How is AI most often used in real estate?

The Pilot Paradox: Why Most Real Estate Firms Are Stuck in Experimentation Mode

Here's a puzzle: nearly half of real estate firms are experimenting with AI, yet almost none have made it stick. According to Keyway's state of AI adoption survey, 45% of firms are running AI pilots while only 9% have achieved successful enterprise-level deployment.

Researchers call this the Pilot Paradox, and it's not a motivation problem. As Keyway puts it, the primary bottleneck isn't organizational willingness—it's data infrastructure. A striking 76% of firms report significant data infrastructure limitations, and only 8% consider themselves fully data-ready for AI adoption. Pilots run on clean demo datasets; production runs on the messy, fragmented reality of a firm's actual records.

Trust creates the second wall. Even when the technology works, decision-makers hesitate to rely on it. The same survey found that 44% of investment committees distrust AI-generated analysis, citing hallucinations, unreliable outputs, and data privacy as their primary concerns. For AI to move upstream into underwriting, valuation, and investment decisions, Keyway argues it must be explainable, auditable, and grounded in traceable data—trust, not technology, is the true gating factor.

So where does AI actually succeed today? In the low-risk, high-volume work that doesn't require a committee's sign-off:

• Writing listing descriptions — the top use case for 75% of REALTORS® who use AI
• Creating social media posts — used by 56% of AI adopters
• Drafting emails and follow-ups — used by 52%
• Handling routine resident inquiries through chatbots in property management

The pattern is clear: content and communication tasks clear the trust bar, while financial analysis does not. That's also why the industry has settled on augmentation over automation. NAR's guiding philosophy is that AI should "enhance, not replace" the trusted human expertise at the heart of every transaction, and PwC and ULI's Emerging Trends research finds most firms still in preliminary stages, with job transformation far more common than replacement.

The stakes for getting past the paradox are enormous. Morgan Stanley estimates $34 billion in operating efficiencies for real estate by 2030, driven largely by labor cost savings in management, sales, and administrative support. Firms that standardize their data first—and build trust through explainable, human-in-the-loop workflows—will capture that upside. The rest will remain stuck running pilots.

For teams ready to move beyond experimentation, the practical path starts with the workflow itself: audit your data, pick high-volume low-risk tasks, and verify the tools and agent skills you deploy actually perform. That verification step matters everywhere in AI adoption—published research has found roughly 26% of published agent skills carry at least one vulnerability, which is why SkillzBench is building a free directory that security-scans and benchmarks agent skills so firms can integrate them with confidence.

Where AI Is Actually Working Today: The Top Three Real Estate Use Cases

When real estate professionals reach for AI, they overwhelmingly reach for words. According to the NAR Technology Report, 75% of REALTORS® who use AI turn to it for writing listing descriptions, 56% for social media posts, and 52% for drafting emails and follow-ups.

The pattern is clear: the industry's first wave of AI adoption is concentrated in high-volume, low-risk communication tasks. A listing description can be reviewed, edited, and approved in seconds, and a flawed social post costs little compared to a flawed valuation. That low-stakes nature makes these tasks ideal proving grounds.

• Listing descriptions (75%) — turning property facts into compelling marketing copy at scale
• Social media content (56%) — keeping a consistent presence without hours of drafting
• Email and follow-up drafting (52%) — accelerating client communication and lead nurturing

Generative AI excels here because these tasks share three traits: they repeat constantly, they follow recognizable formats, and a human can quickly verify the output. Contrast that with underwriting or investment analysis, where Keyway's research found 44% of investment committees distrust AI-generated analysis. Trust, not capability, is what separates the adopted use cases from the aspirational ones.

Notably, 81% of REALTORS® adopt technology primarily for efficiency and productivity — not to shrink headcount. This reflects the industry's guiding philosophy at NAR that AI should "enhance, not replace" human expertise. JLL echoes this, noting that role augmentation and selective displacement happen simultaneously rather than as uniform replacement.

The result is a co-pilot model: agents draft faster, respond sooner, and market more consistently, while judgment-heavy work — negotiation, pricing strategy, client relationships — stays human. As teams standardize these workflows, many are extending them further with agent skills and connectors; SkillzBench is building a free directory and MCP tool for exactly that kind of workflow extension, so you can browse, verify, and test skills before wiring them into your stack.

For businesses planning integration, the lesson from these numbers is simple: start where volume is high and risk is low. Build confidence with content wins first, then expand toward higher-stakes decisions once data infrastructure and governance catch up.

Beyond Marketing Content: Customer Service, Operations, and the Path to Financial Decisions

Marketing content may dominate today's headlines, but the more consequential AI deployments are happening behind the front desk. From resident chatbots to HVAC tuning, operational use cases are quietly reshaping how buildings actually run.

In multifamily property management, AI is consolidating routine resident interactions under a single chatbot interface. PwC's research with ULI notes this shift allows leasing and service staff to support multiple properties from one location, reducing the need for redundant onsite headcount while keeping overall team size stable through expanded software maintenance roles. Beyond service, AI is being applied to energy management and risk identification, flagging cash flow volatility, climate exposure, and location-specific threats before they escalate.

The financial upside is substantial. Morgan Stanley estimates roughly 37% of real estate tasks are automatable, projecting $34 billion in operating efficiencies industry-wide by 2030, with the largest gains coming from labor cost savings in management, sales, and administrative support. The same analysis points to a potential 34% increase in operating cash flow for brokers and service firms that deploy AI effectively.

Yet the path from operational wins to high-stakes financial decisions remains blocked by a trust gap. Keyway's adoption survey found that 44% of investment committees actively distrust AI-generated analysis, citing hallucinations, unreliable outputs, and data privacy concerns. For AI to move upstream into underwriting, valuation, and investment decisions, it must be explainable, auditable, and grounded in traceable data. Trust, not technology, is the true gating factor.

This is why the industry's prevailing model remains augmentation over automation. AI serves as a co-pilot that surfaces insights and drafts outputs, while humans retain final authority on consequential calls. Organizations that want to extend AI deeper into their workflows can accelerate this transition by standardizing data first, then layering in automation. Teams experimenting with these workflows often package their best prompts and processes into reusable agent skills, and SkillzBench is building a free directory where you can browse and benchmark those skills before deploying them in production.

How to Integrate AI Into Your Real Estate Workflow: A Practical Roadmap

Knowing where AI works in real estate is one thing; building it into your daily operations is another. The gap is real: industry research shows 45% of firms run AI pilots, yet only 9% achieve enterprise-level deployment. Here is a practical roadmap, ordered from quick wins to full-scale rollout.

Step 1: Standardize your data first. The most common reason AI initiatives stall is not resistance to change—it is fragmented data. The same research found 76% of firms report significant data infrastructure gaps, and only 8% consider themselves fully data-ready. Audit your listings, client records, and property files for inconsistent formats before layering AI on top. As Keyway puts it, the winners will be "those who standardize data first—not those who simply adopt more tools."

Step 2: Start with high-volume content tasks. These are the lowest-risk, highest-frequency jobs where AI already proves itself daily. Among REALTORS® who use AI, NAR's technology report found 75% use it for listing descriptions, 56% for social media posts, and 52% for email follow-ups. These quick wins build internal confidence and deliver the efficiency gains that motivate 81% of technology adopters in the first place.

Step 3: Deploy chatbots for customer service. Once content workflows run smoothly, extend AI into resident and client services. PwC and ULI note that consolidating resident inquiries under a dedicated chatbot lets staff serve multiple properties from a single office, while Morgan Stanley found companies using AI this way reported higher satisfaction from both clients and teams.

Step 4: Add human-in-the-loop protocols for financial decisions. Trust is the gating factor for underwriting, valuation, and investment work—44% of investment committees distrust AI-generated analysis. Before AI touches financial decisions, require outputs to be explainable, auditable, and grounded in traceable data, with humans making the final call.

Step 5: Establish AI governance. NAR flags data bias, privacy concerns, and unclear rules as the industry's key risks, and offers policy templates brokers can adapt. Your governance framework should cover:

• Data privacy and client information handling
• Bias mitigation in AI-assisted valuation and pricing
• Human review requirements for high-stakes decisions
• Approved tools and acceptable use guidelines

As your workflow matures, agent-based tools can take over more of the routine execution—drafting, scheduling, and research tasks—under your governance rules. SkillzBench is building a free directory and MCP connector for exactly this: discovering, security-scanning, and benchmarking agent skills before they touch your operations, so you extend AI with confidence rather than guesswork.

Scaling AI Responsibly: Governance, Trust, and What Comes Next

The gap between AI experimentation and operational reality is widening. While 45% of firms run pilots, only 9% achieve enterprise-level deployment, primarily due to data infrastructure gaps. Industry analysis identifies trust, not technology, as the true gating factor for scaling AI responsibly.

To bridge this divide, firms must establish robust governance frameworks before expanding usage. NAR highlights that data bias and privacy concerns remain significant risks, noting the lack of clear, consistent rules governing AI use in real estate transactions. National Association of Realtors guidelines emphasize that AI should enhance, not replace, trusted human expertise. This "augmentation over automation" model ensures that AI serves as a co-pilot, preserving the professional judgment required for high-stakes decisions.

Implementing these controls requires specific structural adjustments:

• Adopt internal AI usage policies that explicitly address data privacy, bias mitigation, and copyright compliance.
• Require explainability and audit trails for any AI output used in underwriting or valuation, ensuring traceable data sources.
• Maintain human-in-the-loop protocols for financial decisions to mitigate risks associated with hallucinations or unreliable outputs.

As workflows evolve, the industry is preparing for agentic AI, which plans and acts with minimal supervision. PwC notes that while first-wave tools focus on content creation, agentic systems are emerging in predictive analytics and inventory management. This shift demands rigorous validation to ensure agents operate within safe, defined parameters. Research from PwC and ULI confirms that job transformation, rather than replacement, is the standard trajectory. Professionals who master these governance workflows will outperform those who ignore them, as AI becomes the invisible engine behind faster, more effective teams.

For teams looking to integrate these capabilities, agent skills can automate much of this verification and workflow management. SkillzBench is building a free directory and MCP connector to help professionals discover, verify, and benchmark these AI skills safely. By tracking upstream changes and security scans, the platform ensures that the tools you deploy are both effective and trustworthy. As you prepare for the next phase of AI adoption, prioritize skills that are proven, verified, and aligned with your specific operational needs.