Accounts payable (AP) has spent the better part of two decades focused on automation. Organizations have invested heavily in eliminating paper, reducing manual data entry, accelerating invoice processing, and creating workflows that can move transactions from receipt to payment with fewer human touches. During Ardent Partners’ recent webcast “From AP Automation to AP Intelligence: How AI Is Reshaping Accounts Payable,” featuring Andrew Bartolini, Founder and Chief Research Officer for Ardent Partners, and Graeme Chard, SVP of Product Marketing for Medius, our industry experts discussed how those investments have delivered meaningful improvements, yet the automation journey is far from complete.
This is Part One of a three-part series exploring how many AP processes remain partially automated or manual, particularly when transactions fall outside predictable rules. The emergence of artificial intelligence is now creating an opportunity to take AP beyond simply processing transactions faster and toward understanding what is happening across the entire function.
Automation in an Unpredictable Environment
This represents an important shift in the role of AP. Traditional automation works best when processes are predictable. Organizations establish rules, configure workflows, and allow technology to execute those activities consistently. That model has proven highly effective for structured transactions, particularly purchase order-based invoices and other activities where the required information is standardized. Chard explained that the challenge is that AP does not operate entirely within predictable boundaries. New suppliers emerge, invoices contain inconsistent information, exceptions require investigation, supplier information changes, and transactions frequently require context that cannot be captured through a simple rule.
AI can begin to address those limitations by expanding the boundaries of what AP can automate. Chard said that instead of replacing existing automation, AI can help organizations apply intelligence to transactions that have historically required human intervention. It can interpret less structured information, identify relationships between data points, support matching and coding decisions, and provide recommendations that help employees resolve exceptions more quickly.
The goal is not simply to create a new form of autonomous AP. It is to increase the number of transactions that can move through the process with minimal human intervention while improving the quality of decisions along the way.
A Deliberate Approach for Measurable Results
The market is already moving in this direction, Bartolini noted. Research from Ardent Partners found that 58% of AP organizations are either actively using or piloting AI. That adoption is significant, with organizations generally approaching AI deliberately rather than attempting to transform the entire function overnight. Productivity and task automation remain the leading objectives, followed by efficiency improvements and better forecasting, shared Bartolini. This suggests that most organizations are beginning with practical use cases where AI can deliver measurable improvements before expanding into more sophisticated applications.
Invoice processing provides a useful example. Traditional AI and machine learning have already helped organizations achieve high levels of straight-through processing for structured, PO-based invoices. Chard said generative AI and large language models can extend those capabilities by helping employees address the transactions that do not fit established patterns. AI can identify possible matches, interpret different descriptions for the same goods or services, recommend coding, and provide relevant context to approvers. Rather than forcing employees to search across contracts, purchase orders, invoices, and email conversations, Chard said AI can bring the relevant information together at the point of decision.
This evolution also changes how AP should think about performance. For years, Bartolini shared, metrics such as cost per invoice, processing time, and touchless rates have been central measures of AP efficiency. Those metrics will remain important, but intelligence introduces a broader set of possibilities. AP can increasingly help the organization understand payment behavior, supplier activity, exceptions, working capital opportunities, and emerging risks.
The transition from automation to intelligence therefore represents more than a technology upgrade. Chard explained that it is an opportunity to elevate AP from a transactional processing function into a more informed business partner. Automation helped AP work faster and more efficiently. AI can help AP understand more, make better decisions, and ultimately contribute more strategic value to the enterprise.
