Trending Topics
Trending
Jeff Chancellor, CMO at Oversight: “In finance, the real advantage comes from the context around the model”
Finance has spent years automating the mechanics of getting money, transactions and processes from A to B. But as those systems have become faster and more connected, another problem has emerged: knowing what is happening across them, and whether it should be happening at all.
For Jeff Chancellor, CMO at Oversight, that is the next gap finance technology needs to address. Automation can accelerate transactions and remove manual work, but it can also allow mistakes, control failures and potentially risky activity to move through an organisation at the same speed. Traditional periodic reviews and rules-based controls were not designed for an environment operating at that volume and velocity.
Chancellor argues that this is where Finance Risk Intelligence (FRI) can play a role, bringing activity from systems including ERP, procurement, payments, expenses and accounts payable together to provide a more continuous view of financial risk. But he is equally clear that simply adding AI to finance processes is not enough. As foundation models become increasingly accessible, he believes the lasting advantage will come from domain knowledge, historical outcomes and the context needed to distinguish genuine risk from normal business activity.
That distinction becomes particularly important as enterprises move towards agentic AI. Chancellor argues that autonomy should be earned rather than assumed, with AI given greater freedom only when decisions are repeatable, evidence is clear and the appropriate controls are in place.
In this interview, Chancellor discusses the gap between finance automation and financial intelligence, what FRI adds to traditional analytics and controls, where agentic AI can deliver genuine value, and how finance leaders can separate meaningful AI capabilities from another layer of automation. He also looks ahead to an operating model in which technology handles scale, intelligence focuses attention and people apply judgment where it matters most.
Finance has spent much of the past two decades automating transactions, standardizing processes, and making operations more efficient. What did the industry fail to modernize along the way?
We’ve gotten very good at making finance move faster. We’ve automated transactions, connected systems, and reduced a great deal of manual work. But we have not made the same progress in understanding the risks moving through those systems.
That matters because speed cuts both ways. If something goes wrong, such as a control gap, an unusual transaction, or a simple mistake, it can travel through the business just as quickly. A periodic review was never built to keep up with that kind of volume and velocity.
So, I think the next phase of finance transformation is less about asking, “What else can we automate?” and more about asking, “Can we see what is happening clearly enough to know where we need to pay attention, and can we respond before there is a loss?”
We’ve modernized the work faster than we’ve modernized the judgment around the work. That is the gap finance has to close next.
Everest Group describes a widening “risk gap” and identifies Finance Risk Intelligence (FRI) as an emerging response. What makes FRI fundamentally different from traditional finance analytics, audit, or controls?
Those tools all play an important role. Analytics can explain what happened; rules-based controls catch risks you already know to look for; and audit provides assurance and human judgment.
What Finance Risk Intelligence (FRI) adds is a more continuous view of what is happening across the finance environment and the context to make sense of it.
Instead of looking at ERP, expense, procurement, payments, cards, and accounts payable in isolation, FRI can bring activity across those systems together. That makes it possible to spot patterns or behavior as they emerge, determine which warrant attention, and then connect that insight to the appropriate response.
For me, that is the simplest distinction. Reporting tells you what happened. Controls check for what you already know to look for. Intelligence helps you decide what matters now and what to do about it. That is why I do not think of FRI as just another finance application. It is part of a broader shift in how finance operates, from automating work to understanding it to acting on that understanding.
Foundation models are becoming increasingly accessible to every software company. If the underlying models become more commoditized, where does durable advantage come from in financial AI?
I don’t think the foundation model itself will be the lasting differentiator. In finance, the real advantage comes from the context around the model, including what the system has learned from actual decisions and outcomes over time. Risk is highly situational. A transaction that is completely normal for one employee, vendor, or business unit may be a red flag somewhere else. A general-purpose model may recognize a pattern, but that does not mean it understands whether the pattern represents meaningful financial risk.
That is where domain experience and feedback matter. At Oversight, our intelligence draws on more than two decades of finance-risk experience, labeled outcomes, and billions of enterprise transaction signals. The value is not just having a large amount of data. It is knowing what happened next. Was the activity actually risky? Was it acceptable? How was it resolved? Which action worked?
Over time, those outcomes become institutional knowledge. As the underlying models become easier for everyone to access, I think that knowledge becomes more valuable, not less.
The question I would ask is: what does this system know about my financial environment that a general-purpose model would not know?
Agentic AI has become one of enterprise technology’s most heavily promoted ideas. Where should finance leaders be genuinely excited, and where should they remain cautious?
There is real potential, but I wouldn’t equate autonomy with intelligence.
An agent that can execute a workflow is useful. An agent that understands why an action is appropriate, operates within defined controls, documents its actions, and knows when not to act is considerably more valuable.
That distinction matters in finance because the consequences vary. Following up on a missing receipt is very different from taking action involving a material payment, suspected fraud, or conflicting evidence. We should not treat those decisions as if they carry the same level of risk.
I tend to think autonomy has to be earned. When a decision is repeatable, the evidence is clear, and confidence is high, you can give the technology more autonomy. When the situation is ambiguous or the financial consequences are greater, human judgment needs to be more involved.
The objective is not autonomous finance for its own sake. It is to remove repetitive work while preserving accountability, explainability, and control.
If every enterprise software provider now says it has AI, how should finance leaders distinguish genuine intelligence from another layer of automation?
I would look at a few things.
First, what does the system actually know? Does it have finance-domain knowledge, historical context, and evidence that it can distinguish meaningful risk from noise?
Second, what happens once the AI finds something? If the end result is just another alert or dashboard, you may have added technology without solving much. The more useful question is whether the intelligence can lead to an appropriate, governed action, with clear escalation when a person needs to step in.
And third, can you measure whether it made a difference? Did it prevent leakage? Reduce manual review? Resolve issues faster? Cut false positives? Improve the quality of decisions? Saying “we added AI” does not produce a business outcome.
Enterprise readiness matters, too. It includes explainability, auditability, security, integration, and data governance. In finance, these are not features you bolt on later. They are part of whether the technology is usable and trustworthy in the first place.
A CFO can get pretty far by asking a vendor three simple questions: What does your AI know? Why did it reach that conclusion? What happens next?
Oversight is beginning to connect Finance Risk Intelligence with agentic action. How do you prevent that from becoming the same “agent washing” we are seeing elsewhere in enterprise technology?
Start with the intelligence, not the agent.
An agent shouldn’t wander through a financial environment looking for something to do. The intelligence should first identify a specific issue and explain why it matters. Then, if an agent is the right way to handle it, give that agent a bounded task and the policies, thresholds, exclusions, and escalation rules it needs.
The order is important: understand the issue, make the decision, then act.
That is also why I would not define agentic AI by whether software can string together several steps. That is a technical capability. For an enterprise finance team, the bar is higher. Was the action appropriate? Can you explain it? Does the organization remain in control? Are the results improving over time?
The meaningful measure is not how many agents an organization can deploy. It is whether the organization can trust the actions those agents take.
Looking three to five years ahead, how do you think AI will change the operating model for enterprise finance?
I think the biggest change will be in how people spend their time.
For years, the core of the finance technology stack was the system of record. ERP and other financial applications gave companies a structured way to capture transactions and establish what happened.
We are now adding a system of intelligence on top of that. AI can analyze activity, recognize patterns, and surface items that deserve attention without requiring people to review everything themselves.
The next step is governed action. For routine, well-understood situations, the intelligence will increasingly be able to trigger or carry out a response. People will spend more of their time on cases where judgment, consequence, or business context really matter.
That does not mean ERP goes away, nor does it mean finance professionals go away. It means their expertise is applied differently.
The most successful finance organizations will not simply automate more work. They will establish an operating model in which technology handles scale, intelligence focuses attention, and people apply judgment where it matters most.
Recommended reads:
About The Author
Ricardo Oliveira
NEXT UP
