Chatbots, in other words, are just thestart. In a way, they can be seen as the lowhanging fruit. Page reports that RET
Ventures has a couple of portfolio compa-
nies looking at expanding the chatbot’s
typical use to other scenarios. “Existing ten-
ants have questions too and the same tech
can be applied to them. Perhaps they
believe there is a leak in their bathroom. An
AI-based chatbot can help them get to the
bottom of that and schedule an appoint-
ment to fix it if necessary.”
Revenue management—how owners
price apartment or housing units across
portfolios—is another area investors are
eyeing for AI, he continues.
There hasn’t been the same advance-
ments in this area as the chatbots but Page
reports that several companies are making
progress on cracking this nut. “Legacy
revenue management applications basi-
cally look at comparable buildings in
order to price units but they do not get to
the granular level, such as how to price a
unit that looks out over an overpass, for
example.” In other use, AI could conceiv-
ably process which businesses are nearby
and which demographics would need to
be targeted in order to maximize revenue
for a particular building, he says. “Or you
look at renewals and concessions and feed
that data into a model. The system then
tells you that this renter has a 90% chance
of renewing so there is no need to reach
out and offer a concession as his lease
nears the end.”
There are few commercial applications
on the market that do this now, Page says.
He has heard of certain landlords, such as
apartment owners and single-family rental
home operators, developing such applica-
tions in house after hiring their own data
scientists to lead such an initiative. These
projects could go down one of two paths,
Page says. Either they will be competitive
intellectual property that the owner doesn’t
want to share with a competitor or it will be
productized and sold broadly.
One argument for the latter is that themore broadly a particular revenue management model is shared the more accurate itbecomes. “We think eventually these companies will become comfortable spinningout the solutions they are developing inhouse,” Page says.
Some companies are not keeping their useof AI under wraps but instead openly incorporating it into pieces of its operations.Walker & Dunlop, for example, is using AIfor property valuations and appraisals. In2019 it acquired Enodo, a tech startupfocused on the underwriting process. Thenin January 2020, Walker & Dunlop andGeoPhy launched a joint venture calledApprise to focus on appraisals.
But even with whole divisions dedicatedto adding automation and higher tech tounderwriting and appraisals, experts findthat tackling the tasks incrementally is stillthe best approach.
It is appealing to think that AI couldautomate the valuation of an entire building, says Marc Rutzen, CEO and SVP ofInformation Technology for Enodo. Butmany investors wouldn’t trust the end resultof such a complex task, he says.
“We found it easier to apply and morereadily accepted to use AI to handle smallerdecisions in the underwriting process,”Rutzen says.
Consider the documents used in a valua-
tion process—rent rolls and operating
statements for instance. “These documents
are completely different and in different
formats. We build algorithms that pick out
relative data, such as rents and unit num-
bers of the rent roll documents, and then
run a series of predictive algorithms to see
if the rents are too high or too low.”
Or the system could be used to see if
expenses are historical outliers, Rutzen
says. “We can point out to the underwriter
that a certain expense is higher than it
should be.” The system isn’t doing the job
of the underwriter, he says. Rather it is help-
ing him get to a final conclusion faster.
The value of the AI can be seen in thetime savings for humans who still need tomake final decisions. Before it deployedEnodo, Walker & Dunlop’s average time toprocess a rent roll was 55 minutes. Now itcan process a rent roll in about five minutes,according to Rutzen. For operating statements, it took about 90 minutes to processcurrent operations and four years of historicoperations. Now it is down to 15 minutes.
“Multiply that by the thousands of dealswe do and the thousands more we look atand that is a pretty substantial time savingsin workflow automation,” Rutzen says.
Similar time savings also benefits Walker& Dunlop in the appraisal side of the business, says Meghan Czechowski, the valuation lead for Apprise and managing director of the firm’s midwest region. “Over thepast four years the volume of multifamilytransactions have spiked 74%, while thenumber of appraisers has decreased by10%. The industry’s professionals are trying to keep up with ever escalating data, butthey are doing it manually.” Applying AI toautomate some of these processes can leadto valuable savings, she says.
The Apprise application pulls lead unitmix information from multiple sources intothe system, and can compare these sourcesdirectly to map discrepancies and anomalies,she explains. “The constant updates forapartment unit mixes from industry standard resources coupled with our automatedprocessing of actual rent roll informationinto unit mixes allows the system to recordunit type and floor plan level changes over aperiod of time allowing appraisers a strongerbase for trend analysis and support for rentforecasting within our appraisal reports.”
WE ARE SEEING A LOT OFADOPTION AROUND BUILDINGNATURAL LANGUAGEPROCESSING TOOLS FORBOTS, FOR EXAMPLE. THESEBOTS ARE ABLE TO HAVEADVANCED CONVERSATIONSMUCH LIKE A LEASING AGENTCOULD.
ALEC PAGEVICE PRESIDENT OF RET VENTURES