KEY TAKEAWAYS
- LPs increasingly use AI adoption as a proxy for operating quality during due diligence.
- The new ILPA Reporting Template 2.0 makes fees and expenses comparable across managers, putting a spotlight on your overhead.
- Sponsors should expect questions on headcount per $1B of assets, automated functions, audit trails, and AI error management.
- Strong answers require 12 to 18 months of actual operating work; you cannot fake an audit trail or a lower headcount ratio during a meeting.
IN THIS ARTICLE
For decades, raising a real estate fund came down to track record, relationships and the deal pipeline. That’s changing. Limited partners (LPs), the pension funds, endowments, family offices and wealth platforms that invest in funds, are now asking general partners (GPs) how they use CRE AI.
They’re not asking out of curiosity. LPs increasingly see AI adoption as a proxy for operating quality: whether a sponsor can control costs, scale without bloating overhead and keep up with competitors. A sponsor who can’t answer these questions well signals operational risk.
Here are the 10 AI questions middle-market CRE sponsors should expect in their next fundraise, what each one is really testing, and what a strong answer sounds like.
Why LPs are asking about AI for commercial real estate now
Three shifts have put AI on the LP due diligence checklist.
Fee transparency is now standardized. The Institutional Limited Partners Association (ILPA) released version 2.0 of its Reporting Template in January 2025, its first update since 2016. Funds now report fees, internal chargebacks and partnership expenses in a fixed format that LPs can compare across managers. For the first time, an LP can line up your overhead against a peer’s, line by line.
Capital is concentrating. According to PitchBook, 46% of private-fund capital raised in 2025 went to just 10 funds. Middle-market sponsors are competing for a shrinking slice, and cost efficiency is one of the few ways to stand out.
Fees keep falling. Average private equity management fees hit a record-low 1.61% in 2025, according to Preqin. LPs know that sponsors with leaner, AI-enabled operations can accept lower fees and still stay healthy. They want to know whether you’re one of them.
These norms start with the largest institutional LPs and move down market. Family offices and wealth platforms increasingly use the same ILPA templates and diligence questions when they evaluate smaller sponsors.
The 10 CRE AI questions LPs will ask
1. How many people do you employ per $1 billion of assets, and how has that changed?
What they’re testing: whether your productivity is improving or drifting with the industry. The NAREIM/Ferguson Partners benchmark puts the median investment manager at 9 full-time employees per $1 billion of assets, excluding property management staff.
Strong answer: “We were at 12 in 2024, we’re at 8 today, and we’re targeting 6 by the end of 2028. Our head of AI operations owns that target.”
Weak answer: “We don’t track that. Our headcount plan is built property by property.”
2. How does your G&A compare to the closest public REIT, and how much of it is variable?
What they’re testing: whether your overhead (general and administrative costs) can scale with assets without growing at the same rate.
Strong answer: “Our G&A is 4.1% of revenue, versus 5.8% at our closest public comparable. More of it is variable today than two years ago, because AI handles work we used to hire for.”
Weak answer: “Our G&A grows roughly in line with assets.”
3. Can you itemize your internal chargebacks under ILPA 2.0, and which savings come from AI?
What they’re testing: whether you’ll pass the new transparency standard without embarrassment.
Strong answer: “Here’s the itemization. We can show which line items are lower because of AI and which are traditional overhead.”
Weak answer: “We’re still working on our ILPA 2.0 reporting.”
4. Which functions are automated today, and how much capacity does that add?
What they’re testing: whether you’ve moved from AI pilots to production. According to JLL, 92% of CRE teams are piloting AI, but only 5% have achieved all their AI program goals.
Strong answer: “We have AI workflows in production across leasing, accounting, investor reporting and lender reporting. Each runs with human review, and together they add the capacity of about a dozen employees.”
Weak answer: “We have several pilots underway. Nothing is fully rolled out yet.”
5. Which functions are still done only by people, and why?
What they’re testing: whether your remaining cost base is a deliberate choice or just legacy.
Strong answer: “Investment committee decisions, investor relationships, lender negotiations and anything licensed or legal. Everything else has AI doing the first pass, with a person reviewing it.”
Weak answer: “Most of our functions are still manual because we haven’t figured out the AI part yet.”
6. Show us the audit trail for an AI action from the last 90 days.
What they’re testing: whether your AI governance is real or just talk. LPs want to know who approved an action, what changed and what would have happened if the reviewer said no.
Strong answer: “Here’s an approval record from last month showing the reviewer, the inputs and the result. Here’s our escalation policy and how we train reviewers.”
Weak answer: “We haven’t built out that governance yet.”
7. How do you manage the risk of AI errors in valuations, forecasts and investor reporting?
What they’re testing: whether you’ve thought about AI mistakes, inconsistent outputs and model changes, or are exposed to them.
Strong answer: “Every consequential output is reviewed by a person before it’s used. We cross-check results across more than one AI model, monitor quality over time and test the system independently every quarter.”
Weak answer: “We rely on our vendor’s quality controls.”
8. How has on-site and staff retention changed, and how much is due to AI?
What they’re testing: whether AI is making your firm a better place to work or creating turnover risk.
Strong answer: “Retention is up since 2024. We credit removing repetitive work and the better schedules that after-hours AI coverage made possible.”
Weak answer: “We haven’t looked at that.”
9. Will you commit to a productivity target in your next LP letter?
What they’re testing: whether you’ll hold yourself accountable to a measurable operating goal.
Strong answer: “Yes. Our 2029 target is a specific number of employees per $1 billion of assets. It’s in our annual letter, and we’ll report progress every quarter.”
Weak answer: “We prefer not to publish operating metrics.”
10. Compared to a peer of similar size and strategy, what operating advantage will we see in three years?
What they’re testing: whether you can describe a real, differentiated position for 2029 and beyond.
Strong answer: “Our closest peer has no board-level AI plan. Three years out, our operating cost per unit will be meaningfully lower than theirs on the same asset class, and here’s the plan that gets us there.”
Weak answer: “We deliver strong performance. Operating comparisons vary by market.”
You can’t fake these answers: how to prepare
Every strong answer above depends on work that’s already done. You can’t talk your way to a lower employee-per-asset ratio, a real audit trail or a retention improvement in the middle of a diligence call. Those take 12 to 18 months of operating work to build.
That means the time to prepare for your next fundraise is now. A practical starting point:
- Score yourself honestly. Go through all 10 questions and write the answer you’d give today. Mark each one strong or weak.
- Fix the governance first. Questions 6 and 7 are about human review and audit trails. Build that layer before you scale AI, not after.
- Get workflows into production. Pick a handful of high-volume workflows, such as investor reporting, rent collection follow-up or month-end reconciliation, and run them live with human review.
- Start measuring now. Questions 1, 2 and 8 need a baseline. Record today’s headcount per $1 billion of assets, G&A ratio and retention so you can show the trend later.
- Name an owner. Put one executive in charge of AI operations so there’s a name to give when LPs ask who’s accountable.
Sponsors who do this work now will walk into their next LP meeting with answers. Those who wait will walk in with explanations.
Frequently asked questions about CRE AI and LP due diligence
Do LPs really ask about AI for commercial real estate in due diligence?
Yes, and increasingly. Institutional LPs and their advisers now ask whether sponsors use AI in reporting, leasing, pricing and asset management. Family offices and wealth platforms are adopting the same diligence frameworks.
What is the ILPA Reporting Template 2.0?
It’s the Institutional Limited Partners Association’s standardized format for reporting fund fees and expenses. Version 2.0, released in January 2025, requires itemized internal chargebacks and partnership expenses in a format LPs can compare across managers.
How does AI in commercial real estate affect fundraising?
Sponsors with AI-enabled operations can show leaner overhead, scale without proportional cost growth and offer more competitive fees. Those are exactly the things LPs compare under the new reporting standards.
How long does it take to prepare strong answers to LP AI questions?
Usually 12 to 18 months, because the answers depend on results like lower overhead, workflows in production and measurable retention changes.
Your next fundraise starts now
LPs are adding CRE AI to the list of things that separate great sponsors from average ones. The sponsors who answer these 10 questions with confidence will have a real edge in a market where capital is concentrating in fewer hands.
The good news is that every one of these answers can be built. It just takes starting before the LP meeting, not during it.
Want to walk into your next fundraise with strong answers?
Sophia CRE helps CRE sponsors put AI workflows into production with human review and a full audit trail built in. Book a conversation to see where your firm stands today.
Real audit trails. Real human review. Real results for your next LP meeting.
