CFOs and finance leaders will have to transform their teams to better prepare for a future where artificial intelligence (AI) agents are a fundamental part of their organisation.
This also involves managing a hybrid workforce where both employees and AI agents work together, said Tim Wakeford, vice president of product management for Workday.
Organisations will have to create an agent “system of record”, to ensure each agent is given access only to the data and systems it needs to fulfil its role, Wakeford said in an interview with FutureCFO, at the sidelines of the Finance Think Tank event in Singapore, hosted by FutureCFO and Workday.
“You would never give every employee access to the company’s payroll data, so why would this be different for an AI agent,” he said.
Organisations, including CFOs, will want to maintain control and the ability to manage this based on their company’s policies, he added, pointing to Workday’s Agent System of Record.
Announced in February, the AI platform is touted to enable organisations to manage their AI agents, including governing, managing, and optimising the ROI of these agents.
It provides a way to onboard new AI agents, such as defining their roles and responsibilities, and tracking their impact, Workday said.
It added that the centralised system gives IT and business leaders “transparency and control” over the impact of AI agents on work.
The Agent System of Record is able to monitor agent activities, enforce policies, and track costs in real-time, so the performance and ROI (returns on investment) of AI agents can be optimised, according to Workday.
The software vendor supports 75 million customers globally, processing some 1 trillion transactions a year.
AI agents will have personas that can be aligned with roles that exist within an organisation.
Enterprises need to understand what these roles are as they transition towards an agentic AI environment, said Benjamin Patricks, Accenture’s Asia-Pacific Workday financials and adaptive insights lead.
It makes agents more relatable to what they actually do for the organisation, Patricks explained. It also helps define the roles of AI agents within the organisation, much like a job description, he added.
CFOs then can better determine where an agent will make a better fit and where an employee will be better placed, and what actions each agent should take, he said.
Ability to understand context, decide autonomously
It is important that organisations also measure the value AI agents generate and ensure there is ROI, Wakeford said.
Agentic AI will have a profound impact on all parts of the business, especially finance, he noted.
While the technology still is nascent and continues to make advancements today, the ability of AI agents to automate with reasoning and autonomy will profoundly change how finance operates, he said.
It is a fundamental shift from RPA (robotic process automation) and traditional AI, where these systems are built to carry out pre-defined activities and draw predictions from historical data.
AI agents can potentially unlock significant business value with their ability to digest data, understand context, and then take action to drive a business outcome, Wakeford said.
CFOs want insights that drive sustainable growth, which means they need accurate real-time analytics to drive decisions that fuel growth, he noted.
At the same time, they look for transformation that can power operational excellence as well as resilience to manage risks. This will enable them to quickly adapt without sacrificing accounting controls whilst seizing market opportunities, he said.
Achieving these goals requires a robust AI strategy that encompasses building the best agents to support employees and that are embedded in functions tied to real business outcomes, Wakeford said.
By taking over basic tasks, AI agents allow employees to be precise and focused, so they can manage business outcomes, said Lee Eng Khan, managing partner at audit and accounting firm, PKF-CAP.
AI agents can further guide employees to make better business decisions, Lee said during a panel discussion at the event.
For instance, agents can review draft service contracts and generate a risk impact analysis, which then can provide employees advice on terms that should or should not be included in the contract, he said.
Ultimately, humans still are the ones making the final decision, Wakeford stressed.
AI agents perform their assigned tasks, including evaluating contracts and providing recommendations on next actions to take, and then route these to the relevant employees to check and approve the final action.
“What we always do is bake that human into the loop, so it’s the human approving the transaction,” he said. “Your approval process is far faster, but it is still a human in the loop and approval process. So you are outsourcing the effort, but retaining the intellectual control over each transaction.”
Need for business redesign
Organisations also should take the time to understand where and how they can adopt AI agents most effectively.
There is a general misconception that any new technology will be the panacea for any business problem, Wakeford said.
Organisations cannot simply deploy an AI agent to perform any task, without first assessing whether it is feasible to do so, he said.
“There will always be some things that need people [to manage] and business process that needs to be redesigned,” he said. “It’s not just an AI agent solution to everything. That’s probably the biggest misconception.”
Patricks further noted that AI and agentic AI will become so integral in everything an organisation does, that it will require a major structural change.
“It isn’t just a piece of technology,” he said, adding that businesses will have to review various key functions.
“For example, who should manage AI agents? Is it IT or HR, or will it be a separate part of the organisation that will deal with agents?” he posed.
Organisations will have to look at the fundamentals they have in place, including their data and cloud platforms, and ensure their underlying infrastructure and technology are ready to support AI and agentic AI, Patricks said.
