Published
- 10 min read
Claude Cowork and the End of Enterprise Software Patience
The Short Version
Anthropic launched Claude Cowork on 12 January 2026 and followed up on 30 January with 11 open source plugins targeting specific business functions. The legal plugin triggered what Bloomberg described as “a $285 billion rout in stocks across the software, financial services and asset management sectors,” with Thomson Reuters falling 18%, RELX dropping 14%, and software stocks experiencing their worst day since April 2025.
The significance is not the capability itself but the strategic positioning. As LawNext observed, “For the first time, a foundation-model company is packaging a legal workflow product directly into its platform, rather than merely supplying an API to legal-tech vendors.” Anthropic is no longer just plumbing; it is now competing with its own customers.
One might reasonably ask why. The uncharitable answer is that Anthropic’s enterprise customers have proven themselves rather spectacularly incapable of shipping anything useful with the tools they have been given. Microsoft, to pick an example entirely at random, has spent the better part of two years promising Copilot would transform enterprise productivity. The result has been a series of underwhelming demos and a product that, by most accounts, struggles to reliably summarise a meeting. Anthropic, meanwhile, appears to have vibe coded a functional contract review system in a fortnight. When your API customers cannot manage to build what you can knock together over a long weekend, eventually you stop waiting for them.
What Is Cowork?
Cowork brings Claude Code’s agentic architecture to non developers. According to Anthropic’s support documentation, “Instead of responding to prompts one at a time, Claude can take on complex, multi-step tasks and execute them on your behalf.”
Key characteristics:
- Runs on macOS desktop (research preview)
- Direct local file access without manual uploads
- Sub agent coordination for parallel workstreams
- Creates polished outputs including Excel spreadsheets with formulas, PowerPoint decks, and formatted documents
- Long running tasks without context limits or timeouts
- Virtual machine execution environment for security
Current limitations (per Anthropic’s documentation):
- macOS only
- No memory across sessions
- No project support
- No GSuite connector support
- Plugins saved locally (organisation wide sharing coming)
- Not recommended for regulated workloads
Availability: Pro, Max, Team, and Enterprise plans. Plugins are open source on GitHub.
Plugin Architecture
According to Anthropic’s plugin announcement, each plugin bundles four components:
| Component | Function |
|---|---|
| Skills | Domain knowledge Claude draws on automatically when context requires |
| Slash Commands | Explicit actions triggered with /command syntax |
| MCP Connectors | Integrations with external tools via Model Context Protocol servers |
| Sub agents | Specialised agents for complex subtasks |
The GitHub repository notes that “Every component is file-based — markdown and JSON, no code, no infrastructure, no build steps.”
Plugin structure:
plugin-name/
├── .claude-plugin/plugin.json # Manifest
├── .mcp.json # Tool connections
├── commands/ # Slash commands
└── skills/ # Domain knowledgeThe 11 Official Plugins
All plugins are open source on GitHub.
1. Productivity
Manages tasks, calendars, and daily workflows.
Connectors: Slack, Notion, Asana, Linear, Jira, Monday, ClickUp, Microsoft 365
Commands:
/updatescans emails, calendars, and chats to update task lists/competitive-briefcreates automatic competitive briefs/metrics-reviewanalyses performance metrics
2. Sales
Manages the entire sales cycle from prospecting to close.
Connectors: Slack, HubSpot, Close, Clay, ZoomInfo, Notion, Jira, Fireflies, Microsoft 365
Capabilities:
- Automated prospect research
- Sales call preparation
- Personalised follow ups
- Competitive analysis
- Sales forecasting
- Pipeline summary recording
Example prompt: “Prepare a brief for the Acme Corp call tomorrow: include the last 5 CRM interactions, recent company news, competitive advantages against main competitors, and closing suggestions.”
3. Legal
The plugin that triggered the market panic. According to the plugin page, it “automates contract review, NDA triage, compliance workflows, legal briefings, and templated responses — all configurable to your organization’s playbook and risk tolerances. Built for commercial counsel, product counsel, privacy/compliance, and litigation support teams.”
Connectors: Slack, Box, Egnyte, Jira, Microsoft 365
Commands:
/review-contractruns clause by clause review with GREEN/YELLOW/RED flags and specific redline suggestions based on configured playbook/triage-ndaprovides rapid NDA pre screening (standard approval, counsel review, full review)/vendor-checkchecks vendor agreement status/briefgenerates contextual briefings (daily brief, topic research, incident response)/respondcreates templated responses for common requests (GDPR, discovery holds)
How it works: The system analyses the entire contract before flagging issues because clauses interact with each other. An uncapped indemnity may be partially mitigated by a broad limitation of liability elsewhere in the document.
Explicit disclaimer: Anthropic frames this as assistance, not advice. The Daily Upside reports that the company states “All outputs should be reviewed by licensed attorneys.”
4. Finance
Supports accounting, reconciliations, and financial analysis.
Connectors: Snowflake, Databricks, BigQuery, Slack, Microsoft 365
Capabilities:
- Journal entry preparation
- Variance analysis
- Financial statements
- Local tax regulation research
- Chart creation (waterfall, trend analysis)
- Forecasting models
- Account reconciliation
Example prompt: “Analyse cash flow for the last 3 quarters from the attached CSV and create a waterfall chart highlighting major variations between periods.”
5. Data
Query, visualise, and interpret datasets.
Connectors: Snowflake, Databricks, BigQuery, Hex, Amplitude, Jira
Capabilities:
- Natural language queries on business data
- SQL query generation
- Automatic visualisations
- Data validation and cleaning
- Statistical analysis
- Automatic PowerPoint deck creation
- Export in multiple formats
Commands:
/write-querygenerates SQL from natural language
6. Marketing
Manages content creation and campaign execution.
Connectors: Slack, Canva, Figma, HubSpot, Amplitude, Notion, Ahrefs, SimilarWeb, Klaviyo
Capabilities:
- Brief to content draft workflow
- Brand consistency review
- Automatic SEO optimisation
- Editorial calendars
- Campaign performance reports
- Social and email campaign planning
7. Customer Support
Automates ticket routing and response drafting.
Connectors: Slack, Intercom, HubSpot, Guru, Jira, Notion, Microsoft 365
Automatic priority routing:
| Priority | Action |
|---|---|
| Critical | Immediate escalation to engineering plus management notification |
| High | Assignment to senior support with 4 hour SLA |
| Medium | Knowledge base search plus response draft |
| Low | Automatic template response plus documentation |
Additional capabilities:
- Draft responses
- Package escalations
- Research customer context
- Convert resolved issues into knowledge base articles
8. Product Management
Creates product documentation and manages roadmaps.
Connectors: Slack, Linear, Asana, Monday, ClickUp, Jira, Notion, Figma, Amplitude, Pendo, Intercom, Fireflies
Capabilities:
- PRD (Product Requirement Documents)
- Technical specifications
- Visual roadmaps
- Feature prioritisation (RICE, MoSCoW)
Commands:
/write-specgenerates product specifications
9. Enterprise Search
Simultaneous hybrid search across company tools.
Connectors: Slack, Notion, Guru, Jira, Asana, Microsoft 365
Searches across:
- Chat (Slack, Teams)
- Documents (Google Drive, SharePoint)
- Internal wikis (Confluence, Notion)
Features:
- Automatic deduplication
- Synthesis with citations
- Document export
- Schedulable for recurring questions
Example: “What are customers complaining about this week?” Claude simultaneously searches support tickets, sales emails, Slack chats, and synthesises main themes with citations to original sources.
10. Biology Research
Accelerates R&D in life sciences.
Connectors: PubMed, BioRender, bioRxiv, ClinicalTrials.gov, ChEMBL, Synapse, Wiley, Owkin, Open Targets, Benchling
Capabilities:
- Queries only peer reviewed sources (PubMed, Nature, Science)
- Configurable for specific domains (oncology, genomics, pharmacology)
- Scientific literature synthesis with citations
- Methodology comparison between studies
11. Plugin Manager (Meta Plugin)
Creates and customises other plugins.
Process:
- Define the workflow in natural language
- Plugin Manager identifies necessary capabilities
- Generates related skills and slash commands
- Creates the complete plugin structure
- Distributes to team
Installation and Usage
From marketplace:
- Open Claude Cowork on desktop
- Click Plugin in left sidebar
- Click + then Browse plugins
- Select plugin
- Install
Activating commands: Type / in chat, go to Plugin, and select from available commands.
Custom plugins: Click + then Upload and provide the plugin ZIP file.
Why the Market Panicked
The selloff was not about the specific capabilities. Contract review and NDA triage tools have existed for years through vendors like Harvey (valued at $8 billion), Legora ($1.8 billion), Thomson Reuters, and RELX. The legal tech industry has been promising AI transformation since approximately the Jurassic period, with results that could charitably be described as “incremental.”
The shift is strategic, and the market understood it immediately:
-
Foundation models are now competitors. As LawNext noted, “Many legal AI vendors have built their products on the ‘model + wrapper + workflow’ model, assuming that the model layer remains a neutral player. But now Anthropic is effectively bundling its own ‘model + wrapper + workflow’ – circumventing the legal vendor and going straight to the customer.”
-
Distribution advantage. The legal plugin ships alongside sales, marketing, finance, and other enterprise functions. For in house legal departments dealing with NDAs from sales contracts, this creates an effective and streamlined distribution channel.
-
Pricing pressure. Why pay for expensive enterprise legal software when the foundation model provider includes it in the subscription? Thomson Reuters has built a rather comfortable business charging eye watering sums for Westlaw access. That business model assumes customers have no alternative. They now have one.
-
Open source as standard setting. Anthropic is following the MCP playbook: open source the infrastructure, let it become industry standard, position as ecosystem foundation. In December 2025, Anthropic donated MCP to the Linux Foundation as part of the newly formed Agentic AI Foundation. VentureBeat reported that “OpenAI has quietly adopted structurally identical architecture in both ChatGPT and its Codex CLI tool.”
Market commentary:
Mike Archibald, portfolio manager at AGF Investments, told Reuters: “I think Anthropic came out with some plug-ins to address the legal space. Obviously, that’s where Thomson Reuters generates a good chunk of their revenues. Sometimes the market just shoots first and asks questions later.”
Morgan Stanley analysts led by Toni Kaplan wrote: “Most of the investors we have spoken with recently are overwhelmingly bearish on TRI as the consensus opinion worries that the company will be unable to maintain the same level of growth within its legal segment given increased competition from specialized AI tools.”
Schroders analyst Jonathan McMullan observed: “The selling pressure in software and data analytics reflects a deepening structural debate, accelerated today by Anthropic’s legal automation tool challenging incumbents like RELX. Investors are aggressively repricing these areas as the historical ‘visibility premium’ erodes.”
Jefferies described software sentiment as the “worst ever,” while Anurag Rana of Bloomberg Intelligence called it “radioactive.”
Market impact (3–4 February 2026):
| Company | Drop | Notes |
|---|---|---|
| Thomson Reuters | 18% | Record single day loss |
| RELX | 14% | Biggest drop since 1988 |
| Wolters Kluwer | 13% | |
| LegalZoom | 19.7% | |
| FactSet Research | 10.5% | |
| Morningstar | 9% | |
| Goldman Sachs Software Basket | 6% | Biggest one day decline since April 2025 |
The selloff spread beyond legal to data analytics, financial services, and broader software. Business development companies like Blue Owl Capital fell 13% for nine consecutive sessions.
What It Means
For legal tech vendors:
The existential question is whether value lies in proprietary data (case law, legal research databases) or in the workflow software layer. If primarily the software layer, they are now competing against tools users can customise themselves without paying enterprise licensing fees to companies whose primary innovation in the past decade has been finding new ways to increase said fees.
Artificial Lawyer noted: “High quality legal tech tools have little to fear here…for now. Although, perhaps better customisation of what’s on offer could go some way to addressing the limitations of what’s on offer? As noted before by AL, the real risk is for those vendors selling commoditised legal AI skills. Those indeed face something of an existential threat.”
The phrase “for now” is doing considerable heavy lifting in that sentence.
For enterprises:
Cowork plugins require some technical skill to deploy, but most large firms and corporate legal departments can handle implementation with existing tech teams. The plugins are customisable to organisation specific playbooks and risk tolerances. More to the point, they actually work, which puts them ahead of approximately 90% of enterprise software.
For the broader software industry:
This is the first clear example of the market repricing based on AI disruption rather than AI enablement. Software companies are no longer automatic beneficiaries of AI advancement; they may be direct casualties. The comfortable assumption that AI would simply make existing software more valuable has collided rather violently with reality.
Axios summarised the broader concern: “The software selloff dragged down the entire market on Tuesday—it’s the first example of how the market will respond when presented with evidence that AI could disrupt or even replace an entire industry.”
The enterprise software industry has spent years selling the promise of AI transformation whilst delivering chatbots that cannot remember what you asked them thirty seconds ago. Anthropic, now projecting $9 billion in annual revenue and with Claude Code alone generating $1 billion in run rate revenue within six months of launch, has demonstrated what happens when someone who actually understands the technology decides to stop waiting for the middlemen to figure it out.
Have a comment?
Send me a message, and I'll get back to you.