AI in Commercial Real Estate: Why Pilots Stall and What Works
Key answer
Most CRE professionals already use general AI tools, but few firms have AI built into their workflows. Pilots stall on messy lease data, unclear owners and tools that sit outside daily work. Firms that succeed start with one repeatable workflow, like lease abstraction or LOI review, and run it inside the AI their team already uses.

By now you’ve used ChatGPT. You’ve seen the headlines. You probably have a few people on your team who swear by it and a few who ignore it entirely.
Here’s what the data actually shows: ChatGPT alone hit 400 million weekly users (OpenAI, February 2025).
But individual use and business adoption are two completely different things. Across the OECD, only 20% of firms (2025) have actually deployed AI at an organizational level. And in commercial real estate — shopping centers, retail, industrial — the gap is even more glaring.
76% of CRE firms say they’re exploring or implementing AI (Deloitte, 2024). Only 5% say they’ve actually hit their goals (JLL, 2025).
That’s not a technology problem. That’s a strategy and execution problem. And it’s exactly what this piece is about.
What Is an AI Agent in Commercial Real Estate?
Before we get into the real estate angle, let’s be clear about what “agentic AI” actually means — because it’s not just a chatbot with a nicer interface.
A traditional AI tool answers a question. You ask, it responds. Full stop.
An AI agent does something more: it receives a goal, breaks it into steps, takes actions, evaluates the results, and keeps going until the job is done. It can call other tools, loop back when something doesn’t look right, hand off subtasks to specialized agents, and escalate to a human when judgment is genuinely required. It’s the difference between a calculator and a colleague.
And that distinction matters enormously in Commercial Real Estate, where the work isn’t a single question — it’s a chain of interconnected tasks buried in thousands of pages of documents.
For example, a lease abstraction agent reads a lease file and its amendments, extracts the controlling terms, calculates dates, cites source clauses and flags uncertainty for review. An LOI agent compares business terms with the landlord’s playbook. Both return a finished file rather than just a chat answer.
Why CRE Is Such a Good Fit — and Still Getting It Wrong
If you run shopping centers or a growing retail portfolio, you are sitting on some of the most AI-ready data in any industry: leases, rent rolls, tenant sales reports, maintenance logs, deal pipelines. This is exactly the kind of information AI was built to work with.
The opportunity is real. JLL surveyed 1,500+ senior CRE decision-makers in 2025 and found that nearly 90% are running AI pilots. Two years ago that number was 5% (2023). The acceleration is genuinely historic.
And yet — only 5% (2025) are achieving their goals. Most pilots quietly stall. Most tools go underused. Most organizations are paying for AI and not getting value from it.
So what’s actually going wrong?
Why AI Pilots Die (And It’s Almost Never the Technology)
Here are the patterns we see fail repeatedly:
No specific problem to solve. Teams adopt AI because they feel behind, not because they’ve identified a broken workflow. A tool without a job becomes shelfware.
Messy data. AI can’t abstract what it can’t find. If your leases are in a mix of PDFs, email attachments, and someone’s desktop, no AI tool will save you until that’s fixed.
IT owns the rollout. When AI is treated as a technology project rather than a business change, business teams don’t adopt it. Your IT person can configure a tool. They can’t change how your leasing team actually works.
Piloting instead of deploying. A pilot that works in one property or one market proves the concept. It is not the destination. Organizations that stay in pilot mode indefinitely get no ROI and build no capability.
Skipping the people part. IBM’s research found that lack of AI skills (33%, 2023) and data complexity (25%, 2023) are the top barriers to deployment — not access to tools. The tools are accessible. Preparing your people to use them is the work.
What “Ready” Actually Looks Like
Before picking tools, be honest about four things:
Your data. Can you actually find and access your leases, rent rolls, and deal history? Are they consistent? If your data is scattered and dirty, AI will surface that problem faster than anything else.
Your workflows. Do you know how work actually gets done on your team today? You can’t automate — or reimagine — what hasn’t been mapped. AI doesn’t fix broken processes. It accelerates them.
Your leadership. Is there a decision-maker who owns this with a real business outcome in mind? AI initiatives without executive ownership and clear KPIs die in committee.
Your people. You don’t need data scientists. You need people who are curious, capable, and willing to learn — and a plan to actually train them.
The Four-Quadrant Framework: Where to Start and What to Build Toward
Not all AI investments are equal. The smartest approach is to sequence your adoption based on two dimensions: how much value it delivers and how hard it is to implement.
Quadrant 1 — Start Here: Low Hanging Fruit (Easy + Moderate Value)
These are productivity tools that any team can use immediately with almost no implementation overhead. Deploy these now. They build AI fluency and return time to your team within weeks.
Claude or ChatGPT for drafting emails, summarizing documents, and meeting notes
AI meeting transcription (Otter, Fireflies) so nothing falls through the cracks
AI scheduling and inbox management tools
This isn’t transformational. It’s the warm-up. But skipping it means your team arrives at more complex AI tools cold.
Quadrant 2 — Do Next: High Value + Easier to Implement
These tools deliver serious ROI and don’t require you to reinvent your operations. They work on data and workflows you already have.
AI lease abstraction for shopping center owners — agents that pull key dates, clauses, options, and obligations from leases in minutes, not days. For any shopping center owner managing 20+ leases, this alone pays for itself quickly.
Automated rent roll analysis — flag anomalies, track expirations, and model scenarios without a spreadsheet marathon
AI-assisted due diligence — review LOIs, PSAs, and estoppel certificates at speed
Lease comp analysis — pull and summarize market comparables from deal databases
Tenant sales reporting — AI that consolidates, cleans, and flags underperformers from tenant sales data
These use cases have a clear input, a clear output, and proven vendor solutions. The implementation barrier is low. The value is high. Start here after your productivity layer is in place.
Quadrant 3 — Build Toward: High Value + Requires Workflow Redesign
These are the most transformational opportunities — but they require you to rethink how work gets done, not just add a tool to an existing process.
End-to-end deal pipeline automation — from prospect to executed lease with AI at every step
Predictive tenant retention modeling — identifying at-risk tenants before they give notice, using sales trends, foot traffic data, and lease history
Dynamic rent optimization — pricing vacancy based on real-time market signals and portfolio-wide data
Automated portfolio risk scoring — continuous monitoring of lease expiration concentration, tenant credit risk, and market exposure
These are genuinely game-changing. They also require clean integrated data, cross-functional buy-in, and 12-24 months of serious investment before you see full returns. McKinsey’s 2025 research found that organizations seeing the greatest impact from AI were more than three times more likely (2025) to have redesigned their workflows — not just added tools. Plan accordingly.
Quadrant 4 — Avoid for Now: Complex + Low Return
Fully autonomous lease negotiation. AI replacing broker relationships. Custom model-building on thin proprietary datasets. These exist in the hype layer, not in the practical value layer. Avoid until the foundational work is done.
Build vs Buy: Best Option Is In Between
If you are a shopping center owner, asset manager or landlord rep broker, building a proprietary AI model is rarely the best starting point. Here is why:
A custom AI build requires data engineers, machine learning expertise, and months of development. By the time you’re done — 18 to 24 months — the commercial tools have lapped you. The AI landscape moves faster than any internal development timeline.
However, subscribing to a buffet of AI tools that may or may not work, while adding data-exposure risks, is not ideal either.
Use the Claude or ChatGPT subscription your firm already has, then add tested CRE agents through a connector. Start with one repeatable workflow, such as lease abstraction or an LOI business-terms review. Agents read files from the folders you authorize and save finished abstracts, LOIs or reports back to them. Your team reviews flagged exceptions; IT approves access, data-processing terms and security settings. There is no need to commission a white-label application or migrate the portfolio into another system.
When you’re evaluating vendors, the five things that actually matter:
Your data stays yours. Does the vendor train on your data? Any vendor that can’t give you a clear contractual guarantee that your lease terms, deal flow, and tenant data stay private is disqualified. Full stop.
CRE-specific, not generic. A general-purpose AI tool requires significant configuration to handle leases, rent rolls, and deal structures. Purpose-built CRE tools deliver value faster.
Fast implementation. The right vendor gets you live in weeks, not months. If they’re quoting a six-month implementation timeline, keep looking.
No training required to use it. Your leasing team and asset managers shouldn’t need a certification to open the tool. If it requires ongoing IT support to use day-to-day, it won’t be used.
Integrates with what you already have. AI tools that require manual export and import of data get abandoned quickly. Look for native connections to your existing systems.
How to Actually Implement This: Three Things That Work
Stand up an AI committee — with the right people in it
This doesn’t need to be formal or large. You need one senior person who owns AI outcomes, representation from leasing, asset management, and property operations, and a clear mandate to prioritize use cases and measure results.
The committee sets strategy and tracks ROI. It is not a technology committee.
Get IT out of the driver’s seat
IT should handle security, vendor contracting, and infrastructure. That’s it. Business units need to own their AI implementations. A lease abstraction solution that your leasing team doesn’t trust and hasn’t been trained on will not be used — no matter how cleanly IT implemented it.
Put AI Activators inside every team
The model that works: identify one high-performer per department who is curious about AI and trusted by their peers. Train them deeply on the tools relevant to their team. Let them teach, adapt, and surface new use cases from the inside.
This is not a technical role. It’s a translation role. And it is the fastest way to turn a tool license into actual behavior change.
The Bottom Line
Commercial real estate firms will be separated by which landlords, asset managers and brokers build practical AI capability early and which wait for it to be obvious. The advantage comes from making repeatable work reliable, not collecting more tools.
The solutions exist. The use cases are proven. The vendors are ready.
What’s missing in most organizations is the intentionality to sequence it right, invest in the people side, and stop treating AI as an IT project.
The 5% (2025) of CRE firms hitting their AI goals aren’t smarter than everyone else. They’re just more deliberate.
Sources: McKinsey State of AI 2025 | JLL Global Real Estate Technology Survey 2025 | OECD ICT Access & Usage Database 2026 | IBM Global AI Adoption Index | Deloitte CRE Outlook 2024 | Wharton Human-AI Research 2025 | Fortune Business Insights
Further reading: The Shadow Partner, a guide for CRE leaders on governing AI adoption.
FAQ
What is an AI agent in commercial real estate?
Software that completes a multi-step task, like abstracting a lease file or checking an LOI against a playbook, and returns a finished file.
Where is AI used in CRE today?
Lease abstraction, clause reports, LOI drafting and review, market research and tenant prospecting.
Should CRE firms build or buy AI?
Most buy the model (Claude or ChatGPT) and add tested CRE agents on top, instead of building from scratch.
Built for every side of the lease
Asset Managers
Portfolio visibility, critical dates and lease-up support from clean lease data. Bring your leases to the call.
Lease Administration
Long-form abstracts across every amendment, plus clause reports and lease Q&A. Bring your toughest lease file.
Tenants & Corporate Real Estate
Acquisitions & Due Diligence
Bulk abstraction sized to your diligence window. Bring your deal timeline.
Landlord Rep Brokers
Every LOI checked against each landlord’s terms and your deal history. Bring a landlord’s master lease form and a recent LOI.
Every subscription includes guided setup and training. New agents arrive as they ship.


