As real estate agents increasingly use AI to generate text for property descriptions and social media marketing, their brokerage leaders have bigger fish in mind.
New brokerage-funded tools rolling out this year look to expand AI’s role in the brokerage from mere text generation to images and video, as well as database management and back-office document processing.
Leaders at real estate companies such as Compass, Anywhere and other large brokerage networks in recent months have publicly discussed how they’re planning to grow AI’s place in their business operations.
In this week’s second installment of our midyear AI series, Intel lays out insights from 55 real estate AI power users — agents and brokerage leaders across the country who say that the new tech is already providing a “significant” boost in their productivity.
Intel examines what this group is getting out of these tools that the typical agent isn’t. Intel also explores what these power users can teach us about how the technology’s place in the brokerage may evolve in the months ahead.
Read the complete breakdown in the full report.
What’s in progress
In last week’s AI report, Intel reviewed the rise in AI adoption at brokerages, which until now has mostly been fueled by agents using free-tier, third-party tools such as the default version of ChatGPT, rather than tools provided by their brokerage.
But even as this trend emerged, America’s largest residential real estate companies were working behind the scenes to develop custom tools that they hope will lead to productivity gains in more areas of brokerage operations.
CEO Robert Reffkin told his company’s investors in July that his company was preparing to debut a beta-test version of “Compass AI 2.0” as early as the fall.
Although details of the functionality were scarce, Reffkin said that the new AI-powered tools would focus first on agent productivity. But in time, he added, the new tools would be “deployed across the organization to make us more efficient.”
This focus on back-office efficiency appears to be a common thread in many of the public statements brokerages have made about their ongoing investments in AI.
Anywhere Chief Financial Officer Charlotte Simonelli said on a July earnings call that her company was using AI to take aim at a near-total automation in brokerage document processing.
“One-third of our Coldwell Banker brokerage documents submissions are now automated,” Simonelli said at the time, “with a path to 90 percent by the end of this year.”
Other firms, such as the Real Brokerage, are expanding internal offerings for agents and taking the additional step of AI-fueled services for clients as well, including AI-enhanced phone lines, property information and tour-scheduling.
But agents at these and other brokerages might not need to wait in order to take advantage of some of these productivity gains, the latest Intel survey results suggest.
Dozens of agents across the U.S. dished to Intel recently about how they were going beyond simple text generation to embed AI into additional parts of their workflows.
The next frontier
Back in December, Intel surveyed large groups of agents who got varying returns from their AI experience. In its analysis of this survey data, Intel focused on how power users were wringing additional value out of now-outdated versions of these models.
Since then, the state of the art has evolved as the major AI companies — including OpenAI, Google and Anthropic — unveiled newer model versions with “reasoning” or “chain-of-thought” style techniques that break prompts and responses into clearer component parts before trying to solve a problem.
These improvements led to benchmark gains in areas like coding, mathematics and comprehension across various documents and file types, even as the models remained limited and inconsistent across many domains.
And among power users — Intel’s term for those agents who said they’ve seen “significant” productivity gains from AI — the new capabilities have led to big leaps in usage for a number of more specialized brokerage tasks.
Here are some of the areas where brokerage AI power users are increasingly relying on AI:
Generating images for social media or marketing materials
- Power users: 24 percent in December → 48 percent in August
- Non-power-users: 10 percent in December → 12 percent in August
Managing leads or tasks in database
- Power users: 21 percent in December → 35 percent in August
- Non-power-users: 9 percent in December → 9 percent in August
Digital staging with listing photos
- Power users: 17 percent in December → 38 percent in August
- Non-power-users: 10 percent in December → 15 percent in August
Analysis or summary of market data
- Power users: 45 percent in December → 70 percent in August
- Non-power-users: 16 percent in December → 22 percent in August
Extracting data from property photos
- Power users: 3 percent in December → 25 percent in August
- Non-power-users: 2 percent in December → 7 percent in August
Interestingly, today’s power users also appear to be getting savvier when it comes to understanding generative AI’s potential to “hallucinate,” or otherwise relay inaccurate information.
While these models have largely been making gains in factuality over the years, their creators caution that they are still prone to mistakes and require humans to double-check their outputs.
- In December, 17 percent of AI power user respondents described having “a high degree of confidence” in model accuracy.
- By August, none of the power users Intel surveyed had high confidence that their models wouldn’t err or hallucinate. Still, 2 in 3 power users said they expected that the models they use would produce factual errors only “in rare cases.”
For brokerage leaders, the next stage in AI adoption appears to be about making these functionalities more reliable, and more accessible to even the casual AI user. And in the months after that, leaders hope, even more robust back-office uses might be realized.
Intel will continue to track these trends in the months ahead.
Methodology notes: This month’s Inman Intel Index survey ran from Aug. 20 to Sept. 3, 2025, and received 474 responses. The entire Inman reader community was invited to participate, and a rotating, randomized selection of community members was prompted to participate by email. Users responded to a series of questions related to their self-identified corner of the real estate industry — including real estate agents, brokerage leaders, lenders and proptech entrepreneurs. Results reflect the opinions of the engaged Inman community, which may not always match those of the broader real estate industry. This survey is conducted monthly.
