What Your Controller Still Owns When AI Touches the Month-End Close

By Douglas Kohn, MBA, CPA

I read Yardi's announcement yesterday morning with my coffee, and one line caught my attention. Its new Virtuoso Enterprise platform includes a "Month-End Close Agent" that, in Yardi's words, "runs 30 reports automatically across your books, AR, AP, bank reconciliations, security deposits, rent roll and GL hygiene, and surfaces issues in plain language with suggested next steps."

Thirty reports is a good share of a property accountant's close checklist, at least on paper.

It isn't just Yardi. In the last few weeks, several vendors that real estate owners and contractors use have announced AI features aimed at accounting tasks, not just answering questions. So let's talk about what that means on the actual books.

What actually changed this fall

On the real estate side, AppFolio announced on September 29 that its Accounting Performer is now available and, in AppFolio's words, "automates bill entry, financial close, and budgeting." AppFolio says its beta customers reported saving an average of 17 hours per month; that's the company's figure, and results will vary by portfolio.

Yardi's October 5 release also lists an invoice approval agent, a lease audit agent that Yardi says identifies billing gaps, and AI invoice coding.

On the construction side, Adaptive, a construction accounting software company, announced a $30 million Series B on September 17. The company says its agents code invoices to job, phase and cost code, match them against purchase orders, prepare billings and assemble the monthly WIP, and that more than 750 construction companies use it.

The line in Adaptive's release I'd underline for every owner: the agents "never post to the general ledger on their own: a person reviews every action and provides the judgment and approval before it touches the books."

That's a sensible design. But it moves the work; it doesn't remove it. The review step is now the control.

The catch: AI is only as good as the data under it

The finance surveys this fall say the same thing in different ways. Protiviti's 2026 Global Finance Trends Survey, released August 26, found that 77% of finance organizations now use AI, but only 14% are doing it under a defined strategy, and only 35% say they're effective at measuring the return.

Construction looks similar. Grassi's 2026 Construction and A&E survey found more than half of firms already use AI, yet only 25% are very satisfied with the financial and operational information they use to run the business. Job costing accuracy was among the most common reporting obstacles.

In other words, adoption is running ahead of results. That's not a reason to avoid these tools. It's a reason to treat them as help with the prep work, while the judgment and the sign-off stay with the controller.

A useful piece this month came from Elliot Powell, director of data and AI at Pacific Pile & Marine, writing in Construction Dive. His team's AI tool couldn't say which revision of a submittal was current, "since we didn't know the answer ourselves." His advice: "Treat your first months of AI use as a data audit."

I'd say the same thing about your general ledger.

What a controller actually does with AI on the books

Here's where I'd focus if you're turning on any of these tools, whether you run a rental portfolio or a contracting business.

1. Review AI-coded GL entries by exception, and document that you did. An AI that codes 500 bills a month will be right most of the time and confidently wrong some of the time. Set a review threshold (dollar amount, new vendors, any bill coded to a capital account or a balance sheet account) and have a human approve those every time. Then scan GL detail monthly for drift, like repairs creeping into capital improvements or one property's utilities landing on another entity. The review itself needs evidence, such as initials, a date or a system approval log, because your auditor or lender will ask how you know the coding is right.

2. Keep the three-way match deterministic. Use AI to read the invoice and propose a match to the PO and the receiving record. Don't let it decide that a 4% price variance is "close enough." Your tolerance rules should be written down, set in the system, and owned by accounting, not by the model. For subcontractor pay applications, the match is really against the subcontract: billed-to-date versus the scheduled value, approved change orders and the retainage percentage. If the AI calculates retainage, tie the retainage payable subledger to the GL every month.

3. Treat the WIP schedule as a judgment schedule, not a report. An agent can assemble the WIP. It can't decide your estimated cost to complete. Under percentage-of-completion (the cost-to-cost input method most contractors use under ASC 606), revenue, gross profit, and your over/under billings all flow from that estimate. So the controls stay the same: a monthly cost-to-complete review signed by the PM, a check that every approved change order is in both the contract value and the estimate, and a review for jobs where cost is booked to the wrong job or phase. Miscoded cost doesn't just misstate one job. It overstates percent complete on one job and understates it on another, and both lines on the WIP are wrong.

4. Tie lease data to the rent roll and the GL. A tool that flags possible billing gaps can be useful; missed escalations and unbilled recoveries are real money. But any change that comes out of a lease audit should flow through a documented lease-change process: someone compares the lease or amendment to the setup in the system, updates the rent roll, and checks the straight-line rent schedule for operating leases where the payments changed. The same goes for leases where you're the tenant. AI-extracted dates, renewal options and rent steps feed your ASC 842 schedule, so a person should check those fields against the signed document before the schedule is recalculated.

5. Rewrite the close checklist, don't just shorten it. If an agent now runs your bank recs, deposit reconciliation and GL hygiene reports, the checklist line changes from "prepare bank rec" to "review agent bank rec output, clear flagged items, sign off." Save the outputs in the close binder so the audit trail shows what the tool found, what a person decided, and who approved it.

6. Keep access tight. Both Yardi and AppFolio say their AI works within your existing permissions and controls. If so, your permissions are now your AI controls. Review who can approve bills, edit vendors and post journal entries before you turn on anything that acts on a user's behalf.

The bottom line

These tools may save time on close prep, but only on top of a clean chart of accounts, consistent cost codes, clear approval rules and documented review. Otherwise AI just produces wrong numbers faster. Use it for the legwork, and keep the judgment and sign-off with a person who understands the books.

At Ultramar, my focus is fractional CFO and controller work for real estate and construction companies: setting up the close, the job-cost and WIP process, lease and rent roll controls, and the review steps that need to be in place before tools like these touch the books. SoFla Prime Consulting does fractional CFO, controller and accounting work for South Florida businesses and nonprofits that want the same discipline in their books.

Thinking about turning on one of these features, or already did and unsure what to review? I'm happy to talk it through.

Sources

Yardi unveils Virtuoso Enterprise, a complete AI platform for real estate, ahead of YASC San Diego. Yardi via PR Newswire. October 5, 2026. https://www.prnewswire.com/news-releases/yardi-unveils-virtuoso-enterprise-a-complete-ai-platform-for-real-estate-ahead-of-yasc-san-diego-302898226.html

AppFolio Introduces the New Realm-X, Built in the Platform That Already Knows Your Business. AppFolio via GlobeNewswire. September 29, 2026. https://www.globenewswire.com/news-release/2026/09/29/3371181/36405/en/appfolio-introduces-the-new-realm-x-built-in-the-platform-that-already-knows-your-business.html

Adaptive Raises $30 Million Series B Led by Tidemark to Bring Agentic Accounting to Construction. Adaptive via PR Newswire. September 17, 2026. https://www.prnewswire.com/news-releases/adaptive-raises-30-million-series-b-led-by-tidemark-to-bring-agentic-accounting-to-construction-302881537.html

CFOs Turn to AI to Better Synchronize Finance and Enterprise Priorities, Report AI ROI Challenges: Protiviti Global Finance Trends Survey. Protiviti. August 26, 2026. https://www.protiviti.com/us-en/press-release-cfos-turn-ai-better-synchronize-finance-and-enterprise-priorities

Construction and A&E Firms Are Investing in AI, but the Bigger Challenge Is Data Confidence. Carl Oliveri, Grassi. September 17, 2026. https://www.grassiadvisors.com/blog/ai-in-construction-why-data-confidence-comes-first/

AI didn't fix our construction data problem. It exposed it. Elliot Powell, Construction Dive. September 30, 2026. https://www.constructiondive.com/news/ai-doesnt-fix-data-problem-construction-need-unified-language/831774/