Everyone's Piloting, Almost Nobody's Finished

88% of CRE firms are running AI pilots. 5% have actually finished one. Here's the gap that matters in 2026.

We ran the numbers on where CRE is actually at with AI in 2026, using what operators asked us for instead of a survey we paid for. The read is uncomfortable in both directions, and it changes what you should do next.

- Sasha Deneux

The Take: being "in" on AI stopped being the flex

For two years the anxious question was "are we even doing AI yet." That question is dead. Per JLL's 2025 Global Real Estate Technology Survey, published October 28, 2025 and drawn from more than 1,500 decision-makers across 16 markets, 88% of investors and owners have started piloting AI, up from under 5% in July 2023. In under three years, piloting went from a rounding error to near-universal. So if you feel behind, read the next number before you panic.

In that same survey, only 5% of firms report achieving all of their AI goals. 47% got to two or three. Firms are running roughly five pilots at once across 56 use cases, and 87% say their tech budget went up because of AI. So the spend is real, the pilots are real, and the finished work is rare. Everybody's on the field. Almost nobody's crossed the line.

That reframes the "am I behind" feeling. You are not behind the pilots. You are behind the results, and so is nearly everyone else. JLL's own read backs it: more than 60% of firms call themselves unprepared to scale past pilots. So here's the operator move. Stop counting pilots. Pick the one workflow where a finished loop changes your week, and finish that one. A license is not a system. The gap between 88% and 5% is the gap between owning the tool and owning the loop. We wrote the pilot-to-production framework up in The CRE Tech Leader's Guide to AI Implementation.

The Teardown: reading your own demand honestly

We just published our 2026 state-of-AI read, and I want to walk the method, because most "state of AI" reports are a vendor survey wearing a data costume, and you can build a far more useful read from what people actually ask you for. Here's the whole thing end to end.

Start with first-party demand, not a survey. Over the last several months, 8,092 distinct operator sessions viewed or requested our free CRE-AI templates. That's real hand-raising, not a poll where people say what sounds smart. Every count is "viewed or requested," never "downloaded," because the honest label is the load-bearing one.

Cut it by who's asking. Sorted by persona, investor and acquisitions sessions outnumbered broker sessions roughly 4.5 to 1: 4,853 versus 1,066 distinct sessions. Developers were 311, institutional and LP 48. One discipline point here: we publish the ratio, not percentage shares, because the buckets are topic-affinity groupings and a single operator can land in more than one. Shares would fake a precision we don't have. The ratio is honest.

Then read what the buy side actually wants AI for, from real calls. This is the finding that surprised me. When acquisitions teams tell us what they want, they name sourcing before underwriting. In our recorded calls, off-market sourcing and owner-finding came up in 13 distinct conversations. Underwriting and pro-forma automation came up in 6. Those are floors, the count of calls where the intent was explicit, never a guess at hidden demand. Underwriting matters. It's just not the first thing operators reach for. Finding the deal is.

Layer the outside surveys on top, attributed and dated, as context, never as your own number. JLL for the adoption gap. Your first-party signal is the spine, third-party surveys are the frame, and you never blur the two.

State the bias out loud. Ours is LinkedIn-weighted (57% of views came through LinkedIn), and the curve is a distribution spike with a long tail, not organic month-over-month growth. So we never say "adoption is accelerating" from it. Naming the bias is what makes the number usable instead of decorative.

The honest caveat, same as we tell clients: this is demand, not a forecast. It tells you where operators are pointing AI right now, not where transaction volume is moving. But the direction is actionable. If your buy side is going to finish one loop this year, the data says point it at sourcing, because that's where the demand stacks. The built version of that workflow is The Off-Market Sourcing Agent, free, the full 7-stage architecture.

Signal

JLL: 88% of investors and owners are piloting AI, but only 5% report hitting all their goals; 47% got two or three (2025 Global Real Estate Technology Survey, published October 28, 2025, 1,500+ decision-makers, 16 markets). Why it matters: this is the whole story of 2026 in two numbers. Adoption is wide, completion is shallow, and the readiness wall between them is where the next year gets decided.

We launched the Off-Market Deal-Flow Index for Q3 2026, mapping where CRE buyers say they're hunting off-market, with the sample sizes printed unflinchingly (n=1 rows and all). Why it matters: it's demand, not transaction data, and we say so. Off-market sourcing was the single most-stated ask across our calls at 13 distinct conversations, and the US Midwest was the most-repeated single region. If that's your lane, it's a contested one.

Deloitte: roughly 75% of respondents plan to increase real estate investment over the next 12 to 18 months, and Deloitte points at reliable data and application readiness, not enthusiasm, as the path to getting value from AI (2026 Commercial Real Estate Outlook, published September 29, 2025, 850+ C-level respondents, 13 countries). Why it matters: the capital's coming back. The firms that turn it into returns are the ones whose data is ready to feed a model, not the ones with the most enthusiasm.

EliseAI's multifamily read: 92% of multifamily operators have implemented AI, 32% think competitors are moving faster, and 72% fear slow adoption hurts NOI (State of AI in Multifamily, published September 26, 2025, 280 executives; vendor survey, multifamily lane). Why it matters: the peer-anxiety is documented, not invented. But "implemented" is not "finished," and fear of falling behind is a worse compass than picking one loop and closing it.

From NextAutomation

The move this issue argues for, finish one loop, and point it at sourcing, is exactly the kind of engine we build and run with acquisitions teams: the off-market sourcing system, wired into your stack, refreshed on a schedule, with a human approving every contact. If your firm is sitting at "we've piloted it" and wants to get to "it runs every week," book a call and bring your buy box. Want to read the full 2026 read first? It's the state of AI in CRE report.

Book a call

Everyone's piloting. Almost nobody's finished. The firms that pull ahead this year won't be the ones with one more experiment, they'll be the ones who took a single loop all the way home.

Next week: the inbound side of sourcing, a practical walkthrough of what to prioritize when the deals come to you.

- NextAutomation Team