Anthropic this week introduced Claude for Small Business, a set of 31 prebuilt skills and data connectors inside the Claude Cowork experience in the Claude app, with a contract-analysis feature called /review-contract emerging as the most consequential tool for owners handling day‑to‑day agreements and financial workflows.

AI integration

Claude for Small Business is organized as structured AI applications—skills that run repeatable tasks against information pulled from connected services. Within the Claude app, users select Cowork, open Customize, add the small business plugin, and then invoke commands such as /review-contract. The contract workflow accepts an uploaded agreement and returns a full risk scan in roughly five minutes, surfacing red flags, problematic clauses, and practical edits in plain language that can be shared with counterparties during negotiations.

Anthropic positions the package as a way to extend the same AI access typically associated with large enterprises to smaller firms. In remarks to ZDNET, Lina Ochman, who leads US SMB and product-led growth GTM at Anthropic, framed the rollout around expanding AI literacy and utility for companies that have historically lacked the time and resources to implement it at scale.

Pricing ties to the Claude Pro tier, which the author notes as the minimum subscription required to run Cowork and use the new skills. Once enabled, skills operate like guided playbooks. The contract tool, for instance, was run across several real‑world documents: a home‑improvement agreement from a window company, a membership form from a national fitness chain that made cancellation difficult through shifting dates, and a caregiving services contract evaluated during a period of family stress. In each case, the analysis highlighted risk points and remedies clearly enough to support tougher negotiating positions.

Technology use case

Beyond contract review, the broader skills catalog links to widely used business systems. Connectors span QuickBooks (accounting), PayPal and Stripe/Square (payments), HubSpot (CRM and email), Mailchimp (marketing), and Canva (creative assets). Example commands include:

• /business-pulse for a cross‑functional snapshot drawing on QuickBooks, PayPal, Stripe/Square, HubSpot, and email context.
• /cash-flow-snapshot for a view of receivables, payables, payment timing, and fixed costs.
• /close-month for month‑end reconciliation and a narrative profit‑and‑loss summary.
• /invoice-chase for overdue‑invoice reminders tied to QuickBooks and PayPal data.
• /margin-analyzer for product or service margin diagnostics using payment and cost inputs.
• /plan-payroll for payroll readiness, factoring in cash position and staged reminders.
• /quarterly-review for a QBR‑style briefing on revenue, margin, customer health, opportunities, and risks.
• /run-campaign for campaign creation that blends sales analysis, content briefs, Canva assets, and HubSpot sends.

As with many out‑of‑the‑box templates, the usefulness of each skill depends on whether a business already operates on those same systems. Workflows geared to PayPal for transactions and QuickBooks for accounting, for example, will naturally get more immediate value than firms using different stacks. Still, because skills are saved as editable text descriptions, they can serve as starting points for teams that want to adapt instructions to their own tools and processes.

Data access and permissions

The rollout also underscores a recurring tension in AI adoption: how much system access to grant an assistant. The author describes a cautious approach, beginning with Mailchimp as a lower‑risk connector and setting permissions to allow reading data but not writing changes. Claude respects such scoped access at the time of connection, enabling users to constrain actions for each app.

QuickBooks access raised additional concerns. Although write permissions could be turned off within Claude, the workflow also depended on how QuickBooks itself allowed the connection to operate. When the author disabled write privileges at the QuickBooks layer, Claude would not function; re‑enabling the connection was necessary to proceed. After completing tests, both the Mailchimp and QuickBooks connectors were disconnected as a precaution, reflecting the broader industry conversation about principle‑of‑least‑privilege and the practical realities of third‑party app permissions.

Hands‑on findings

The Mailchimp connector, as tested, proved limited. It could not surface core subscriber metrics, identify lists, calculate growth, or infer style from past campaigns. It could generate a new draft mailing from a prompt, but the result required subsequent manual formatting and review.

QuickBooks integration showed more promise. A “business pulse” briefing assembled a single‑screen summary of the company’s key indicators—similar in content to standard QuickBooks reports but packaged in a conversational, narrative format. A targeted scan also surfaced a bookkeeping anomaly: an overpayment recorded 14 years earlier, where $36 in Florida sales tax had been remitted in excess of what was owed. While not mission‑critical, the finding illustrates how a generalist assistant can comb through ledgers to catch small items that might otherwise remain buried.

Market impact

For small businesses that conduct frequent transactions, manage invoices, and reconcile accounts across multiple payment channels, an AI layer that unifies context from QuickBooks, PayPal, Stripe/Square, and email can compress routine work into short, auditable runs. The inclusion of margin analysis, cash‑flow snapshots, and quarter‑in‑review summaries places emphasis on operational finance—areas where timely pattern detection and narrative explanations can support better‑grounded decisions without adding headcount.

The contract‑review capability stands out because it translates dense legal language into concrete risks and suggested counter‑terms. The analysis is both immediately actionable and easy to share with a vendor or service provider inside the same chat thread. The author, who has reviewed many agreements with attorneys, reports that the clarity and structure of Claude’s output compared favorably with far more expensive professional evaluations. At the same time, standard caveats remain in force: generative systems can err or hallucinate, so the results still require human scrutiny, and the tool should be treated as an aid rather than a substitute for legal counsel.

Industry response

Anthropic frames the release around raising the baseline for AI accessibility in the small‑business segment, which accounts for a substantial share of the US economy and private‑sector employment. In practice, the new library lands as a mixed bag of early building blocks: some skills are tightly aligned with mainstream finance and marketing tasks, while others will need customization or deeper connectors to feel indispensable.

That said, the /review-contract workflow is a clear signal of where small‑business assistants can have an outsized effect: extracting risk, clarifying obligations, and prompting informed negotiations, all within minutes and at the price of a Claude Pro subscription. For many owners, that combination—repeatable analysis, transparent outputs, and integrations across accounting and payments tools—may be enough to justify a trial, provided they are comfortable with the data‑access model and keep firm guardrails in place.

The bottom line: Claude for Small Business packages a broad set of operational skills, but it is the contract‑reviewer that most directly turns AI into a practical, back‑office control—one that small firms can apply on demand, alongside finance and marketing checkups, to reduce friction and surface issues before they become costly.