AI Sales Reps
A 32% Win Rate at 27% of the Work: Reading That Benchmark as a Real Estate Broker
Brokers react to the win rate. The number that should change how you staff a lead desk is the effort figure behind it. A five-step method for running an agency lead desk against a near-parity benchmark.
by Jerrod Anthraper
Two numbers, held together. Above a 32% win rate, achieved with roughly 27% of the work a human sales team would have spent getting there. That is the benchmark our framework had to clear across six to nine weeks of testing before we were willing to put it in front of a paying client.
Brokers react to the first number. The second one is the number that should change how an agency staffs its lead desk, and it is the one almost nobody asks about.
Near parity is not the impressive part
A 32% win rate is not a shocking figure on its own. Plenty of good agents beat it on well-sourced leads, and any broker who has been doing this a while has an agent who beats it consistently.
The interesting part is the denominator. Near-human results at roughly a quarter of the effort means the binding constraint on your lead desk was never talent. It was attention — a finite number of agent-hours getting spread across a lead volume that does not respect anybody's schedule. When the effort cost of covering a lead drops by three quarters, coverage stops being a rationing decision and starts being a default.
Most agencies never see this, because they measure the lead desk in closings, and closings are a lagging, heavily filtered number that arrives 30 to 90 days after the decision that caused them.
The leak is coverage, not conversion
Real estate demand is structurally nocturnal and weekend-heavy. People browse listings after dinner, tour on Saturday, and call about a yard sign on a Sunday afternoon while parked across the street from it.
Somewhere between 40% and 60% of inbound leads arrive outside business hours and typically go unanswered. In most agencies those leads are not lost dramatically. They sit in a queue until Monday, get a call from an agent who is already behind, and quietly convert at a fraction of the rate of a lead answered on Saturday at 2:40 p.m. while the caller is still standing in front of the house.
The agency sees a conversion problem and buys better leads. The actual problem was that nobody was there.
Why agencies mis-diagnose this
Two reasons, both structural rather than personal.
First, the people who could answer are the people worst positioned to. An agent at a showing cannot pick up. An agent at a closing table cannot pick up. An agent's busiest hours and the lead desk's busiest hours are the same hours, and that conflict never resolves on its own — it just gets absorbed into a vague sense that leads were bad this quarter.
Second, most brokerages have no owner for the lead desk. Leads route to agents individually, so nobody ever sees the aggregate. There is no screen where a broker can see that 41 inquiries came in over the weekend and 12 of them got any response at all, because that number lives scattered across fourteen agents' phones.
Five steps to run a lead desk against the effort benchmark
1. Define the win before you measure anything
For a brokerage, the win at the lead-desk stage is the appointment set and attended: the showing, the listing consultation, the buyer consult. Not the closing.
If you measure your lead desk in closings, you are measuring your agents' closing ability and your lead desk's response quality tangled together, on a three-month delay. Separate them cleanly. The lead desk owns appointments set. The agent owns what happens inside the appointment. Now you can tell which one is actually broken.
2. Split your lead flow into lanes and fix the worst one first
Most agencies have three distinct lanes: portal inquiries, sign and listing calls, and sphere or referral introductions. They behave differently, they arrive at different hours, and they deserve different handling.
Score each lane on one thing — what percentage got any response inside fifteen minutes last month. Then automate or restaff the worst lane and only the worst lane. Doing all three at once guarantees you will not be able to tell what worked.
3. Write down five qualification fields, and stop there
The temptation is to build a long intake script because more data feels safer. Resist it. On a first contact, five fields are enough to route intelligently: buy, sell, or both; the area or the specific address; timeline; financing status (cash, pre-approved, not yet); and two windows of availability this week.
Anything past those five belongs to the agent, and asking it too early costs you the lead. A first contact is not an intake form. It is a booking with just enough context attached to be useful.
4. Name one escalation trigger and one handoff format
The trigger should be a single written rule. For most offices it is a seller asking about a listing appointment, or any caller naming a specific address they want to see this week. When the trigger fires, a named person's phone rings — not a shared inbox.
The handoff format matters as much as the trigger. An agent should receive the five fields, the appointment time, and one line of context, in something they can read at a stoplight between showings. If your handoff is a raw call transcript, agents will stop opening it by the second week.
5. Measure the effort side, not just the wins
Track agent-hours spent on leads that never became an appointment. If you cannot measure it cleanly, estimate it — ask three agents to log a week honestly and multiply.
That number is your real baseline and it is the one that moves first. Appointment volume usually takes a few weeks to shift. Reclaimed hours change immediately, and reclaimed hours are what let a ten-agent office behave like a twenty-agent office without adding headcount or desk fees.
The proof
Data point: across six to nine weeks of testing, the framework held above a 32% win rate while requiring roughly 27% of the work a human sales team would have spent — near parity on results, at a fraction of the effort.
Near parity is the honest claim, and it is worth being precise about what it does and does not mean. It does not mean an automated first contact will outsell your best agent in a listing presentation. It means the first-contact job — answer, qualify, book — can be done at close to human quality without consuming human hours. That happens to be exactly the job your agents are worst positioned to do at 8 p.m. on a Saturday.
What this looks like running
James, the AI sales agent we deploy for agencies, is built to execute that specification: answers inbound inquiries in 60 to 90 seconds at any hour, works the five fields, books to the right agent's calendar, escalates on the trigger you defined, and hands off in a format an agent can actually read between showings. After the transaction closes, the same system runs the review request and the referral ask that most agents mean to send and never do.
If you want a read on whether any of this applies to your office, run step five before you talk to a vendor. Ask three agents to log, for one week, the hours they spent on leads that never turned into an appointment, then multiply by your headcount. That number is what the effort benchmark is describing. If it comes back small, you have your answer and you can skip the call entirely.