Insights from a live expert session with Mark D. McDonald, former Gartner research leader on AI in finance
AI agents are no longer theoretical.
Finance teams are already using them to retrieve data, trigger actions, and support decisions that were previously handled by people. What started as experimentation is quickly becoming embedded in day-to-day workflows.
The challenge is not adoption. It is understanding what changes when financial work is executed by systems that operate at speed, across processes, and with limited visibility.
This live expert session with Mark D. McDonald explored that shift in detail, focusing on how AI agents actually work, where risk begins to surface, and what finance leaders need to do now to maintain control.
If you are evaluating or deploying AI in finance, this is a conversation worth exploring in more detail.
Session Host
Mark D. McDonald is the founder of Finance Next and a former Gartner research leader focused on AI in finance. He has advised finance leaders on the adoption of AI and advanced analytics, helping organizations develop practical roadmaps for implementation, governance, and long-term value creation.
With a background spanning finance leadership, technology, and operations, including nearly two decades at Siemens, Mark brings a practical perspective on how finance teams can responsibly scale AI while maintaining control and performance.
What Are AI Agents and Why They Change Finance Work
One of the most important clarifications from the session is that AI agents are not a new type of intelligence. They are a new way of organizing work.
An AI agent is best understood as a program designed to execute a specific task. What makes it different is how it operates within a broader system. Agents can access enterprise data, interact with other systems, and coordinate workflows without direct human intervention.
This creates a step change from traditional automation.
Instead of predefined processes, finance is moving toward dynamic systems where tasks are delegated, executed, and completed across multiple interconnected components.
That shift introduces new capabilities, but also new dependencies. When execution becomes distributed across systems, understanding how decisions are made becomes more difficult, and control models built for linear workflows start to break down.
AI Agent Risks in Finance: Where They Show Up
A key theme throughout the session is that AI risk does not present itself as a single, obvious failure point.
It shows up through operational consequences.
- Errors can propagate across workflows rather than remaining isolated.
- Gaps in auditability make it harder to trace how outputs were generated.
- Regulatory pressure increases as organizations struggle to explain system-driven decisions.
These are not hypothetical concerns. They reflect what finance teams are already encountering as adoption accelerates.
Importantly, these risks do not stem from a single failure. They can occur at multiple points across the system, from incorrect task routing to breakdowns in multi-step workflows.
This is what makes them difficult to manage using traditional approaches. The issue is not a single control failure. It is a mismatch between how work is executed and how it is governed.
Why Governance Is Falling Behind AI Adoption in Finance
Adoption is moving faster than oversight.
Research shared in the session highlights the scale of this gap. A majority of organizations are already using AI in finance, and most expect AI agents to become standard tools within the next few years.
At the same time, governance frameworks are still evolving. In the US, regulatory bodies are largely deferring to existing control structures. In other regions, guidance is emerging but remains focused on specific use cases rather than system-wide oversight.
The result is a growing disconnect.
Finance teams are scaling automation and AI-driven execution, but oversight models remain periodic, sample-based, and dependent on the systems generating the activity. That model was not designed for environments where execution is continuous and system-driven.
This is the underlying issue that surfaced throughout the discussion. What many teams experience as errors, audit gaps, or regulatory pressure is the visible impact of a deeper governance gap.
A Practical Framework for Managing AI Agent Risk
Rather than treating all AI agents the same, the session introduced a structured way to assess and manage risk.
The framework evaluates agents across two dimensions: task complexity and risk exposure. Based on these factors, agents fall into four categories ranging from simple clerical tasks to high-impact, consequential decision-making.
Each category requires a different level of control.
Lower-risk agents may require monitoring and basic validation. Higher-risk agents require stronger oversight, including audit trails, human review, and clear accountability structures.
This approach allows finance teams to prioritize where controls are most needed, rather than applying a one-size-fits-all model.
It also reinforces an important point. Simplicity does not mean safety. Some of the highest-risk scenarios involve routine tasks with direct financial impact.
What Finance Leaders Should Do Next
The takeaway is not to slow down adoption.
AI agents are already in use, often without formal oversight. Ignoring them does not reduce risk. It increases it.
The more practical path is to bring visibility and structure to what is already happening.
That starts with identifying where agents are being used, understanding their role in financial processes, and assessing their risk profile. From there, organizations can begin to align controls with how work is actually executed.
As execution scales, teams can extend oversight beyond periodic reviews, using platforms like MindBridge to continuously analyze financial activity, provide independent visibility across transactions, and surface risk as it emerges.
Watch the Full Session
This recap highlights the key themes, but the full session goes deeper into how AI agents operate, how risks emerge across workflows, and how finance teams can apply a structured governance approach.
If AI is entering your finance environment, these are not future considerations. They are immediate ones.
Watch the full session to see how leading finance teams are approaching this shift and what it takes to scale AI without losing control.
FAQ: AI Agents in Finance
AI agents in finance are software programs that execute tasks such as data retrieval, transaction processing, and decision support across financial workflows with limited human intervention.
AI agents can introduce risks such as error propagation across processes, gaps in auditability, lack of explainability, and increased regulatory pressure.
Traditional automation follows predefined rules within fixed processes, while AI agents can interact across systems, make decisions, and execute tasks dynamically.
Governance is challenging because AI agents operate across multiple systems at speed, making it difficult to apply traditional controls designed for linear, human-driven workflows.
Finance teams can manage AI agent risk by identifying where agents are used, assessing risk levels, and aligning controls to how work is executed. As automation scales, many teams also implement continuous monitoring across financial activity using platforms like MindBridge to maintain visibility and detect risk earlier.