Can finance trust AI to act?
Reflections from Gartner® Finance Symposium/XPO™ on the move from automated to autonomous AP
By Daniel Ball, SVP Global Analyst Relations, Medius
At this year's Gartner® Finance Symposium/XPO™ in London, I had the opportunity to speak with CFOs, finance leaders, and transformation teams from organisations at various stages of their AI journey.
What struck me wasn't the excitement around AI. That part was expected. What stood out was how quickly the conversation evolved. A year ago, finance leaders were asking whether AI could automate invoice processing, approvals, or supplier communications. Today, most assume it can.
Instead, they're asking different questions: How far can we trust AI to make decisions and take action? What safeguards do we need in place to make us comfortable allowing decisions to run autonomously?
That's a fundamentally different conversation, and one that will define the next phase of finance transformation. At Medius, we see this evolution happening most clearly within accounts payable.
The automation conversation is maturing, and control starts in AP
For more than a decade, finance organisations have used some sort of automation. The results have been meaningful, too. Processes that once required hours of manual effort can now be completed in minutes. And teams are working more efficiently, reducing manual errors, and getting better visibility across operations.
Accounts payable has been at the center of much of that progress, and for good reason. Financial control starts in AP.
Invoice capture, matching, approvals, and payment workflows have all become increasingly automated. Yet despite these advances, many AP teams still spend considerable time managing exceptions, responding to supplier inquiries, investigating discrepancies, and ensuring compliance with policies. Our own research shows that 87% of AP professionals must respond to vendor emails, handling an average of 35 emails per day, work that can consume up to 25 hours a week.
The process may be automated, but people are still orchestrating the work. That's where the next evolution begins.
Automation executes tasks. Autonomy acts.
At Medius, we see the evolution of AI in finance as a progression from traditional automation to autonomous systems.
Traditional automation follows set instructions. When conditions change, the process often stops and waits for a person to step in. Autonomous systems are different. They understand context, evaluate options, take action within defined policies, and only escalate when necessary.
Today, 54% of enterprises report that automation is limited to specific tasks. By 2028, 27% expect to operate primarily without human intervention, with AI systems making decisions and acting independently.
Gartner, 2026 Gartner CEO & Senior Business Executive Survey, Gartner Research, conducted March–November 2025, published 2026.
Consider a common exception: an invoice arrives without a valid purchase order number. This usually triggers a series of manual tasks like emails and follow-ups, and resolving the issue can take days. An autonomous approach can identify the problem, request the missing information, validate the response, and automatically return the invoice to the workflow. Human intervention is only needed when the risk or ambiguity exceeds set limits.
We're seeing similar advancements in governance and risk management. Traditionally, issues like supplier data changes or potential fraud were handled through periodic reviews. Autonomous agents can now continuously monitor activity, enforce policies, and escalate issues before transactions are even executed.
Autonomous capabilities also extend to payment optimization. AI can analyze unpaid invoices, identify early-payment discount windows, evaluate cash-flow impact, and recommend payment timing, all while operating within existing financial controls.
The goal isn't just faster processing. It's about work moving forward without constantly waiting for human intervention. And in the process making things more secure, more consistent, more accountable, more proactive. In short, it’s about more intelligent accounts payable.
Trust is the gatekeeper to autonomous finance
Of course, none of this matters if organizations don't trust the technology.
While much of the market conversation focuses on AI capabilities, we believe the more important discussion is about governance.
Trust—not technology—is the true gatekeeper of autonomous finance.
Finance leaders need confidence that autonomous systems will operate consistently, transparently, and within defined policy boundaries.
At Medius, we believe autonomous finance requires several foundational elements:
Explainable decision-making with clear audit trails
Policy guardrails that define what AI can and cannot do
Human oversight for high-risk or unusual situations
Continuous monitoring for anomalies and behavioral drift
Full override capabilities when intervention is required
Compliance controls that align with existing regulatory standards
The organizations making the greatest progress with AI are not eliminating governance. They're embedding governance directly into how AI operates.
Context matters more than data
Another theme that recurred in discussions at the event was the importance of context.
Finance teams have no shortage of data. What they need is the ability to understand relationships across that data.
An invoice doesn't exist in isolation. It connects to suppliers, contracts, purchase orders, payment terms, approval hierarchies, historical transactions, and organizational policies.
For autonomous systems to make reliable decisions, they need access to that broader context. Without context, AI can automate individual tasks. With context, AI can make informed decisions. That's an important distinction as organizations evaluate where autonomous capabilities can create meaningful business value.
From managing exceptions to owning outcomes
Today, many AI applications in finance focus on discrete tasks. An AI model extracts invoice data. Another recommends coding. Another identifies anomalies or flags risks.
These capabilities create value, but they operate within narrow boundaries.
The next stage is the emergence of autonomous agents that coordinate activities across workflows and execute outcomes rather than individual tasks.
From intelligent automation to autonomous finance
agents
Enterprise-grade confidence
Tasks executed automatically
Policies applied step-by-step
People run the process
agents
Increasing autonomy
Work completes across workflows
Decisions execute within policy
People handle exceptions
agents
Outcome ownership
Cross-domain financial supervision
Predictive governance
Working-capital optimization
Over time, these capabilities will evolve further. Instead of task-specific agents, finance organizations will increasingly rely on role-based agents that oversee broader responsibilities and business outcomes.
The shift is subtle but significant. We're moving from systems that help people complete work to systems that help ensure work gets completed.
The question finance leaders should ask next
One of the most encouraging developments in today's AI conversation is that finance leaders are approaching autonomy thoughtfully.
People are looking beyond automation to understand where autonomy creates value, what oversight needs to be in place to deliver that value safely, and how that Venn diagram can create transformational organisational returns.
At Medius, we believe the future of finance won't be defined by how many tasks organizations automate. It will be defined by how confidently finance leaders can delegate routine decisions to intelligent systems operating within trusted controls.
GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.
