If you've been waiting for a sign that AI is no longer just for the technical side of a business, this is it.

---

What Claude Cowork Actually Is

Claude Cowork is Anthropic's agentic workspace tool — built for persistent, multi-step work sessions rather than single-question prompts. The difference matters in practice:

With a standard Claude conversation, you build context each time. You paste your document, explain your task, get output, and when you come back the next day, you start over.

With Cowork, sessions are persistent. The AI maintains context across your work — files, instructions, ongoing projects — so you return to where you left off instead of rebuilding from scratch. For a firm owner who might be reviewing a client file in the morning, drafting follow-up in the afternoon, and checking compliance tomorrow morning, that persistence is a meaningful workflow change.

The GA release — available for all paid Claude tiers (Teams at $30/user/month and Enterprise) — also added enterprise administration features: SCIM-based role controls, expanded analytics dashboards, and OpenTelemetry support for integration with monitoring systems. For a 5-person firm, these additions are primarily background infrastructure. The core product is what matters.

---

The One Addition That Directly Affects Accounting Firms

Alongside the Cowork GA release, Anthropic announced an integration with Intuit (spring 2026) that surfaces QuickBooks and TurboTax data directly inside Claude conversations.

For an accounting firm, this closes a workflow gap that has existed since AI tools became useful for financial analysis: the export-paste cycle.

Currently, if you want to analyze a client's QuickBooks data in Claude, you export a report, paste it into the conversation, and run your queries. Every session starts fresh. The Intuit integration changes this — QuickBooks financial data and TurboTax context become accessible directly in Claude without manual export steps.

The practical use case: a client advisory conversation where you can ask cross-period questions, flag anomalies, and compare categories against prior years — all without building context from exports. This is the first time a mainstream accounting data source has been available inside a conversational AI without custom integration work.

If you're on a Claude Teams plan and your firm uses QuickBooks, this is worth testing before your next client advisory meeting.

---

Claude Managed Agents: The Path for Firms That Want Background Automation

Also launching simultaneously (April 8, 2026): Claude Managed Agents in public beta.

This product is different from Cowork. Where Cowork is a workspace tool you interact with directly, Managed Agents are background AI workers — autonomous agents that execute multi-step workflows while you're not involved, at $0.08 per session hour.

The business model shift: previously, deploying an autonomous AI agent required your own server infrastructure, API management, and someone to maintain it. Managed Agents removes that infrastructure requirement. You configure the agent and review its output. Anthropic runs the infrastructure.

For a small professional services firm, the practical applications:

• Document processing pipeline: an agent that receives incoming client documents, categorizes them, extracts key information, and routes them appropriately — running continuously without manual intervention
• Contract review queue: an agent that performs first-pass review of incoming contracts, flags clause deviations from your standard template, and prepares a summary for attorney review
• Client intake automation: an agent that processes new client inquiry forms, pulls together background information, and prepares a structured intake summary before your first call

At $0.08/session hour, a simple document processing agent running 8 hours a day costs roughly $20/month in compute. The business case writes itself if the agent is handling work that would otherwise take staff time.

---

What the Non-Engineering Majority Tells Professional Services Firm Owners

Let me return to the most important data point, because it deserves more than a passing mention.

When Anthropic reports that non-engineering teams account for the majority of Claude Cowork usage in early enterprise deployments, they're describing who is actually doing agentic AI work inside organizations right now. It's not the developers. It's the operations leads, the finance teams, the in-house legal departments.

These are the people your clients employ.

If your enterprise clients are deploying Claude Cowork to their finance and legal teams, those teams are building AI workflow literacy right now. In 12 months, when you're working through their financials or advising on a contract — or competing with an internal team for budget — they will be more AI-literate than they were. Some of them may be more AI-literate than you.

This creates two different pressures depending on the kind of firm you run:

For accounting and consulting firms: Your enterprise clients' finance departments are learning to do what you do — financial analysis, data synthesis, advisory work — with AI assistance. The value you deliver has to be above what a finance team with Claude Cowork access can do on their own. That threshold is rising. Not in years. In months.

For law firms: Your clients' in-house legal teams are the heaviest non-engineering Cowork users in enterprise deployments. They're reviewing contracts, summarizing case documents, and drafting policy language with AI tools. The work they're sending to outside firms is increasingly the work that AI can't do — complex litigation, novel legal questions, jurisdiction-specific expertise. The general-practice small firm that doesn't specialize faces a version of this problem every time an in-house team decides not to refer.

---

The Practical Starting Point

If you're running a professional services firm and haven't done this yet, here's what to do this week:

1. For accounting firms: Sign up for a Claude Teams trial. Try the Intuit integration with one client's QuickBooks data. Specifically, run a cross-period analysis question you normally answer by manually comparing export reports. See what compresses.

2. For law firms: Download Claude Cowork and run one client file through a persistent session — case background, relevant documents, your working notes. Ask questions across all of it without rebuilding context. Note where the persistence changes what you can do.

3. For consulting and other professional services firms: The non-engineering majority usage figure is your signal that this tool is for people in your role. Pick one workflow that involves synthesizing information across multiple documents or conversations, and run it through Cowork for one week.

The GA release means this isn't experimental anymore. The data showing non-engineering teams leading adoption means this isn't for someone else's firm. Both of those things are true at the same time.

---

The Crossing Report covers AI adoption for professional services firm owners every Monday. Subscribe here.