Three legal AI vendors converge on the same lawyer's daily workflow in 2026 — Anthropic's Cowork legal plugin (open source, free, released February 2026), Microsoft Copilot for legal (new lawyer-targeted capabilities April 15, 2026), and Spellbook (post-$50M Series B March 2026 at $350M valuation). Each occupies a structurally different position. Cowork ships the workflow logic openly, with the cost being the underlying Claude subscription. Copilot integrates AI into the Microsoft 365 stack 90%+ of firms already run. Spellbook owns proprietary contract review depth plus the Canadian Bar Association exclusive. This comparison is operational fit, not vendor character — neutral on each, opinionated on which fits which firm size and practice mix.

The three positioning frames — what each vendor actually does

The vendors look like substitutes from the outside. They aren't. They occupy structurally different positions:

- Anthropic Cowork legal plugin. Open-source plugin that runs on Claude Cowork. Two commands: /review-contract (clause-by-clause review against configured negotiation playbook with GREEN/YELLOW/RED flags) and /triage-nda (NDA categorization for standard approval / counsel review / full review). Configuration per-org playbook + risk tolerance. Cost: the underlying Claude subscription. Per Anthropic's pricing, Pro at $17-20/user/month, Team Standard at $20-25/seat/month, Team Premium at $100-125/seat/month, Enterprise at $20/seat + usage rates. Plugin itself is free. - Microsoft Copilot for legal. AI assistant integrated into Word, Outlook, Teams, Excel, and OneNote. Per Microsoft's pricing, Copilot for M365 Enterprise add-on at $30/user/month annual on top of E3 ($36/user/month) or E5 ($57/user/month). Capabilities specifically for legal include contract comparison in Word, audit trail with track changes, multi-document summarization, response suggestions in Outlook with key links and updates. - Spellbook. Vertical legal AI vendor focused on contract drafting and review, integrated into Word as an add-in. Per Spellbook's pricing page, all tiers are quote-only with 7-day free trial. Per Artificial Lawyer and aiapps coverage, industry estimates suggest $180-300/seat/month with $199/seat/month enterprise minimum 10 seats — these are NOT vendor-confirmed and should not be quoted as facts.

The second-order read: each vendor charges for a different thing. Cowork charges for model access (the workflow ships free). Copilot charges for AI integration into the productivity suite. Spellbook charges for vertical legal expertise embedded in the workflow. Comparing per-seat prices alone misses what each is actually selling.

The third-order read: the buying decision isn't "which vendor is best." It's "which combination fits the firm's existing infrastructure and practice mix." Most firms running this comparison correctly end up choosing 2 of the 3, not 1.

Cowork's structural strength: open-source workflow plus foundation model access at consumer-tier prices. The workflow logic ships free; the cost is the Claude subscription the firm may already have for general AI use.

Where Cowork wins:

- In-house legal departments. Per the in-house deployment checklist, 50-person in-house teams land at $12,000-15,000/year on Claude Team Standard plus the free plugin — replacing legal-ops tooling that historically priced at $50,000+ annual minimum. - Mid-market firms with significant transactional volume. The /review-contract command configured against the firm's playbook handles first-pass clause review faster than manual review. - Firms running the Anthropic stack already. Adding the plugin to existing Claude deployment is configuration-only — no procurement, no security review.

Where Cowork doesn't win:

- Firms needing Westlaw or Lexis content embedded in workflow. Cowork doesn't have proprietary research content. Firms with significant litigation practice need a content-bearing tool layered on top. - Firms with no Claude deployment. Standing up Claude plus configuring the plugin plus training lawyers takes 60-90 days. Faster paths exist if the firm wants AI tooling shipping in 30 days. - Firms with very specific contract templates and negotiation playbooks across many practice groups. The configuration burden scales with practice complexity. Spellbook's pre-built clause library may save configuration time at the cost of customization depth.

The second-order read: Cowork is the structurally cheapest option for firms already running the Anthropic stack. For firms not on Anthropic, the hidden cost is the deployment infrastructure required to make the plugin useful. Per the procurement checklist for mid-market firms, the deployment infrastructure costs are real.

Copilot's structural strength: AI integration into the productivity stack 90%+ of law firms already run. Lawyers don't change tools to use AI — AI appears inside Word, Outlook, Teams.

Where Copilot wins:

- Firms with deep M365 deployment. Copilot inherits the existing license, security review, audit posture, and identity infrastructure. Per the Microsoft Foundry deployment analysis, procurement cycle compresses from 60-180 days to 30-60 days. - Cross-functional workflows. Document drafting in Word + email response in Outlook + meeting summary in Teams + spreadsheet analysis in Excel — Copilot handles all of them. No vendor needs to integrate separately. - Firms with significant practice in regulated industries. Microsoft's compliance posture (SOC 2, ISO 27001, HIPAA, FedRAMP, GDPR) is broader than vertical legal AI vendors. - Bing AI Performance integration. Free per Bing Webmaster Tools, shows which queries surface the firm's content via Microsoft Copilot citations. Vertex's first-party data shows 2,100+ Copilot citations of aivortex.io in 30 days with top grounding query "Harvey AI legal."

Where Copilot doesn't win:

- Firms wanting deep contract review specialization. Copilot's contract comparison in Word is general-purpose; Spellbook's clause library is legal-specialized. For high-volume contract review, Spellbook's depth wins. - Firms running Google Workspace primary. Copilot doesn't help. Vertex AI is the right path. - Firms wanting to deploy at consumer-tier prices. Copilot is $30/user/month minimum on top of E3 or E5 — expensive compared to Cowork's $17-25/user/month total deployment.

The second-order read: Copilot is the structurally cheapest option for firms already on M365 because it's pure-add to existing infrastructure. The procurement battle is shorter and the deployment infrastructure already exists.

The third-order read: per Bing AI Performance citations data, Copilot is where lawyers actually search. Firms not deployed at the integration layer miss not just AI generation — they miss AI search visibility on their own content.

Spellbook — strengths, weaknesses, fit

Spellbook's structural strength: deep contract review specialization plus the Canadian Bar Association exclusive partnership covering ~40,000 Canadian lawyers, judges, notaries, and law students.

Where Spellbook wins:

- Firms with significant Canadian practice. Per Spellbook's CBA partnership analysis, the 2-year exclusive provides effective monopoly on Canadian small/mid-market AI contract drafting and review tooling. - Mid-market firms with 50%+ transactional practice. Spellbook's pre-built clause library plus precedent learning saves configuration time the Cowork plugin would require. - Firms wanting Word add-in deployment without full M365 Copilot deployment. Spellbook installs as a Word add-in independently of Microsoft Copilot. - Firms wanting vertical-specialist support. Spellbook's customer success team specializes in contract review workflows; general-purpose AI vendors (Cowork, Copilot) don't have the same vertical depth.

Where Spellbook doesn't win:

- Firms outside contract review work. Spellbook is contract-focused. Litigation, regulatory, and compliance practices need different tooling. - Firms wanting the cheapest possible deployment. Per industry estimates ($180-300/seat/month, NOT vendor-confirmed), Spellbook is structurally more expensive than Cowork. The premium pays for vertical specialization. - Firms not running Microsoft Word as primary drafting tool. Spellbook's primary integration is Word. Firms on alternative drafting tools have less integration depth.

The second-order read: Spellbook's CBA exclusive creates structural pressure for non-Spellbook firms in Canada. The 2-year exclusivity window is the period when Spellbook builds depth that's hard to replicate.

The third-order read: per the Spellbook Series B analysis, the $50M funding plus track to $100M ARR signals Spellbook is investing in distribution depth specifically. The competitive battle is whether Spellbook's vertical depth holds up against Cowork's open-source pressure plus Copilot's distribution scale.

Side-by-side comparison — by firm size and practice mix

Solo and small firms (under 50 lawyers):

- Best fit: Cowork legal plugin on Claude Pro at $17-20/user/month per Anthropic's pricing. Plugin is free; total annual ~$240 per user. Replaces basic-tier vendor tooling for transactional work. - Secondary: Microsoft Copilot if firm runs M365 already and values AI integration into Word/Outlook. $30/user/month on top of M365 base. - Spellbook fit: marginal at this size. Industry-estimated $180-300/seat/month is hard to justify for sub-50-lawyer firms unless contract review is the primary practice.

Mid-market firms (50-500 lawyers):

- Best fit: combination of Cowork + Copilot. Cowork plugin handles configurable contract workflows on Claude Team. Copilot handles in-suite integration across Word, Outlook, Teams, Excel. Annual run-rate ~$140,000-208,000 for 200-lawyer firm per the procurement checklist. - Add Spellbook if 50%+ transactional practice or significant Canadian work. Spellbook's vertical depth plus CBA exclusive justifies the premium for these specific use cases.

BigLaw and AmLaw 100:

- Best fit: combination of all three plus internal tooling. Cowork for configurable transactional workflows, Copilot for in-suite integration, Spellbook for high-volume contract review specialization, plus Foundation model API for internal-tool builds. - Plus Foundry, Bedrock, or Vertex deployment depending on existing cloud relationship per the Microsoft Foundry deployment analysis. - Plus enterprise-tier pricing negotiation with each vendor — at this scale, list prices are starting points.

By practice area:

- Transactional M&A: Cowork plus Spellbook (clause depth) plus internal tooling (custom diligence workflows). - Litigation: Copilot for in-suite integration plus Thomson Reuters CoCounsel post-rebuild for Westlaw-grounded research (per industry estimates $75-500/user/month per Costbench March 2026, not vendor-confirmed). - Regulatory and compliance: Copilot for cross-functional workflows plus internal tooling for firm-specific compliance frameworks. - Contract review at volume: Spellbook's vertical depth justifies the premium where contract throughput is the practice's primary work product.

The second-order read: most firms running this comparison correctly end up with 2 of the 3 vendors. Picking 1 only fits very specialized firms. Picking all 3 fits BigLaw scale where the per-seat premium is absorbed by the value of comprehensive tool coverage.

The Bottom Line: My take: Cowork legal plugin, Microsoft Copilot, and Spellbook occupy structurally different positions in the legal AI stack. Cowork ships the workflow logic openly with cost being the Claude subscription. Copilot integrates into the Microsoft 365 stack 90%+ of firms already run. Spellbook owns vertical contract review depth plus the Canadian Bar exclusive. The right choice for most firms isn't one — it's a combination. Solo and small firms: Cowork on Claude Pro is the floor; Copilot if already on M365. Mid-market: Cowork plus Copilot is the standard combination; add Spellbook for transactional-heavy practice or Canadian work. BigLaw: combination of all three plus internal tooling on top of Foundation model API. The vendor war isn't "who wins" — it's "which combination fits which firm size and practice mix."

AI-Assisted Research. This piece was researched and written with AI assistance, reviewed and edited by Manu Ayala. For deeper takes and the perspective behind the research, follow me on LinkedIn or email me directly.