What 200 working tenant reps say about AI in their workflow.

Over the last six months we've sat down with roughly 200 working tenant-rep brokers, boutique founders, JLL/CBRE seniors, the Cushman teams in the secondary markets, and asked the same five questions about AI in their workflow. The answers are more consistent than we expected, and not always what the trade press has been writing.

What's actually in their workflow

Lease abstraction is universal. Every broker we talked to has either started using an AI abstraction tool or is in the procurement process. The breakdown was roughly 40% on a paid AI tool today, 35% experimenting with the free tiers of one or more, 25% still on the paralegal-team workflow but planning to switch within the year. Two years ago that 40% number would have been zero.

LOI drafting is the second-most-adopted category, but adoption is much shallower. Brokers told us they'll use AI to draft a first LOI on a new requirement, then hand-edit every counter from there. The reason is straightforward: an LOI gone wrong has bigger consequences than a lease abstract gone wrong. The trust threshold is higher. We expect this to shift as the drafting tools get better, but it'll be slower than abstraction.

What's not in their workflow yet

Comps research is still mostly the legacy CoStar workflow, even at the firms that are otherwise enthusiastic AI adopters. The reason: nobody has fully replaced the depth of the historical database, and brokers don't trust an AI summary of comps the way they trust a database query they can audit. This is changing, the AI-native tools are building their own comp inventories, but it's a 24-36 month transition, not a 12-month one.

Prospecting and lead-gen tools are universally tuned out. Every broker we asked about "AI prospecting" had either tried one and stopped, or had heard from a peer who tried one and stopped. The signal-to-noise ratio is too low. Real prospecting is still relationships and referrals, and brokers do not believe that's about to change.

What surprised us

The most interesting finding: brokers consistently underestimate how much AI they're actually using. We'd ask "do you use AI in your work?" and they'd say "not really." Then we'd ask about specific workflows and find that the same broker uses ChatGPT or Claude 3-5 times a day for email drafting, market research summaries, and quick calculations. The AI use is below the conscious threshold for most working brokers.

This matters because it predicts the second-order adoption pattern. Workflow tools that lean into the existing ChatGPT/Claude habit (via MCP connectors and integrations) will spread faster than workflow tools that ask brokers to learn a new interface. The brokers are already in the chat window. Meet them there.

What firm leaders should take from this

Three implications. First, lease abstraction adoption is past the tipping point, if your firm hasn't picked a tool, you're behind. Second, LOI drafting is the next category to watch, but the trust gap is real and you should expect a slower rollout. Third, the integration story is more important than the model story. The vendors who win the next 18 months are the ones who plug into the existing tools brokers already use, not the ones who try to replace them.