 2024 (Gartner, 2025)
- 33% of enterprise software applications will include agentic AI by 2028, up from less than 1% in 2024 (Gartner, 2025)
- 19.2% of Canadian businesses use AI, and the rate reaches 40.4% in finance and insurance (Statistics Canada, Q2 2026)
- Cybersecurity and confidentiality concerns are the most reported barrier limiting AI use (Statistics Canada, Q2 2026)
A note on method: these figures come from public studies by Gartner and Statistics Canada. Scopes vary from one study to another and none of them describes your specific situation.
An AI agent without an approval threshold is not automation: it is a signing authority nobody signed.
Should you freeze AI agents until the framework is perfect?
No. A freeze pushes teams toward ungoverned tools, and the market moves on without you. The pattern is documented: our work on shadow AI shows that usage almost always precedes policy.
Watch the label too. Gartner calls it “agent washing”: vendors rebrand existing assistants or RPA as “agents.” The firm estimates only about 130 vendors, out of thousands, offer real agentic capabilities. Requiring a demonstration of the controls is the fastest test.
At PlanAxion, we apply the same filter to every use case: problem to solve, expected value, available data, required effort. An agent that fails this filter does not need governance. It does not need to exist.
Where should you start to govern your AI agents?
With one bounded finance process, a written scope and thresholds validated by the finance leadership, not with a fifty-page theoretical framework. Pick a measurable process, document the baseline, define the four controls, then expand.
That is the approach of our AI workshop: 4 weeks, 5 steps (prepare, identify, prioritize, validate the data, decide and deliver), with governance built in at the prioritization stage. The investment varies with scope and is confirmed during a short exploratory call.
Frequently asked questions about AI agent governance
What is AI agent governance?
AI agent governance is the set of rules framing AI systems able to act on their own: a permitted scope of action, thresholds above which a human approves, logging of every action and a regular review of exceptions. It extends to agents the internal controls already applied to employees holding delegated authority.
What decisions can an AI agent make without human approval?
The ones your written scope allows. In finance, a high-confidence payment match under a capped amount is the typical example. Anything touching an amount threshold, an exception or an unusual entry should escalate to a human. Gartner predicts 15% of day-to-day work decisions will be autonomous by 2028.
What is agent washing?
Agent washing is the rebranding of existing products, assistants, chatbots or RPA, as “AI agents” without substantial agentic capabilities. Gartner estimates only about 130 vendors, out of thousands, offer real agent capabilities. Asking for a demonstration of approval thresholds and the audit trail remains the most revealing test.
Why are agentic AI projects being canceled?
Gartner identifies three main causes: escalating costs, unclear business value and inadequate risk controls. The firm predicts that over 40% of agentic AI projects will be canceled by the end of 2027. A use case that is bounded, measurable and governed from day one reduces each of those risks.
