The $200/mo paralegal is becoming a $30/mo subscription.

Most brokerages still allocate a paralegal, or a junior associate doing paralegal work, to keep up with lease abstraction. The fully-loaded cost is usually $60-80K a year, allocated across 15-30 brokers. Per broker, that's something like $200-300 a month of overhead, before you count the time the broker spends QA-ing the work.

That cost is going to $30 a month. We don't think this is controversial. We think every brokerage will look at it the same way every law firm has already looked at it: yes, the AI will write the first draft of the document, and yes, the senior person will review it, and the line item that goes away is the junior person who used to write the first draft.

The pricing-power question

The interesting question isn't whether this happens. It's how the savings get distributed.

If you're an AI vendor, you have a choice: charge usage-based prices that capture most of the savings, or charge flat subscription prices that leave most of the savings with the broker. CoStar took the first approach in their original product, charge per-seat, charge per-search, capture the data leverage. The new generation of AI-native tools, including DealDesk, is taking the second approach. Flat $30-49 per broker per month, all-you-can-abstract.

The reason for that pricing model isn't generosity. It's that flat pricing wins distribution. If a broker can sign up, abstract 200 leases in a month, and pay the same $30 they'd pay for one abstract, they'll bring it to their team. If the price scales linearly with usage, they'll think twice about every click. Flat pricing makes the tool free at the margin, which is the only way to actually displace the existing paralegal workflow.

Why $30 instead of $5

We get this question from price-sensitive brokers. Why isn't this $5 a month? The answer is that the floor is set by the inference cost. A reliable lease abstract, the kind where a broker can stop QA-ing every line, runs through a frontier model with extended thinking and citation grounding. That's not a $0.10 inference. It's closer to $1-2 of compute per abstract, and a working broker abstracts 5-15 leases a month.

$30/month is roughly cost-plus on the inference, with margin to fund the engineering team and the customer success ops. It's not a charity. It is, however, a tenth of what the incumbent paralegal stack costs, and that's the comparison that matters.

What this does to firm overhead

If you're running a 30-broker firm, the math: $200-300/broker/month of paralegal cost goes to $30-50/broker/month of AI subscription. That's $50-75K a year of recovered margin per firm. For a boutique with sub-10% net margins, that's the difference between a fine year and a great one.

It's also the difference between hiring an additional broker and not. Most firm operators we talk to are reinvesting the recovered margin into headcount, not banking it. The paralegal economy isn't shrinking the firm. It's shifting where the people sit.