Every conversation about the finance team of the future turns to AI within a sentence. You tend to hear two feelings at once. There’s pressure to move fast, because the board is asking what finance is doing about AI. No one wants to be the team that falls behind.
Underneath it, there’s a quieter fear of moving the wrong way. Finance is the team that closes the books and keeps money moving. If something breaks here, it isn’t a glitch you fix later. It’s a customer invoice that goes out late, or never goes out at all. Both feelings are reasonable, and they pull against each other. That’s why so many teams freeze, or rush ahead on hope.
Where AI actually lands in finance
Most of the writing about AI in finance skips past that tension. It tells you finance is becoming more strategic, that people will process less and analyze more. The shift is real. In PwC’s AI Agent Survey, 79% of executives say their company already uses AI agents somewhere. Only 34% run them in accounting and finance.
That gap is the interesting part. Agents are showing up across the business faster than in the finance function. So the question for a finance leader isn’t whether this arrives, but where to start once it does. And “become more strategic” is no help there. It doesn’t tell you which task to hand over on Monday, or how you’ll trust it once you have. The real version of this shift is smaller and more concrete than the headlines suggest. It comes down to one skill: judgment about what to automate and what to keep close.
That’s the real change behind “doers to reviewers.” The finance team of the future does less of the doing and more of the reviewing. Getting there is a series of deliberate calls about where a person still needs to stand.
The question before handing any task to an agent
It isn’t “can AI do this?” It’s “if this goes wrong, will I catch it before it matters?” If yes, hand it over and stop doing it by hand. If no, that’s the work to keep close while you build the checks to change the answer.
Maya Cohen, CEO and Co-founder, Monto
What counts as execution, and what counts as judgment?
The shift starts with a split that sounds obvious and turns out to be the whole game. Some finance work is execution: it follows rules, repeats constantly, and looks the same every time.
Logging into a customer AP portal, reformatting an invoice, checking a status for the hundredth time, matching a document to a field. None of it needs you specifically. Getting it right matters, which is a different thing, and it’s exactly what software is good at.
The rest is judgment. It needs someone who knows your business and your customers. A credit decision, a disputed charge, a negotiation, a call on how hard to push a slow-paying account. No rule captures that work, and it’s the part of the job you were actually hired for.
Most teams get the split wrong in one of two ways. They treat the whole function as too sensitive to touch, so nothing gets automated and the pile grows. Or they try to automate everything, including the calls that need a human. Neither holds up. The skill is telling the two apart, task by task, and it rarely makes it onto a transformation slide.
Key takeaway
Execution follows rules and repeats, so it is safe to automate. Judgment needs a person who knows your customers, so it stays. Sorting your work into those two piles is the whole shift.
Where the shift actually starts: accounts receivable
This is the part the strategy pieces tend to miss. When you look for where this plays out first, it isn’t forecasting or the month-end close. It’s accounts receivable (AR). That’s where finance still does the most pure execution, and almost none of it is in your control.
Selling into large enterprises means billing through their systems. Your customers require you to submit invoices through a customer AP portal. That could be Coupa, Ariba, Tungsten, Tipalti, or hundreds of others. A supplier with enough enterprise customers manages dozens of them, each with its own fields, formats, and rules. Any customer can change theirs with little warning. So the financial process turns into a grind: logging in, re-keying invoice data, matching each invoice to its purchase order, checking statuses one portal at a time. Often you find out about a rejection weeks after it happened.
Which portal work to hand off first
That grind is the clearest case of execution in the whole finance function. It’s high volume and tightly ruled and repetitive by design. That’s exactly why it should be the first thing you hand off. Move manual portal work to AI that handles customer AP portal work. It lifts the least valuable, most time-consuming part of AR off your team’s plate, with little downside.
That order matches what AR teams tell us. The portal mechanics are the easy part to hand over. Submitting invoices and checking statuses are tasks nobody guards, because nobody enjoys them. What people hold onto is customer knowledge: the specific, ever-changing rules each customer runs in their portal. They also keep the investigative work of figuring out why something broke and making sure it doesn’t recur. There’s often disbelief that software could know a customer that well, which is fair. So the agent learns each customer’s rules and takes the repeatable submission work. The person keeps the relationship and the judgment that comes with it. The expertise stays where it belongs, with the people who built it.
How do you decide what’s safe to hand over?
You don’t need a complicated system to decide. You ask one question: if this task goes wrong, will I catch it before it matters?
When the answer is yes, hand it over and stop doing it by hand. Portal-based order-to-cash passes that test easily. A well-built agent can validate each invoice against the portal’s current rules before it sends anything. Anything off gets flagged for a person to check in near real time. The agent catches a mistake at submission, not three weeks later as a rejection. So the cost of being wrong stays low, and the time you get back is high. When the answer is no, keep that task close for now. If an error could slip through unnoticed and do real damage, use the freed time to build the checks that turn that no into a yes.
That’s how a finance team moves fast and slow at once. Quickly on the work that’s safe to get wrong, carefully on the work that isn’t. It isn’t a grand transformation initiative. It’s a long series of honest calls about where judgment still earns its place. For many finance organizations, that sorting is the real work of finance transformation, whatever the consultants put in the deck.
What it does to the people on the team
The fear underneath all of this is that automating the work automates the jobs. In AR, the opposite tends to happen. Submitting invoices and chasing statuses never took real skill. Handing them to an agent doesn’t shrink the team. It frees experienced team members in the finance department to work the disputes, difficult accounts, and customer relationships where judgment changes what gets paid and when. Cleaner submissions mean fewer errors and fewer rejections, so invoices go out right the first time. That helps reduce days sales outstanding (DSO) and steady cash flow. Fewer touches also mean lower processing costs, and those savings compound over the long term as volume grows. The work that stays is harder, more human, and worth more to the business.
Here is the shape that change takes. Before, an AR team spends its mornings logged into portals, keying invoices by hand. Today, submission takes minutes. The hours it frees go into collections: working past-due accounts before they age, reaching customers while the issue is still fresh. The timing flips, too. When a purchase order is incorrect or out of funds, the team catches it within a day or two. No more finding out months later, when no one remembers the details. The week moves from keying invoices to collecting money, the job these people were hired to do.
What order-to-cash looks like once the team reviews
Put it together, and order-to-cash stops being a queue of manual tasks. It becomes a supervised flow. An invoice leaves your ERP. An agent submits it in the format the customer AP portal requires, matches it to its purchase order, and tracks it through to payment. Nobody on your team touches it unless something needs a decision. The control doesn’t disappear in that trade; it moves up a level. Instead of one person checking one portal at a time, the team monitors the whole book of receivables in near real time, across every customer, from one place. That’s the operating model the more thoughtful writing about the future of finance keeps gesturing toward. In AR, you can build it today.
Where Monto fits
We built Monto for exactly this part. Monto is the autonomous co-worker that gets you paid by your enterprise customers, end-to-end. It puts a self-learning AI agent on every customer relationship. Each agent learns the exact workflows of the AP portals your customers use, Coupa, Ariba, Tungsten, Tipalti, and 500+ others. Then it carries every invoice from your ERP all the way through to payment. Submission, matching, and tracking become zero-touch. Your team monitors every invoice, purchase order, and payment from a single AI-powered dashboard, stepping in only when a decision requires a person. Execution goes to the agent, judgment stays with you. That’s the whole shift, built into one place.
The bottom line
The finance team of the future isn’t smaller. It’s pointed at better work. Hand the execution to AI, keep the judgment, and start with the portal work in AR. That’s the safest place to get it right.
See where the line between execution and judgment falls.
Keep the judgment. Hand off the portal work.
Book a Monto demo →FAQ
How do you decide which finance tasks to hand to AI?
Run each one through a single test: if this goes wrong, will you catch it before it matters? If yes, automate it and stop doing it by hand. If no, keep it close and build the checks that change the answer. Execution-heavy tasks like portal submission usually pass. Judgment-heavy work, like disputes and credit decisions, usually does not.
Will AI replace accounts receivable teams?
No. It takes the repetitive execution, submission, and status-chasing. People move to the judgment work: disputes, difficult accounts, cash strategy, and customer relationships. The team gets more skilled, not smaller.
What finance work should stay with people?
It’s the work that needs someone who knows your business and customers. Credit decisions, disputes, negotiations, relationships, write-offs, and cash strategy. That is judgment, and it stays yours.
What is the first AR task to give an AI agent?
Invoice submission through customer AP portals. It’s pure execution, high volume, and the source of most late-payment delays. So it returns the most time at the lowest risk.
What does order-to-cash look like with AI agents?
A supervised, autonomous flow instead of a manual queue. An agent submits each invoice in the format the customer AP portal requires, matches it to its purchase order, and tracks it to payment. A person watches the whole book of receivables and steps in only on exceptions.