Industries / Construction and property / Real estate / PropTech
Rental software
This is the software landlords and property managers use to advertise homes, screen applicants, collect rent and keep each tenant's account.
4
Milestones achieved, from 2008 to 2024.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2008
Landlord software moves online
AppFolio, formed in California in 2006, launched its first product, AppFolio Property Manager, in 2008. It runs over the internet, a model called cloud software, instead of being installed on office computers, and was aimed at firms managing 20 to 3,000 rental units. AppFolio's 2015 prospectus says more than 90% of its revenue had come from it, and that it had added online rent payment and tenant screening as paid extras.
Credit
AppFolio
2021
Rent counts toward a mortgage
From 18 September 2021 Fannie Mae's Desktop Underwriter, the software lenders use to check home loans against Fannie Mae's rules, began counting a record of on-time rent read from 12 months of the applicant's bank statements. It applies to first-time buyers with a credit score who have rented for at least 12 months at $300 or more a month, and missed payments are not held against them.
Credit
Fannie Mae
2023
Self-guided tours draw scams
In a self-guided tour a would-be tenant views an empty home alone: they register with an app such as Rently, prove who they are, and get the code to a lockbox that holds the key. In July 2023 KSAT, a San Antonio television station, answered fake ads that copied real homes at lower rents; the fake landlord had reporters open the lockbox through Rently and then pressed them to pay a deposit. Rently, which was not accused of wrongdoing, said it had cut reported security incidents by 47.5% in five months, and in October 2024 it added a service that watches more than 13 social media sites for copied listings.
LessonChecking the renter's identity does not check the landlord's, and a real tour system can make a fake listing look real.
Credit
Rently; KSAT
2024
A screening score is curbed
Tenant screening companies sell landlords a score, or advice to approve or decline, built from an applicant's credit and eviction records. In Louis v. SafeRent Solutions, two Black women with housing vouchers, a government subsidy that pays part of the rent, argued that SafeRent's score ignored the voucher and so turned away Black and Hispanic voucher holders more often. On 20 November 2024 a federal court in Massachusetts approved a settlement of up to $2.275 million under which, for at least five years, SafeRent gives no approve or decline advice for voucher holders unless civil rights experts have validated its model; SafeRent admitted no fault. In May 2024 the US housing department, HUD, had said that screening companies using AI must comply with the Fair Housing Act.
LessonA score that leaves out how a renter will actually pay can screen out people who would pay.
Credit
the plaintiffs and the Community Action Agency of Somerville; US District Court for the District of Massachusetts; HUD
Watch for
AI agents answer renters' calls
RealPage says its AI leasing agent, which answers would-be renters by phone, resolves up to 86% of their inquiries without staff and raises tour bookings by up to 22%. Its footnote gives only "aggregate AI Voice customer results before and after AI", with no period, sample size or definition of resolved.
Achieved whenAn outside party reviews a sample of calls and reports the share answered correctly with no staff involved.
Claimed by
RealPage, December 2025
Rent history doubles approvals
Fannie Mae says changes made at the start of 2025 to how Desktop Underwriter reads 12 months of rent payments doubled the number of potential first-time buyers it found creditworthy. It gave no count of loans made or of how those loans performed.
Achieved whenFannie Mae, its regulator the Federal Housing Finance Agency or an outside study publishes how many loans rent history made possible and how often they defaulted.
Claimed by
Fannie Mae, June 2025
Open
01
Fact-check leasing assistants
A public test of whether the AI assistants that answer would-be renters for apartment buildings, by chat, text or phone, give correct answers. For each of at least 10 buildings, record the facts the building publishes on the day of the test: rent for each unit type, required monthly fees, deposits, pet rules and available units. Ask each assistant the same 30 questions, including the full monthly cost with required fees, and score each answer as right, wrong, or passed to staff. Do not book tours or give real personal details.
The briefCredit · unclaimed02
A rent ledger error test
A public set of 300 made-up tenant ledgers, the running account of rent, fees and payments a landlord keeps for each tenant, with known mistakes planted: late fees charged when rent was paid on time, payments applied to fees before rent, duplicate charges, and fees the lease does not allow. Score the checks in property software and AI reviewers on how many mistakes they find and how many false alarms they raise.
The briefCredit · unclaimed
Open 01 · the brief
Fact-check leasing assistants
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of whether the AI assistants that answer would-be renters for apartment buildings, by chat, text or phone, give correct answers. For each of at least 10 buildings, record the facts the building publishes on the day of the test: rent for each unit type, required monthly fees, deposits, pet rules and available units. Ask each assistant the same 30 questions, including the full monthly cost with required fees, and score each answer as right, wrong, or passed to staff. Do not book tours or give real personal details.
Why it is still open
RealPage says its AI leasing agent resolves up to 86% of inquiries without staff but gives no accuracy figure (see Watch for). Funnel, another seller of these assistants, published a blind study of 169 renters in June 2026 that measured which assistant renters preferred and how knowledgeable each seemed, not whether its answers were right. A search in November 2026 found no outside accuracy test.
Built with what exists
The public websites, chat windows and phone lines of buildings that use these assistants; saved copies of each building's published fees and policies; a speech-to-text model such as ElevenLabs Scribe v3 (see the Note "Fewer word errors") to transcribe phone replies; a spreadsheet for scoring.
First step
Pick three buildings whose websites use different assistants, save their fee and pet pages, and ask each assistant ten questions.
Done looks like
At least 300 questions across at least three products, each answer scored by two people against the building's published facts, with their agreement rate, the share answered right for each product, and every transcript published.
Open 02 · the brief
A rent ledger error test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of 300 made-up tenant ledgers, the running account of rent, fees and payments a landlord keeps for each tenant, with known mistakes planted: late fees charged when rent was paid on time, payments applied to fees before rent, duplicate charges, and fees the lease does not allow. Score the checks in property software and AI reviewers on how many mistakes they find and how many false alarms they raise.
Why it is still open
A property software maker's help pages say wrong settings can keep charging late fees after rent is paid and can apply payments to fees instead of rent, and a New York tenant coalition's survey of 526 tenants found 64% had been charged non-rent fees, some of them late fees on rent paid on time. No public test measures whether software or AI catches such errors.
Built with what exists
Python to generate ledgers in the CSV layouts property software exports; sample lease fee rules; a current AI model as a baseline reviewer.
First step
Write ten ledgers by hand from one sample lease's fee rules, each with one planted mistake.
Done looks like
300 ledgers with an answer key, and published detection and false-alarm rates for at least three checkers.