Most tech companies are leaving significant money on the table. Despite building AI products for their customers, their own finance teams still operate with siloed data and piecemeal automation - what EY calls "Robotic Process Automation 2.0." A shift to agentic AI, where intelligent workflows handle end-to-end processes rather than isolated tasks, can cut finance operating costs by up to 40% and free analysts for higher-value work, according to Amanda Donohue, Principal, Consulting, at Ernst & Young LLP.
The core problem isn't a lack of tools. It's that many finance functions deploy point solutions for narrow use cases like invoice extraction or anomaly detection without redesigning the underlying operating model. "A good deal of AI in finance today is simple automation of parts of the process, not the whole," the EY team said. Moving beyond that requires what they describe as a "design for zero" approach - reimagining processes to achieve outcomes without human intervention, then adding people back only where absolutely necessary.
Redesigning finance from the ground up
Adam Blaylock, EY Americas Financial Accounting Advisory Services TMT Industry and Technology Sector Leader, said the challenge is not vision but execution. "Most tech leaders today can tell you what the future of AI in finance looks like. The challenge is getting from today to tomorrow in a thoughtful, value-driven way, because building a reimagined finance function isn't a simple task."
Agentic AI splits finance work into two categories: routine operational tasks managed by AI with human oversight, and higher-level managed tasks handled almost entirely by people. For an FP&A analyst, this means data consolidation and first-draft commentary arrive automatically. The analyst no longer pulls figures from multiple systems manually. Instead, they review the AI's output and focus on deeper insights, partnering with the business on critical decisions. "When we shift our thinking from eliminating single tasks or processes to designing collaborative outcomes, AI delivers significantly more benefits," Blaylock said.
The entire lead-to-cash process illustrates what this looks like at scale. Agentic AI uses predictive analytics to understand customer preferences, payment risks, and renewal patterns. It automates contract management, credit decisions, invoicing, and collections. It personalizes communication with prospects and customers. The result is a highly automated process that increases sales, manages complex contracts, reduces nonpayment risk, and improves cash flow.
Build vs. buy: where tech firms should focus
Finance leaders face a persistent question: when to use AI embedded in enterprise resource planning software and when to build internal tools. EY professionals believe tech companies should prioritize internal development around order-to-cash and FP&A first. These processes are what make a tech company unique in the marketplace, and off-the-shelf AI solutions rarely fit. For more standardized processes like procure-to-pay and record-to-report, it often makes sense to wait until software providers embed AI into their products.
"There is no 'one size fits all' solution and companies need to take a deep look at their internal processes and supporting tech before moving forward," Donohue said. "The key is to invest in reimagining your critical and complex processes rather than waiting for someone else to reimagine them for you." This decision shapes how quickly a finance team can move from incremental automation to full transformation. For professionals exploring how AI reshapes financial analysis, risk management, and accounting, structured AI for Finance resources can provide a practical starting point.
Compliance and the accuracy imperative
Regulatory ambiguity remains a real obstacle. While internal accuracy can be vetted, regulators are still catching up to the pace of AI adoption. Many tech firms need certainty around internal and external reporting before they can move forward with AI-driven finance operations. Blaylock said that when properly designed and implemented, agentic AI actually improves financial data accuracy because it removes human error and allows more time for review and analysis.
A growing number of clients are shifting toward managed service models, supported by global pools of analysts, engineers, and accounting professionals with access to industry and proprietary tools. Organizations also increasingly need AI validation tools to monitor proprietary model performance. As employees create automated processes with AI-generated code, there is no guarantee those tools deliver accurate data or responses without independent verification.
For CFOs and senior finance leaders building their AI strategy, AI Learning Path for CFOs offers guidance on financial forecasting, automation, and the structural decisions that accompany an agentic AI shift.
Why this matters for finance professionals
The choice is no longer whether to use AI in finance but how quickly to move from piecemeal automation to end-to-end transformation. An AI-first delivery model paired with an outcomes-based strategy can reshape the finance function within months, not years. The three hurdles that delay most teams - tool selection, workforce readiness, and ERP integration - all point to the same underlying need: deciding how work will be divided between AI agents and people, and what the finance function structure looks like once that division is in place. For finance professionals, the practical takeaway is that waiting for software vendors to solve unique, complex processes means ceding the efficiency gains to competitors who build first.
Your membership also unlocks:
