The rise of searching for hard ROI in the CFO role

The adoption race is over. The ROI race just started.

In a June 2025 survey of 183 CFOs, Gartner reported that 84% of finance organisations had deployed AI or planned to, but only 7% reported a high impact. Most AI spend results in increased productivity, efficiency and time back, rather than hard ROI.

Gartner surveyed 204 finance leaders in March 2026 and found 45% of finance AI investment leans toward productivity, and 20% toward decision quality. Shankar Keshav, the analyst who conducted the survey, called this result a perception gap, where “finance leaders report progress on AI adoption, but boards see limited strategic impact.”

Hard ROI must be grounded in cash, whether that is revenue growth, cost savings, or more cash on hand faster. Saving time is meaningful to a degree, but finance leaders must stop mistaking AI deployment for business value creation, and find solutions that can deliver more cash to the business.

Accounts receivable (AR) delivers hard ROI

The Federal Reserve counts $9.8 trillion sitting in trade receivables in the United States, as of the first quarter of 2026. Allianz Research put the average company’s days sales outstanding at 59 days in 2023, across roughly 45,000 listed companies in more than 35 countries. Atradius found 43% of credit-based B2B sales in North America were overdue in 2025.

Accounts receivable is the most acute and universal financial pain in B2B today. AR has been too complex to solve pre-LLMs, and the margin for error is zero. As a result, companies are scared to automate it, but it is common enough that every finance team we speak to is quietly buried in an AR headache.

When it is solved correctly, businesses get paid fast, and the value of AI is clear.

Why a decade of “AR automation” could never go the final mile

Rules-based AR software has existed for years. These software solutions send reminders on a schedule, and for a straightforward invoice that works fine, but anyone who works in AR knows that is rarely the case.

Across the receivables we manage, there is complexity in nearly every payment scenario. 92% of enterprise invoices have to be submitted through a vendor portal or network rather than paid from an emailed invoice, and each portal is its own piece of browser automation. Our data puts edge cases at 39% of the slowdown in cash flow: a missing W-9, an approver out of office, a PO that does not match, a portal that rejected the format. In a NACM and BlackLine survey of more than 400 credit professionals, 44% said they use little or no automation for remittance data, which arrives spread across emails, PDFs, portals and lockboxes.

Almost none of the slowdowns are caused by a customer refusing to pay; they are caused by friction, invoices quietly sitting in individual inboxes for weeks. Pre-LLM software has never been sophisticated enough to handle edge cases, and as a result, the work to collect unpaid invoices goes back to the finance teams.

Collecting is full of nuance, and it’s not so easy to imagine putting an AI agent between you and your customers

An unconsidered message from an agent can damage a relationship that took years to build. LLM software has the unique ability to shape-shift for every customer need, tone, and situation. When deployed correctly, agentic AR tools can take the work off of a finance team’s plate, but it must be done with zero margin for error.

At Monk, every model we build is wrapped in deterministic code and tested against thousands of real cases before it reaches a customer. Cash application runs in three tiers: exact matches apply automatically with no model involved, then customer-specific rules the system suggests once it has seen enough payments, and only what survives both goes to an agent. Every match shows its working and keeps an audit trail. Our voice agent places calls and answers questions about an invoice, and it cannot rewrite an invoice, change a payment status, or accept new bank details over the phone.

None of this nitty-gritty groundwork for our agents is thrilling. But a CFO can get pretty excited when they see an agent elegantly handle a dispute, get cash in the door fast, and when they can see every step the agent took in a clean audit trail.

How AR Automation is the low-hanging fruit to deliver hard ROI with AI

We manage more than $2 billion in receivables. Across our customers, Monk agents resolve 90% of collections with zero human intervention, cash application matches 80% of payments automatically and rises to 95% once teams turn on suggested rules, and outreach reaches customers at a 24% higher response rate than standard dunning.

Teams using Monk see DSO drop by more than 40%, and see an average 37% increase in cash on hand in their first month. Cutting DSO pulls cash forward, and the finance team can spend that cash on growth, and their time on being strategic rather than stuck in manual chasing. For example, at $100M in revenue, a 40% DSO cut is roughly $5M in cash unlocked. ROI in AR is cash you already earned, pulled forward.

The rise of the finance strategist

For years, finance was promised automation, but oftentimes it never went the final mile to deliver hard ROI, and the most critical work landed on a person.

When AI agents can handle the exceptions, the edge cases, the portals, the people who spent every week chasing invoices can free up their time and focus on the most strategic work at hand. Instead of looking back, they forecast, they can ask what the cash position looks like next quarter. Everyone has been imagining what that version of the finance role would be for years, and now it’s possible.