Strategic Intelligence Brief
AI workflow assessment:
Evidence and hypotheses
in Commercial Real Estate
A review of published sources and workflow hypotheses for occupier advisory. Test local results with a proposed two-week assessment.
Prepared by
iFrame Consulting
AI Strategy for Commercial Real Estate
Click a numbered reference to inspect the source and its limits.
Two survey measures
JLL's 2025 occupier CRE survey reports 92% started or planned AI pilots and 5% achieved most program goals. These are separate measures.
Four opportunity hypotheses
Lease abstraction, proposal assembly, market intelligence, and prospect research need local tests. Vendor reports do not establish your results.
Measure before procurement
Recommendation: record current hours, output quality, data access, and review costs before choosing a tool. An assessment cannot guarantee success.
A proposed two-week plan
Use the illustrative discovery and analysis schedule to define tests. Scope and timing depend on access and team availability.
Core Recommendation
Assess workflows and test the hypotheses before procurement. See /cre for the current Assessment offer.
Strategic context and published evidence
JLL announced Falcon, its CRE AI platform, on 29 October 2024. The announcement does not establish RFP speed gains.
JLL's 2025 survey covered 1,000+ senior CRE decision-makers across 16 markets. Among occupier CRE respondents, 92% had started or planned AI pilots in 2025. Separately, 5% reported achieving most program goals.
These measures do not classify the remaining respondents as failures. They do not establish results for every brokerage team.
JLL's 2025 occupier survey
Two separate survey measures. The bars are independent percentages, not parts of a total or a failure rate.
Started or planned AI pilots: 92%
Most program goals achieved: 5%
Started or planned AI pilots
Occupier CRE respondents who started or planned pilots in 2025.
Most program goals achieved
Occupier CRE respondents who reported most program goals achieved.
What this report offers
Four workflow hypotheses for occupier advisory need local tests. A proposed two-week assessment can define baselines and test criteria before procurement.
Four opportunity hypotheses
Test each workflow against your current process. Published vendor reports do not establish time savings, quality, or commercial results for your team.
Click any card to inspect evidence limits and local test criteria
Lease Abstraction & Clause Extraction
V7 describes AI extraction of lease terms. This is a vendor claim; test with your leases.
Vendor claim
Lease Abstraction & Clause Extraction
Workflow Transformation
Hypothesis: extraction tools may help reviewers locate lease terms. Compare extracted terms against the original documents.
Evidence and test criteria
- Measure omissions and incorrect clauses
- Include reviewer effort in the comparison
- Human review cannot guarantee accuracy
Proposal & Pitch Material Assembly
Proposify reports a 17-minute mean creation time across 27 industries. This vendor observation does not prove an AI effect.
Vendor mean; not a CRE benchmark
Proposal & Pitch Material Assembly
Workflow Transformation
Hypothesis: approved content libraries may help assemble proposals. Test preparation and review effort against your current process.
Evidence and test criteria
- Vendor data covers 1,280,657 proposals
- The sample is cross-industry, not CRE-specific
- Measure content quality and total review time locally
Market Intelligence Retrieval
Hypothesis: source-linked retrieval may help advisors find market evidence. No local benefit is established.
Local hypothesis
Market Intelligence Retrieval
Workflow Transformation
Compare retrieval against the current research process using the same questions and approved sources.
Evidence and test criteria
- Check source dates and coverage
- Measure incorrect and unsupported answers
- Include data access and review costs
Prospect Research Automation
Hypothesis: draft dossiers may help prepare for meetings. Test source quality, privacy, and reviewer effort.
Local hypothesis
Prospect Research Automation
Workflow Transformation
Compare draft dossiers with manually prepared dossiers using agreed quality criteria.
Evidence and test criteria
- Check facts against original sources
- Measure actual preparation and correction time
- Review privacy and permitted data use
Implementation questions before procurement
The practices below are recommendations to test with your team. They are not proven success rates or guarantees against failure.
The Implementation Divide
Recommended practices and risks to assess locally. Click a card for details.
Six recommended practices
Workflow-first scoping
Define the workflow problem and acceptance criteria before evaluating tools.
Baseline metrics before AI
Record current cycle time, errors, and throughput. Compare trial results against that baseline.
Human-in-the-loop by design
Assign qualified reviewers. Check extracted terms against source leases and record corrections.
Clear governance from launch
Define output ownership, escalation rules, and permitted data use before a trial.
Phased value delivery
Test a bounded workflow before expanding scope. Use measured results to choose the next step.
Audit readiness first
Assess workflow needs, data access, and governance before procurement. This recommendation does not guarantee results.
Six risks to assess
Solution-first thinking
A tool may solve the wrong problem if the team has not defined the required outcome.
Data quality neglect
Check data quality and source coverage before relying on outputs.
Absent change management
Plan team training and measure whether staff use the tool.
Unrealistic expectations
State test criteria and limits before a trial. Avoid promises that the evidence cannot support.
Vendor lock-in
Check export options and integration requirements before procurement.
Security afterthoughts
Review access controls and data handling before a trial.
Test before procurement
Map the workflow and define a local test. An assessment cannot eliminate every risk or guarantee a successful implementation.
Assessment recommendation
Map workflows, record baselines, and identify governance needs before procurement. Measure whether each test improves the process after review costs.
Estimate current labor cost
Use measured hours and a relevant labor rate. This estimate values current work; it does not predict recoverable savings or the cost of deferral.
Weekly Burn Calculator
Example inputs, not measured baselines. Edit hours and the labor rate to estimate current direct labor cost.
Weekly hours by workflow
Lease Document Analysis
Scanning leases for key terms, clause extraction
Prospect Intelligence
Pre-meeting dossier compilation
Proposal & Pitch Build
Deck creation, market slides, formatting
Quarterly Market Reports
Trend PDFs, portfolio summaries
Annual direct labor cost
$0
Estimate limits
- Hours and rate are example inputs
- Assumes 52 identical weeks
- Excludes costs outside the entered hours
Current work
Labor value, not savings
Measure a trial
- Record actual saved hours
- Include all trial and ongoing costs
- Check output quality after review
Potential savings
Not estimated
This estimate values current work. It does not predict savings or prove that AI can replace these hours.
Governance That Clients Can Trust
In client-facing advisory, speed without control becomes a liability. A clear AI governance model turns risk management into differentiation: your team can explain how outputs were produced, what data was protected, and who approved each recommendation.
Governance Framework
Click a pillar to explore best practices
Data Classification Framework
Establish tiered classification (Public, Internal, Confidential, Restricted) for all client data processed by AI systems. Map each AI workflow to its required classification level.
Access Control & Encryption
Enforce role-based access controls on AI tools. Use appropriate encryption at rest and in transit for client data.
Environment Isolation
Guarantee sensitive client lease data and financials never leave approved, contractually governed environments. Prohibit data from being used for model training without explicit consent.
Vendor Data Processing Agreements
Require DPAs from all AI vendors specifying data retention limits, deletion policies, subprocessor lists, and breach notification timelines. Review annually.
Output Provenance Logging
Every AI-generated recommendation must be traceable to its source data, model version, and timestamp. Set retention periods using applicable jurisdiction, client policy, and legal advice.
Version Control for Models
Track which model version produced each output. When models are updated, maintain the ability to reproduce prior results for client dispute resolution.
Bias & Drift Monitoring
Implement periodic checks for model drift and output bias. Document acceptance thresholds and remediation procedures when outputs deviate from baselines.
Regulatory Alignment Reviews
Use NIST AI RMF as voluntary risk guidance. It is not a legal standard or certification. Review applicable duties with legal counsel.
Risk-Tiered Review Thresholds
Define escalation thresholds by output risk level: Low (automated pass-through with sampling), Medium (team lead review), High (senior advisor sign-off required before delivery).
Reviewer Qualification Standards
Establish minimum qualifications for AI output reviewers by domain: lease analysts for abstraction outputs, senior brokers for market intelligence, partners for client-facing proposals.
Structured Approval Workflows
Implement documented approval chains with digital sign-off. No AI-generated deliverable reaches a client without at least one qualified human review and approval.
Feedback Loop Integration
Capture reviewer corrections and client feedback systematically. Feed validated corrections back into model fine-tuning and prompt optimization cycles.
3 pillars / 12 practices to validate during the 2-week audit
Audit PrerequisiteData Privacy Controls
Protect sensitive client and deal information by design, not by exception. Classify data before it enters any model, restrict access by role, and enforce encryption in transit and at rest across every workflow.
Auditability Requirements
Treat every AI-assisted deliverable like an investment memo: traceable, defensible, and reviewable. Log source documents, model versions, and decision timestamps so recommendations can be reconstructed during client questions, legal review, or internal QA.
Human Review Protocols
AI can draft, but advisors remain accountable. Define risk-based review thresholds, assign qualified reviewers, and require explicit sign-off before any client-facing output is delivered.
What We Know, What We Need to Validate
Documented observations
The cited sources document JLL survey responses, the Falcon announcement, vendor descriptions, and voluntary NIST guidance. Their scope and limits matter.
What needs local proof
Measure current cycle times, output errors, data access, review effort, and team use. Test client-facing outputs before adopting them.
Documented observations
Among occupier CRE respondents, 92% started or planned AI pilots in 2025; 5% reported most program goals achieved. Separate measures, not a failure rate.
JLL announced its CRE AI platform on 29 October 2024. The announcement does not prove RFP speed gains.
Vendor claim; test with your leases. This marketing page is not independent controlled research.
Vendor observation across 1,280,657 proposals and 27 industries. Not a CRE benchmark or proof of an AI effect.
The AI RMF is voluntary. It is not a certification or proof of legal compliance.
What Remains Hypothesis
Actual cycle times, error rates, and throughput in occupier advisory practice are unvalidated without direct assessment.
We cannot confirm which processes cause the most friction without stakeholder interviews during the audit.
System integration feasibility, data quality, and format compatibility require technical discovery.
Change readiness, skill gaps, and cultural receptiveness are internal factors that remain unassessed.
How technology, law, and financial services clients would respond to AI-assisted outputs is hypothesis only.
Returns remain unproven. Measure actual saved hours, output quality, and all costs in a local trial.
Define the next test
Use local evidence to decide whether to proceed, change scope, or defer. An assessment can define tests; it cannot guarantee their results.
The decision: assess and test
Recommendation: agree on scope and test criteria. The two-week schedule below is illustrative.
The Assessment Framework
The Assessment Framework
An illustrative two-week schedule. Agree on scope and access first; timing and results are not guaranteed.
Week 1: Discovery
Foundation building and current-state assessment
Stakeholder Interviews
8-10 sessions with partners, associates, and operations staff to surface pain points
Baseline Metrics
Capture current cycle times, error rates, and throughput for key workflows
Data Quality Assessment
Evaluate data readiness, access controls, and integration requirements
Governance Review
Assess compliance, approval workflows, and privacy protocols for AI
Week 2: Analysis
Synthesis, prioritization, and strategic roadmap
Opportunity Scoring
Rank by impact x feasibility x risk x time-to-value with explicit assumptions
ROI Modeling
Build conservative, realistic, optimistic scenarios with break-even analysis
Implementation Roadmap
Create 90-day milestone plan with success metrics and phased delivery
Executive Presentation
Deliver findings with actionable recommendations and go/no-go framework
Two-Week Timeline
A proposed schedule, subject to scope and availability
Current offer
Local returns
Unproven
Costs and returns
Measure actual saved hours and all costs, including software, integration, training, human review, and maintenance. Redeployed capacity is not automatically cash savings.
Choose the next step
Use test results to decide whether to build, buy, change scope, or defer. See the current Assessment offer.
Should we discuss the proposed assessment now or revisit later?
Agree on scope, access, and test criteria before choosing a schedule.
Download Full Report
Get the complete strategic intelligence brief as a PDF for offline review and sharing.