Alexi

What it is

Alexi is a legal AI platform from Alexi Technologies, a Toronto company founded in 2017 and led by CEO Mark Doble. It runs legal research, drafting, and document analysis across US and Canadian jurisdictions, and it sits in the same buying conversation as Thomson Reuters CoCounsel, Lexis+ AI with Protégé, and vLex Vincent AI — research-corpus AI aimed at litigators, not a contract-review point tool.

What separates it from that group is the deployment model. Firms that take the private cloud option get a fully isolated single-tenant instance of Alexi rather than a tenant partition inside a shared system, and a dedicated VPC is available where a security committee will not accept less. Alexi documents SOC 2 certification, AES-256 encryption at rest, TLS 1.2+ and 1.3 in transit, per-firm data separation with no cross-tenant access, inference-only model usage, and zero-data-retention policies. That posture is why the name reaches a shortlist at firms whose infosec review kills multi-tenant legal AI before the demo.

Why it shows up in Legal Ops stacks

• It posted the highest accuracy score in the only independent benchmark that grades legal research. In the Vals Legal AI Report, updated 15 October 2025, Alexi scored 80% accuracy on 200 US legal research questions — the top result among the seven vendors that participated — against a lawyer baseline of 69%. Authoritativeness came in at 75%, second among participants, and appropriateness at 70%.
• Procurement clears it when it kills the alternatives. Single-tenant isolation moves the security conversation from “trust our controls” to “your instance is your own,” which is a materially shorter path through a firm’s outside-counsel guidelines and client audit obligations.
• It connects to where the work already lives. Alexi documents connectors for iManage, NetDocuments, SharePoint, Word, Excel, Outlook, Teams, and Zoom; e-signature via DocuSign and Adobe Sign; practice management via Elite 3E, Aderant, Litify, Actionstep, and Filevine; and data rooms via Datasite, SS&C Intralinks, and Firmex.

Pricing reality

Alexi publishes no price at all. The pricing path on alexi.com redirects to the homepage, the FAQ says pricing is based on firm size and deployment mode and routes to sales, and Capterra lists Alexi as contact-vendor with no free trial and no free version. The only self-serve entry point is a Guided Trial. Older third-party write-ups still quote a $499 and $949 per-month memo-and-argument plan; those tiers are not on Alexi’s site today, and quoting them back in a negotiation dates you.

Two variables set your number: headcount band and deployment mode. Single-tenant private cloud is a separate and higher line than shared deployment, so make the vendor quote both side by side — the isolation premium is the thing you are actually deciding on, and it is invisible inside a blended per-seat figure. Anchor the conversation with published category numbers: Paxton lists $499 per user per month or roughly $250 effective on an annual commitment, Legora runs about $3,000 per user per year on a 10-seat minimum for a $30K floor, and Westlaw Precision with CoCounsel lands near $428 per user per month for a solo with all-states-plus-federal coverage. An Alexi quote that lands far above the top of that band needs the isolation line broken out to justify itself.

Best for

Litigation teams at mid-market and enterprise firms whose infosec or client audit requirements rule out shared-tenant AI, and whose matters are US or Canadian. As of October 2025 Alexi reported more than 600 mid-market to enterprise firm customers, so this is not an experiment — it is a segment where Alexi has already won the procurement argument repeatedly.

Don’t buy Alexi as a transactional or contract-review engine. The benchmark result that justifies the price is a legal research score; drafting and agreement analysis exist in the product but carry no equivalent independent validation. If Word-native redlining against a negotiation playbook is the actual job, Spellbook or Ivo fit better. Skip it too if your citator workflow is already standardized on KeyCite or Shepard’s and the firm will not run a second research platform.

Versus the alternatives

Thomson Reuters CoCounsel holds the largest US share and brings KeyCite and Practical Law — pick it when citator depth and secondary sources are the requirement you cannot trade away. Lexis+ AI with Protégé is the other incumbent — pick it when Shepard’s is the firm’s citator of record and the Lexis contract renews regardless. Legora is the fastest-growing entrant and is built around collaborative drafting — pick it when several lawyers need to work one document with AI in the room. vLex Vincent AI is the multi-jurisdiction option and is now owned by Clio, the party on the other side of Alexi’s litigation — worth naming out loud if both are on your shortlist. The status quo alternative is Westlaw or Lexis plus a general assistant; choose Alexi over that when single-tenant isolation is a hard gate rather than a preference.

Watch-outs

• Alexi’s US caselaw supply is in active litigation and it lost the emergency round. Fastcase, now part of Clio, sued Alexi on 26 November 2025 in the US District Court for the District of Columbia (case 1:25-cv-04159) over its use of licensed caselaw data, terminated the licence effective 27 November, and stopped daily caselaw updates on 3 December 2025. Judge Richard J. Leon denied Alexi’s emergency motion to restore access on 18 December 2025, finding the claimed harm too speculative to merit emergency relief. Cross-motions for partial summary judgment were argued on 8 July 2026, and no ruling had issued as of 23 August 2026. Guard: before signing, put a citation-currency commitment in the contract — name the jurisdictions, name the update frequency, attach a termination right if it slips — then spot-check ten recent opinions from your own practice area and jurisdiction inside the Guided Trial.
• Alexi has not publicly named a replacement primary-law source. Its own counterclaim describes the vLex and Fastcase corpus as one of only three full-coverage US caselaw databases, which is an argument about how hard that supply is to replace. Guard: ask the vendor in writing where US and Canadian primary law comes from today and who else it is licensed from; treat a non-answer as reason to keep an incumbent research seat running in parallel through the first contract year.
• The 77% aggregate score is less flattering than the 80% accuracy headline. Alexi’s weighted aggregate in the Vals report was 77% against a cohort average of 76% — it won on accuracy, not on the composite. Guard: if your evaluation weighs appropriateness or output shape as heavily as correctness, run your own scored bake-off on live matters instead of buying on the accuracy line.
• There is no public API and no MCP server. Alexi’s integrations are inbound connectors into Alexi; nothing documented lets your own agents call Alexi research the way Harvey now allows. Guard: if agent-accessible research is on your roadmap, keep a separate programmable path for it and treat Alexi as an end-user surface rather than infrastructure.
• Pricing opacity is total, which makes the first quote an anchor rather than a market price. Guard: get the per-seat number, the isolation premium, and the renewal-year uplift in writing during procurement, and benchmark all three against a Paxton or Legora quote for the same seat count.