88% Are Piloting AI in Commercial Real Estate, Yet Complete Success Remains Rare

AI in Commercial Real Estate: From Pilots to Profit

Commercial real estate has been slow to modernize, but that's changing. A new JLL survey shows 88% of investors, owners, and landlords are piloting AI, with most testing an average of five use cases at the same time. More than 90% of occupiers are also running pilots. Two years ago, that figure was just 5%.

Adoption is high. Outcomes are mixed. Only 5% say they've met all their AI program goals, while nearly half hit two to three targets. The shift from simple efficiency plays to revenue-focused use cases is raising the bar and exposing gaps in data, process, and operating models.

What this means for owners, developers, and operators

• AI is now a line item in the tech budget, not a side project. More than half of investors reported meaningful budget increases tied directly to AI.
• Spending is flowing first to strategic advisory, followed by cyber/data security and infrastructure for integration.
• The goalposts have moved from "faster tasks" to better underwriting, higher NOI, and smarter portfolio decisions.

Where AI is creating real value

• Investment and risk modeling: Scenario testing, risk-adjusted returns, portfolio optimization.
• Underwriting and acquisitions: Lease comps, rent rolls, T-12s, sales and debt comps, anomalies flagged in minutes.
• Leasing and revenue: Dynamic pricing, targeted outreach, faster deal cycles, tenant churn prediction.
• Asset and facilities management: Predictive maintenance, energy optimization, work order routing, vendor performance scoring.
• Tenant experience: Concierge chat, issue triage, amenity usage insights tied to retention.
• Construction and development: Schedule risk alerts, change-order analysis, submittal review, quantity takeoff assistance.
• Compliance and security: Policy checks, PII redaction, document controls across portfolios.

Why most pilots stall

• Shifting objectives: Moving from efficiency to revenue impact without changing operating models.
• Data sprawl: Fragmented leases, IoT, and financials with weak governance and lineage.
• Poor workflow fit: Tools don't integrate with existing processes or incentives.
• Security and risk: Incomplete access controls, vendor due diligence, and model oversight.
• Too many pilots: Five pilots at once, none with clear ownership or KPIs.

A practical 12-month plan

• 1) Pick three high-impact use cases: Tie each to revenue or NOI (e.g., underwriting speed-to-decision, energy cost per square foot, lease-up velocity). Assign an owner and a measurable goal.
• 2) Fix the data foundation: Centralize leases, financials, and building telemetry. Define data contracts, access controls, and retention. Start with one region or asset class.
• 3) Standardize vendors and security: Limit tool sprawl. Run privacy, SOC2/ISO, and model risk checks. Decide build vs. buy per use case.
• 4) Pilot with control groups: Compare AI-assisted vs. business-as-usual. Track lift in underwriting throughput, win rates, downtime reduction, or energy savings.
• 5) Integrate into workflows: Embed into your CRM, IWMS, or ERP. Document SOPs, train teams, align incentives, and set approval thresholds.
• 6) Scale what works, retire what doesn't: Productize proven pilots. Monitor model drift, cost, and compliance. Refresh KPIs quarterly.

Metrics that matter

• NOI margin lift per asset
• Leasing cycle time and win rate
• Underwriting accuracy and throughput
• Energy cost and carbon per square foot
• Vacancy duration and retention
• Capex forecast variance and schedule risk

Budget priorities

• Strategic advisory: Align use cases with portfolio strategy and capital plans.
• Security and data: Access controls, PII redaction, audit trails, vendor risk assessments.
• Integration and infrastructure: APIs, data pipelines, and monitoring to move from demos to daily operations.

Governance essentials

• Adopt a clear framework for risk, bias, and accountability. The NIST AI Risk Management Framework is a solid starting point.
• Require human-in-the-loop for high-impact decisions (valuations, approvals, investor reporting).
• Log prompts, model versions, and decisions for auditability.
• Set redlines: data that never leaves your environment, and vendors that don't meet your security bar.

Talent and training

• Stand up cross-functional squads (asset, leasing, ops, IT, data) for each use case.
• Upskill teams on prompt quality, model limits, and data privacy. If you need curated resources, see AI courses by job or the latest practical programs at Complete AI Training.

Bottom line

The industry is all-in on AI pilots. A small fraction are hitting every goal, but many are hitting some. The difference isn't the model-it's focus, data quality, security, and an operating model built for measurable revenue and margin impact.

Pick fewer use cases. Tie them to dollars. Build the plumbing. Then scale with confidence.

Related research: JLL Trends & Insights