← The Intake
The Intake · Weekly briefing

This week in legal AI

Harvey and Legora launched same-named "Intake" front doors for in-house legal requests within hours of each other, Microsoft added legal skills to Copilot, and AI pricing is edging towards pay-per-use, so the question for legal teams is who routes the work and who reviews it.

Week of October 10 – October 16 2026
Category Market intelligence
Reading time 6 minutes
01 · The week at a glance

Legal's front door became a product category this week, and AI is moving to pay-per-use

October 8 · Regulation
Consultation opened October 8 and runs to January 14 2027.
October 13 · Product
Legora's agent logs requests and prepares work for lawyers to review.
October 14 · Product
Early access begins this month, with Carvana, GE Aerospace and Dentsu as design partners.
October 14 · Pricing
Law.com reports firms building tools that can switch between models as prices change.
October 15 · Product
Harvey, Legora, Docusign and Box are among the partners named for plugins.
02 · Product

Harvey and Legora both launched an AI front door for in-house legal requests, hours apart

Legora announced Intake on October 13 and Harvey announced its own Intake the next day, which Legal Cheek called an awkward coincidence. Both promise one place for the business to send legal requests, with an AI agent doing a first pass before a lawyer sees the work. Neither names a customer result.

 
Legora Intake
Harvey Intake
Where requests arrive
LegoraSlack today. Teams, email and MCP are planned.
HarveyEmail, Jira, Slack, ServiceNow, iManage, Google Drive, SharePoint and more, per Harvey's product page.
What the agent does
LegoraLogs the request, asks for missing information, prepares work using the team's Skills and playbooks.
HarveyRoutes each request to an agent, a solution or a lawyer, and takes a first pass at contracts.
Availability
LegoraAvailable now in Slack.
HarveyEarly access begins in October.
Named customers
LegoraNone named
HarveyCarvana, GE Aerospace and Dentsu as design partners
Results published
LegoraNone
HarveyNone. Harvey says it works with 1,600 in-house teams, a figure from the company.

The one customer comment comes from Carvana's Seena Ghebleh, who told Artificial Lawyer the tool "can apply our playbook" and leaves her an initial review to sign off. That is a design partner talking about an early-access product. Neither company says how often the first pass is wrong, or what happens to a request the agent cannot place.

The Flank read

A front door is the right instinct, because the problem is routing, not drafting. Inexpensive work, standard NDAs and low-risk redlines, still lands on expensive resources because nothing sorts it on arrival. But both launches are products that a legal team configures and then supervises itself. Outsourcing that work to supervised agents means the sorting, the playbook and the escalation rules are run for you, with a lawyer reviewing finished work rather than a queue of first drafts.

03 · Product

Microsoft added more than 20 legal skills to Copilot, putting playbook review inside Word

On October 15 Microsoft announced Copilot for legal, built on its earlier Legal Agent and delivered as two plugins: one for Word, one for the multi-step Cowork workspace. In Word it reviews agreements against playbooks, drafts from approved templates and applies proposed revisions as tracked changes. Cowork extracts obligations and deadlines and compares agreements across collections.

An hour, then several minutes
Aneesh Mehta, a senior corporate counsel at Microsoft, said a review that might take an hour can often take several minutes. He works at Microsoft and is describing its own product.
Vendor-reported · no pricing or availability stated

Microsoft also named partners building plugins that bring specialist data into Word and Cowork, within each organisation's access controls. Two of them, Harvey and Legora, sell competing products.

6
Legal research
Bloomberg Law, Trellis, Descrybe, Legal Data Hunter, Courtroom5, Lawstronaut
7
Documents and matters
iManage, NetDocuments, Box, Litera, HighQ, LegalZoom, Solve Intelligence
7
Agreements and e-signature
Docusign, Juro, Adobe Acrobat Sign, LawVu, SignatureAPI, Dropbox Sign, PandaDoc
4
Legal AI and e-discovery
Harvey, Legora, Relativity, Everlaw
The structural question for enterprise legal teams

If contract review comes bundled into the software your lawyers already draft in, what are you still paying a separate vendor for? A review skill makes one lawyer faster. It does not decide which contracts should reach a lawyer at all, and that decision drives the cost.

04 · Structural

Pay-per-use AI pricing is arriving, and legal teams will need to budget for variable costs

Law.com reported on October 14 that law firms are planning for consumption-based AI pricing, after Anthropic moved that way this year. Troutman Pepper Locke is building in-house tools with mechanisms to route to different models ahead of pricing changes. The article is paywalled beyond its opening and the visible text gives no price points.

Per seat
What you pay for
Access, per named user per month
What drives the bill
Headcount
Who watches it
Usually fixed in the contract
Per use
What you pay for
Work done: tokens, tasks or credits
What drives the bill
Volume, and which model handles each task
Who watches it
Someone who can see usage as it happens

One vendor has already built for this. Jylo 3.0, announced October 13, adds credit management that lets administrators allocate budgets, set usage limits and decide which models each team or person can use. It can also give models custom labels that hide the underlying foundation model. Jylo's CEO Shawn Curran said "The customer should be in control." These are vendor descriptions of a release going first to early adopters.

The structural question for enterprise legal teams

Under per-use pricing, a model that costs more per task is a cost every time it touches a standard NDA. Who in your department decides which work needs the expensive model, and which should never have reached it?

05 · Regulation

The UK Law Commission proposes treating AI systems as products under product liability law

The Law Commission of England and Wales opened a consultation on October 8 to update product liability law, which dates from almost 40 years ago and covers only physical goods. Its provisional proposal widens "product" to include software and AI systems, however they are supplied. It is a proposal for consultation, not law, and it is UK only.

October 8 2026
Consultation opens
Provisional proposals cover software, AI systems and online marketplaces.
January 14 2027
Consultation closes
Responses go in through the Commission's online form.
Late 2027
Final recommendations expected
Any change to the law would then need government action.

The Commission's project page lists online marketplaces, fulfilment service providers and UK authorised representatives as potentially liable parties and does not list developers. It also proposes clarifying liability where a product is substantially modified after supply, without saying whether that covers software updates. Those gaps are what an in-house team would want to see closed before relying on it.

The structural question for enterprise legal teams

If AI systems can be products, who carries the liability when an AI tool you bought produces a defective output: you, the vendor, or the model provider behind it? Your current vendor contracts probably answer that by assumption, not by drafting.

06 · So what

AI moved closer to where legal work arrives this week, and who reviews it is still open

Legora and Harvey want to be the place requests land, Microsoft wants review to happen inside Word, vendors are working out how to bill for use, and a UK proposal asks who is liable when it goes wrong. The tools are converging on the front door. Each one still leaves the legal team running it.

Treat "front door" as a routing question
Ask what the agent does with a request it cannot resolve, and who owns the escalation rules.
Check what your existing software already does
Microsoft's skills may overlap with a tool you are about to buy. Test the overlap before signing.
Ask for usage reporting before you commit
Under pay-per-use pricing, spend controls are a procurement requirement, not a feature.
Read your AI vendor liability clauses
A UK consultation is asking whether AI systems are products. Your contracts should say who pays if one fails.
The Flank view

Every launch this week makes the same bet: give the legal team a better tool and let the lawyer drive. But inexpensive work is still done by expensive resources whenever a lawyer is the one sorting, configuring and checking the queue. The cost of that shows up in hours now, and in per-use bills next.

Outsource legal work to supervised agents. Flank's agents know your templates, terms and escalation rules, take the NDAs, standard redlines and procurement agreements, and hand finished work to a lawyer for review before anything leaves.

Subscribe

The Intake

Weekly briefings on what's actually changing in legal AI: the market shifts, regulatory moves, and structural questions that matter for enterprise legal teams. Written by the Flank team.

Subscribe on Substack
Flank

Outsource legal work to supervised agents

Enterprise legal teams use Flank to handle high-volume contracting end-to-end: NDAs, MSA redlines, procurement, triage. Agents that know your templates, terms, and escalation rules. Lawyers review finished work.

Learn more at flank.ai