AI-assisted and agentic legal research, citation grounding, case law analysis, and memo drafting

Perplexity just launched a product built only for lawyers.
It plugs straight into the tools they already use.

It's called Computer for Counsel.
It wires research databases, document repositories, contract tools, and matter-management systems into Perplexity Computer.

Here's what Computer for Counsel does:

It clears the admin work off attorneys' desks
→ It handles research, document gathering, and contract triage in the background.
→ Perplexity points to a Thomson Reuters survey where nearly 75% of lawyers call admin a major time drain.
→ The trade is more hours on judgment, fewer on the work nobody went to law school for.

It picks the right model for each task
→ More than 20 frontier models sit under the hood.
→ Computer chooses the best one for research, reasoning, or contract work.
→ That's the same multi-model setup behind Perplexity Computer.

It links every answer back to a source
→ Outputs cite the case, statute, regulation, filing, or internal document they came from.
→ Attorneys verify accuracy before anything lands in a brief or reaches a client.
→ For a field that watched peers get sanctioned over fake citations, that's the whole sell.

It connects to the legal stack
→ MCP connectors include Box, Carta, Docusign, DeepJudge, and NetDocuments.
→ Inside Microsoft 365 it drafts in Word, pulls files from SharePoint, and reads context from Outlook or Teams.
→ Premium sources include Midpage for case law, with LegalZoom and Deel on limited access.

It assembles the deliverable for you
→ Computer pulls the relevant pieces into a brief, a memo, or a deal summary.
→ Live workflows cover third-party NDA intake, regulatory monitoring, and case research with citation review.
→ Clio and Ironclad are on the way, not live yet.

It's open now to Perplexity Enterprise and Max subscribers.

Perplexity isn't trying to out-Westlaw Westlaw.
It's going after everything that happens before and after the research, and betting that lawyers will trust it because every answer shows its receipts.

Subscribe to The AI Report for daily AI news → https://lnkd.in/eVxtGmfV

Building on Libra by Wolters Kluwer’s early leadership, strong market momentum and advanced AI innovation in Italy, its capabilities are now directly available in One via the Libra add-in.

Milan, July 8 2026 – Wolters Kluwer Legal & Regulatory today announced the integration of Libra by Wolters Kluwer AI workflows in One, its leading legal research platform in Italy. This step further reinforces the strong commercial success and rapid market adoption of the Libra legal AI workspace in the country. With the integration generative AI features are directly available in the research workflow of legal professionals allowing a seamless experience without switching between tools.

I'm a solo trial attorney in Maine. Federal and state criminal defense, family law, state and federal appeals, and CJA panel work. I've been using Claude (Cowork mode almost every time) as a drafting and research assistant, and figured I'd share the setup in case it's useful to other litigators.

I have a dedicated project with standing instructions describing my practice areas, jurisdictions (Maine state court, Maine federal court, First Circuit Court of Appeals), and drafting conventions — formal prose matching Maine and First Circuit filing norms, numbered paragraphs, lead with the strongest argument, and citation priority to Maine SJC, Maine statutes, First Circuit, and Maine federal authority. Every draft it produces starts from that baseline instead of generic legal-writing defaults.

I have a custom-built plugin I put together myself for drafting trial-court motions and supporting memoranda (Maine state and federal court). It researches authority via CourtListener and Google Scholar and outputs a court-formatted .docx.

A standalone skill for drafting appellate brief argument sections (Maine Law Court and First Circuit).

Also a general "legal" plugin pulled from Anthropic's plugin library — NDA triage, contract review against a playbook, compliance checks, vendor agreement tracking, meeting briefings, templated responses to legal inquiries. Honestly this one's built for in-house/corporate legal teams, not litigation, so it's more "came with the toolkit" than something I use day to day.

CourtListener and Google Scholar are wired in for case law and citation research, feeding the motion-drafting skill directly.

The thing I've found most useful: Claude has a persistent memory built from reviewing my actual past filings — motions to suppress, Law Court and First Circuit briefs, motions in limine, sentencing memos. It captures my structural conventions per document type, citation format (Maine's ME-number format, Bluebook signals, First Circuit record-citation conventions), and specific style corrections to avoid. That carries across sessions, so new drafts already sound like mine instead of generic legal-AI output.

While I have used various providers and models over the years, I am only about a level above a noob. I have found that Sonnet can draft a decent federal sentencing memorandum (more difficult than it sounds), a motion to suppress in criminal cases and can review appellate briefs very well (have not drafted an appeal yet, but next on my list when I take another appeal on).

What I am looking for is where should I go next? Areas for improvement? New areas I have not thought of? Missing something obvious to others? Looking for human lawyers in appellate law, family law or criminal law who can give input based on their experience.

Happy to answer any questions if that helps. Thanks!

Descrybe Legal Engine is now available through ChatGPT’s App Directory, bringing structured U.S. primary law, citation intelligence, treatment analysis, and source-grounded AI legal research into one of the world’s most widely used AI assistants.

June marks a major leap forward headlined by early access to the next generation of CoCounsel Legal in the U.S. as well as major releases that push agentic capabilities into more jurisdictions and more of the everyday moments where lawyers are already working. From cross-border research and citation-level verification to a clause library reaching new markets, this month’s innovations reflect our ongoing commitment to Agentic AI grounded in deep legal expertise, and tools built for how modern legal teams actually work.

Starting this week, CoCounsel Legal customers will get early access to the next generation of CoCounsel, ahead of the full launch planned for later this summer starting in the U.S, and followed by rollouts in Canada, UK and Australia. The product will keep getting better. But it is already good enough that keeping it from you any longer felt like the wrong call.

The next generation works the way an attorney works with a teammate. Describe your matter in plain language. CoCounsel Legal creates a plan, dives deep into legal authority, reasons through the legal issues, retrieves what it needs from your own precedents, and from Westlaw and Practical Law, and drafts with citations. When new information changes the picture, it adapts. Instead of getting piecemeal answers, you get a professional standard response that you can iterate on.

Legal research has always been one of the most demanding parts of legal work. Deep Research in the next generation of CoCounsel Legal changes the equation.

Highlights

Deep Research in CoCounsel Legal transforms legal research from manual synthesis to autonomous execution.

The system accesses Westlaw content curated by 1,500+ attorney-editors and validated by AI specialists.

CoCounsel Legal delivers end-to-end legal work from research to drafting in a single workflow.

Less than three months after replacing Lexis+ AI with Lexis+ with Protégé as its flagship legal AI platform, LexisNexis today is announcing a substantial expansion of that platform – what the company is calling its “next evolution,” combining significant build-outs of existing capabilities with several completely new ones.

Courts across the country are adopting rules that require lawyers to disclose when they "use" AI in their filings. The intent behind these rules is understandable. But there is a word doing a lot of heavy lifting in that sentence, and it deserves more scrutiny: "use."

Consider a few scenarios.

I pull up a legal-research platform that happens to be powered by AI. I feed it my adversary's brief and ask it to identify every case cited so I can pull and review each one myself. Did I "use" AI? The tool saved me ten minutes of skimming footnotes, but every word of analysis is mine.

Now suppose I go a step further. I ask the platform to generate short summaries of each cited case so I can triage which ones warrant a close read. I still read the key cases in full. I still do the analysis. But AI helped me decide where to focus. Must I file a disclosure?

Or take a different scenario entirely. I draft a brief from scratch — every argument, every citation, every piece of analysis is my own work. But the brief is two pages over the limit. I run it through an AI writing tool to tighten the prose, cut redundancies, and bring it within the page count. The substance is untouched. Is a disclosure required?

These are not hypotheticals I am inventing to be difficult. These are questions that lawyers are confronting right now, and the current patchwork of court rules does not answer them clearly. The word "use" is doing too much work without enough definition. A lawyer who asks AI to draft an entire brief from whole cloth is "using" AI. A lawyer who asks AI to fix a typo is also "using" AI. Surely those two things are not the same, and surely they should not trigger the same disclosure obligations.

We are all still learning — learning which tools work, which tasks are well-suited to AI, and where the guardrails need to be. That learning is happening in real time, and the rules are struggling to keep pace.

I'd love to hear from this community.

Clients: How do you expect your outside counsel to use AI? Which uses do you prohibit? Discourage?

Practitioners: How are you using AI in your practice? Which tasks have you found it's best suited for? Where do you draw the line on disclosure? And do you think the current rules are getting it right — or do they need to catch up to how lawyers are actually working?

Drop your thoughts in the comments. 👇

Okay, so I use LexisAI for case research. ChatGPT plus for rewording emails, drafting demand letters, preparing lawsuits, etc.

I’m trying to start using Gemini because my friends say it’s helpful.

What are you favorite AI platforms and how do you use it for work?