Legal AI This Week · Episode 6 · November 20, 2026
Who Let the AI Into Your Inbox?
Legal AI tools now want to search your email, calendar and chats. What permissions mean, why "read-only" is the detail to look for when AI connects to your other software, what to ask a docketing vendor about AI-suggested deadlines, and a new AI model its maker says makes things up less often.
Disclosure: this episode’s hosts, Mia Hale and Mark Dawson, are AI-generated voices. The script is written and fact-checked by the AI Tools For Law team. Nothing in this episode is sponsored.
Download the MP3 · 9:24
Stories and sources
News from Nov. 10–16, 2026. Vendor claims are attributed to the vendor; we note what each announcement did not disclose.
AI in your inbox
Top story
Harvey said Nov. 13 that its AI agents can now search Gmail and Google Drive, when you ask or on a schedule, and return answers with citations back to the source email or file, with firm administrators choosing which users get access. Harvey also said it can now search Microsoft Teams channels, chats and shared files. Legora said Nov. 16 that users can type "@Legora" in Microsoft Teams to send a request to the legal team (intake), and that its agentic search across Outlook email and calendar looks through relevant emails and meetings to find the background on a task, help draft a response and suggest next steps. Legora says existing user permissions stay in place throughout, and argues that forwarding attachments and exporting files into separate tools can create copies outside the original permission controls. Before turning on features like these, check who in the firm can see what today, choose on purpose who gets access, and read the cited email before relying on an answer.
- Harvey blog: The Brief, November 2026
- Legora announcement: Microsoft Teams and Microsoft 365
- AI Tools For Law: news, Nov. 16
- AI Tools For Law: news, Nov. 17
MCP, the universal plug
Tools
MCP (Model Context Protocol) is an open standard that lets AI tools pull information from outside software. Clarivate said Nov. 11 that its Derwent patent data and Darts-ip litigation data can now be connected to Anthropic's AI assistant through its IPOne service, joining CompuMark trademark data that was already available there; the announcement gives no pricing. Legora said Nov. 11 that it partnered with DiliTrust, a contract, entity and board-records system for in-house legal teams: shared customers can search DiliTrust records from inside Legora and draft from their own templates, and the companies say the link only reads DiliTrust data, follows DiliTrust's access permissions, and leaves any change, approval or signature in DiliTrust. It starts as an early access beta. Harvey said Nov. 13 that firms can connect their own tools and internal knowledge bases through MCP. When a tool asks to connect to your other software, ask whether it can only read, whose permissions it follows, and where the information it pulls ends up. Check any patent or trademark answer against the official record before relying on it.
- Clarivate announcement
- Legora announcement: DiliTrust
- AI Tools For Law: news, Nov. 11
- AI Tools For Law: news, Nov. 12
Deadline software with AI built in
Industry
Aderant said Nov. 16 that Milana, its cloud docketing and court-calendar tool launched in 2022, has been deployed at more than 170 firms. Aderant says its CompuLaw court rules are maintained by an in-house team of US-licensed rules attorneys, and it highlighted two AI features: one reads documents as they are filed and suggests docketing entries, and another turns a firm's own checklists and procedures into automatic deadline rules for its practice areas. These are Aderant's own figures and descriptions. If you compare docketing tools, or already use one, ask how court rules are kept current, which courts are covered, and how AI-suggested entries are reviewed before they reach your calendar.
Google's Gemini 4 Carbon and Hydrogen
The Short Docket
Google made Gemini 4 Carbon available Nov. 10 and calls it its strongest model, built for long, multi-step work; Google says it is better at working through large sets of dense documents, including for legal work. On Nov. 12 Google released Gemini 4 Hydrogen, a fast, lower-cost model for quick, everyday tasks, and says it cuts fabricated claims by more than five times compared with the older Gemini 3.8 Flash. These are Google's own figures. Fewer made-up facts is not zero, so keep checking every citation and quote before it goes into a filing.
- Google blog: Gemini 4 Carbon
- Google blog: Gemini 4 Hydrogen
- AI Tools For Law: news, Nov. 11
- AI Tools For Law: news, Nov. 13
Terms explained in this episode
- AI agent
- An AI tool that carries out a series of steps for you, like searching, reading and summarizing, instead of answering one question and stopping.
- Agentic search
- Search where an AI agent does the digging through your emails, files or meetings for you.
- Intake
- The front door for legal work, the step where someone first asks the lawyers for help.
- Permissions
- The settings that decide who in the firm can open which files, folders and mailboxes.
- MCP (Model Context Protocol)
- An open standard that lets AI tools pull information from outside software; a universal plug.
- Read-only
- The AI can look at records in another system but can't edit, approve or sign anything there.
- In-house legal team
- Lawyers who work inside a company, rather than at a law firm.
- Company entities
- The separate legal companies a business owns, along with their officers and filings.
- Docketing software
- Software that tracks court deadlines and calculates them from court rules.
- Cloud
- Software that runs on the vendor's servers over the internet, not on a machine in your office.
- Made-up claims (hallucinations)
- When an AI states something false with total confidence.
Transcript
Cold open
Mia: Quick question for every lawyer listening. Who in your firm decided the AI gets to read your email?
Mark: At some firms, maybe nobody has decided yet. The software is asking anyway.
Mia: Nice of it to ask.
Mark: We'll help you answer. Stay with us.
Mia: This is Legal AI This Week, from the AI Tools For Law team. I'm Mia Hale.
Mark: And I'm Mark Dawson.
Mia: Coming up: legal AI tools that want into your inbox, the universal plug connecting AI to your other software, deadline software with AI built in, and a fast new model that its maker says makes things up less often.
Mia: First up. Harvey, which makes an AI platform for legal work, says its AI agents can now search Gmail and Google Drive. You can ask, or set them to run on a schedule.
Mark: Quick definition. An AI agent is an AI tool that carries out a series of steps for you, like searching, reading and summarizing, instead of answering one question and stopping.
Mia: And the answers come back with sources?
Mark: Harvey says they come with citations pointing back to the source email or file.
Mia: Harvey also says firm administrators choose which users get that access. And it can now search Microsoft Teams channels, chats and shared files too.
Mia: Then Legora, another legal AI platform, announced its own Microsoft push. Someone types "at Legora" in Teams to send a request to the legal team.
Mark: That's called intake. It's the front door for legal work, the step where someone first asks the lawyers for help.
Mia: Legora also has what it calls agentic search across Outlook email and calendar. It says its AI looks through relevant emails and meetings to find the background on a task, helps draft a response, and suggests next steps.
Mark: Agentic is the agent idea again. The AI does the digging, not you.
Mia: So the AI reads your calendar and learns how often you move the same meeting.
Mark: It may learn a lot more than that. Which is the point.
Mia: Okay, the obvious question. Should a small firm be nervous about this?
Mark: Not nervous. Deliberate. Your inbox may be the most sensitive place in your practice. Client confidences live there, along with opposing counsel's emails and that one thread you'd rather nobody summarized.
Mia: Both companies do talk about access. Legora says existing user permissions stay in place throughout. Harvey says administrators choose who gets Gmail and Drive search. What does that actually mean?
Mark: Permissions are the settings that decide who in the firm can open which files, folders and mailboxes. In plain terms, the AI is supposed to follow those same lines.
Mia: Which only helps if the lines are drawn well.
Mark: Exactly. If everyone in the office can already open everything, a tool that follows permissions can too. So before you switch anything on, check who can see what today, then pick a short list of users on purpose.
Mia: And here's Legora's case for doing this at all. It says forwarding attachments and exporting files into separate tools can create copies outside your normal permission controls.
Mark: That's worth weighing. Every extra copy is one more place client information sits.
Mia: And those citations back to the email?
Mark: Click them. An answer that points to an email only helps if you read the email.
Mia: Next, three letters you're going to keep hearing. M C P.
Mark: Model Context Protocol. It's an open standard, a shared set of technical rules, that lets AI tools pull information from outside software. Think of it as a universal plug.
Mia: And it was everywhere. Clarivate, which sells patent, trademark and litigation research data, says its patent data and its litigation data can now plug into Anthropic's AI assistant, through a Clarivate service. Its trademark data was already available there.
Mark: So someone doing patent work could ask questions in plain English and get answers drawn from Clarivate's data.
Mia: Clarivate gave no pricing, so that's question one. Then Legora again. It partnered with DiliTrust, a system in-house legal teams use to manage contracts, company entities and board records.
Mark: Company entities are the separate legal companies a business owns, along with their officers and filings.
Mia: The link lets shared customers search those records from inside Legora, ask about a contract's dates or parties, and draft from their own templates.
Mark: And in-house just means lawyers who work inside a company, rather than at a law firm.
Mia: Here's the part I liked. The companies say the link only reads DiliTrust data. Any change, approval or signature still happens in DiliTrust.
Mark: That's the detail to look for. Read-only means the AI can look but can't edit, approve or sign anything in the other system.
Mia: Look, don't touch. The museum rule.
Mia: And Harvey says firms can now connect their own tools and internal knowledge bases to Harvey through MCP as well. So what's the takeaway for a small firm? Most of us aren't wiring up knowledge bases.
Mark: Sooner or later a tool will ask to connect to something you use. When it does, ask three things. Can it only read, or can it change things? Whose permissions does it follow? And where does the information it pulls end up?
Mia: And the patent answers?
Mark: Check them against the official patent or trademark record before you rely on them.
Mia: Now, deadlines. Aderant, which makes business and docketing software for law firms, says Milana, its cloud docketing tool, has been deployed at more than one hundred seventy firms.
Mark: Docketing software tracks court deadlines and calculates them from the court rules. A document gets filed, and the software works out what's due and when.
Mia: And cloud means?
Mark: It runs on the vendor's servers over the internet, not on a machine in your office.
Mia: Aderant says its court rules are kept by an in-house team of US-licensed rules attorneys. It also highlighted two AI features.
Mia: One reads documents as they're filed and suggests docketing entries. The other turns a firm's own checklists and procedures into automatic deadline rules for its practice areas.
Mia: So the AI proposes the deadline. That's a big job to hand to software.
Mark: Fair. Notice the word Aderant uses, though. Suggests. So a person should still review every entry before it goes on the calendar.
Mia: Most listeners don't have a docketing department. Why should they care?
Mark: Because a docketing tool is only as good as its rules. If you compare tools, or already use one, ask how the court rules are kept current, which courts are covered, and how AI-suggested entries get reviewed.
Mia: And if the software gets one wrong?
Mark: The deadline is still yours. No tool takes that off your plate.
Mia: Quick hits. Google released two new models in its Gemini four family. First, Carbon, which Google calls its strongest model, built for long, multi-step work. Google says it's better at working through large sets of dense documents, including legal work.
Mia: Then Hydrogen, a fast, lower-cost model for quick, everyday tasks. Google says it cuts made-up claims by more than five times compared with its older fast model, Gemini three point eight Flash.
Mark: Made-up claims are what people call hallucinations: the AI stating something false with total confidence.
Mia: More than five times fewer. So, still some.
Mark: Exactly. These are Google's own figures, and fewer made-up facts is not zero. Keep checking every citation and quote before it goes into a filing.
Mia: So, Mark. Three moves for Monday morning.
Mark: One: before any AI tool reads your firm's email, check who can see what today, and choose on purpose who gets access.
Mia: Two?
Mark: When a tool asks to connect to your other software, ask whether it can only read, and whose permissions it follows.
Mia: And three?
Mark: Ask your docketing vendor how its court rules stay current, and have a person review every AI-suggested deadline.
Mia: And check your state's ethics rules on all of it. This isn't legal advice.
Mark: That's Legal AI This Week, from the AI Tools For Law team. New episode every Friday at aitoolsforlaw dot A I. Sources are in the show notes.
Mia: Spotted an error? Email hello at aitoolsforlaw dot A I.
About the show
Not legal advice. Corrections: hello@aitoolsforlaw.ai.
Listen on Spotify · Apple Podcasts · RSS. Written version: the newsletter.