On November 2nd, we stood in front of the room at the Our Legal Community AI certification course in Silicon Valley. It was our second run, after London in June. The conversation had moved a long way.
Many use cases were floated, but one use case were mentioned by several folks as their top priority: solving intake.
If this is solved the way that in-house teams plan, this has immediate consequences for law firms. This blog post plays out how solving intake cuts revenue to law firms.
Note: interesting aside, on Oct 13th, both Harvey and Legora launched their intake tools, both named “intake”. By virtue of this, we can say that intake is the hottest problem in town right now.
Eliminating 80% of legal questions
For in-house teams, intake tools are built to serve the business teams — sales, HR, procurement, etc. — they get a “front door” where they can find answers to legal questions, or get routed to the right person when the question needs a lawyer.
Someone at the OLC training day mentioned that they aim to stop 80% of queries from ever reaching the legal team. The perfect system would answer most questions via self-served (e.g. "here's the approved template"), with only a tiny fraction going to the legal team.
Simulating the Goldilocks intake tool
Let's assume there is a perfect intake tool (whether AI, agentic, or otherwise). It handles every simple query flawlessly. It is not sophisticated enough to touch the complex work, so it routes all of that to the legal team. Not too hot, not too cold.
Under that assumption, the volume of complex legal work does not change at all. What changes is how much time the in-house team can allocate to different tasks.
Take a hypothetical in-house team of 10 lawyers with 18,000 hours a year between them, a $6,000,000 external legal budget, and 5,000 queries from their business team every year.
The first key observation: 80% of tickets is not 80% of time. Simple questions are quick one by one. If 4,000 simple queries take about 1.8 hours each on average (including the possible back and forth and phone calls), that's a total of 7,200 hours, or about 40% of the team's capacity.
The second key observation: the agent needs verification. Someone has to check its answers, keep the playbook current, and pick up its escalations. Let’s assume that is 900 hours a year.
The net result is 6,300 hours back (or, roughly 3.5 full time lawyers). Here's where that time might go:
Simple queries (the “80%”): 7,200 hours —> 900 hours per year
Complex work done in-house: 7,200 hours —> 10,000 hours per year
Briefing out and supervising external counsel: 2,700 hours —> 4,200 hours per year
Strategy and governance: 900 hours —> 2,900 hours per year
We have assumed that lawyer hours are fungible, and can be reallocated without retraining.
Who does which work?
When an in-house lawyer is drowning, they send the whole matter to an external counsel.
"Here's the deal. Please deal. Kind regards."
The firm decides the staffing: a partner, a senior associate, plus three juniors on due diligence. The client pays for all of it, because nobody in-house controlled which parts they needed. If we give the in-house lawyer a day back, they can read the matter properly before it goes out, and they carve it up:
Use AI to run the first pass of diligence.
Use AI to draft the disclosure schedules automatically.
Use AI to run the research.
Use in-house capacity to manage the workflow and matter.
Call up the law firm for advice on the negotiation strategy, the judgement, and their read of the market.
Across a year, external work splits three ways:
Some never goes out. The matter is now within the team's capacity, so it stays in-house.
Some goes out, in unbundled pieces. The firm gets the parts that needs its expertise.
Some work never gets done at all. Once someone has time to look closely, a slice of the work turns out to be unnecessary.
What happens to law firm revenue?
In our hypothetical, assuming law firms charge a blended rate of $900 an hour and charge $600 an hour for pure junior work, the change to external spend might look like this:
Starting spend: $6,000,000
Whole matters brought in-house (estimated to be 800 hours): $720,000
Junior work is unbundled and done in-house (estimated to be 2,000 hours): $1,200,000
Work is no longer done because it is unnecessary (estimated to be 500 hours): $450,000
Resultant estimated legal spend: $3,630,000
Law firms change their billing model to try to capture more value (estimated to be +20% of last year’s fees): $4,356,000
That a 27.4% drop to external legal spend, triggered by one (very good, Goldilocks) intake tool.
Of course, this is based on a lot of assumptions. For the math, we are assuming:
The 3.5 full time legal staff were redeployed to different workflows, so the in-house team still had a total capacity of 18,000 hours per year.
Easier access to legal answers via the intake tool does not create more complex legal work for the in-house team.
The intake tool works exactly as advertised (which is a big “if” and probably deserves a whole lot more analysis).
The battle for legal spend
The winner certainly is not the incumbent law firms. Even with a 20% repricing, they collect much less. Their existing leverage model, where partners sell judgement and juniors sell hours, is exactly what gets unbundled.
The winner is not the in-house team either. The lawyers get some time back, but not the financial savings. Lower external spend this year becomes a lower budget next year. Lawyers are made to do more complex work with the same headcount.
If the market continues the trend in the same direction, the bulk of the financial upside might go to whoever turns the work into “products”. AI native law firms can sell diligence, disclosure schedules, and research as fixed-fee services, without the cost structure of traditional firms. Software companies can sell the tools on both sides: intake for in-house teams, diligence engines for law firms, and verification and playbook maintenance as managed services.
