Straight answers
The questions owners ask before they try anything: is this allowed, is it safe, will it hold up. Every claim on these pages is sourced, and where the answer is no, it says no.
Is Claude HIPAA compliant?
No, Claude is not HIPAA compliant by default. Anthropic offers a Business Associate Agreement that covers only its HIPAA-ready services, which means Claude Enterprise after the organization's Primary Owner activates HIPAA compliance and accepts the BAA, and the first-party Claude API once Anthropic enables it on a HIPAA-ready organization. The Free, Pro, Max, and Team plans have no BAA path, so protected health information should not go into them. Even with a BAA in place, only specific listed features are covered, and the practice remains responsible for how its staff actually use the tool.
Can I put client data into Claude?
Often yes, but the plan you are on changes the answer. Anthropic states that by default it does not use inputs or outputs from its commercial products, which include Claude for Work Team and Enterprise plans and the API, to train its models, while on the consumer Free, Pro, and Max plans model training is governed by a setting you choose in Privacy Settings and can change at any time. In both cases, conversations flagged by Anthropic's safety systems can still be reviewed and used to improve its trust and safety models. Before client material goes in, remove identifying details where you can and check what your own client agreement and professional obligations require.
Can lawyers use Claude without getting sanctioned?
Yes, as long as you never treat Claude as a source of legal authority. A language model produces text that looks like a citation whether or not the case exists, which is how the lawyers in Mata v. Avianca drew a $5,000 Rule 11 sanction in 2023 for a brief built on six opinions that did not exist. ABA Formal Opinion 512 does not prohibit generative AI; it says your existing duties of competence, confidentiality, client communication, and reasonable fees still apply, so you own the accuracy of everything you file. The division of labor that holds up: Claude drafts, summarizes material you supply, and rewrites things in plain English, and every citation and quote comes out of your own research service before it reaches a filing.
Do I need Claude Pro for my business, or is free enough?
Free is enough to find out whether Claude helps your business, and most owners should start there. The free plan includes projects, capped at five, plus code execution and file creation that produces Excel, PowerPoint, Word, and PDF files, all under a session usage limit that resets every five hours. Pay for Pro, $20 per month in the US, when you keep running into that limit in the middle of real work or when you need more than five projects. Move to the Team plan, which requires a minimum of two members, only when you need to share projects with staff and manage accounts centrally.
Will Claude replace my staff?
For a US service business, whether that is fifteen people or five hundred, the realistic answer is no. Claude makes the first draft of repetitive written work faster, so the same team absorbs more volume, but it cannot answer your phone line, walk a job site, examine a patient, appear in court, or hold a license. Anthropic's own Usage Policy requires a qualified professional to review AI output in legal, healthcare, insurance, finance, and employment or housing contexts, and it puts responsibility for accuracy on your organization. Anyone attaching a specific headcount reduction to your business without having seen it is selling something.
How do I roll out Claude to my whole team?
Start with one named recurring deliverable on one team rather than seats for everyone. Pick a team whose work is repetitive and written, build the standard prompt and shared project on your own documents, standardize centrally what costs you money when it varies, name one person who owns adoption after launch, and set a thirty-day test you can observe, such as whether a specific recurring deliverable now starts from a draft every time. Claude Team and Claude Enterprise can both host a rollout at this size; Enterprise adds SCIM, audit logs, custom retention, and group-scoped custom roles, and its seat fee covers access only while usage bills separately at API rates. The failure mode is organizational rather than technical: licenses get purchased, nobody owns the change, and the seats sit unused.
What is the difference between Claude Team and Enterprise?
Team and Enterprise are both commercial plans that do not train on your content by default, but Enterprise is where the security and compliance controls live. Team includes SSO, central billing, admin controls, and spend limits, is priced per member per month with usage included, requires a minimum of two members, and caps at 150 seats. Enterprise adds SCIM, audit logs, the Compliance API, custom data retention controls, customer-managed encryption keys, US-only inference, custom roles and groups, IP allowlisting, and the HIPAA-ready configuration with a BAA, and it prices the seat for access only with all usage billed separately at API rates. Enterprise is annual only, with a 20 seat minimum if you buy it self-serve and a 50 seat minimum through sales.
How much does Claude cost for a team?
Claude for a business is sold per seat per month in two commercial tiers. Anthropic's published US prices are a Team Standard seat at $25 billed monthly or $20 billed annually, a Team Premium seat at $125 monthly or $100 annually, and an Enterprise seat at $20 per user per month billed annually with every token of usage billed separately at API rates on top. Team runs from 2 to 150 seats, self-serve Enterprise starts at 20 seats, and sales-assisted Enterprise starts at 50. Anthropic states on those same pages that prices are subject to change, so confirm the live number before it goes in a budget, and expect the license to be the smaller half of what the rollout actually costs.
Will Claude pass our IT security review?
Usually yes, but which plan you buy decides most of the answer. Anthropic's trust center shows SOC 2 Type 2, ISO 27001, ISO 42001, CSA STAR, and NIST 800-171 coverage for both Claude Team and Claude Enterprise, with HIPAA coverage on Enterprise and the API but not on Team, and the underlying reports sit behind an access request form. Single sign-on is available on Team and Enterprise, but SCIM directory provisioning, audit logs, the Compliance API, custom roles, and custom data retention controls are Enterprise only, so a reviewer who requires automated deprovisioning and an audit trail is effectively requiring Enterprise. The gap most reviews miss: Anthropic states that data is stored in the United States and its platform docs list US as the only available storage region, so a non-US data residency requirement is a real blocker rather than a setting.
Can I control what my staff put into Claude?
Less than most buyers expect, and the controls that actually stop someone mid-paste are Enterprise-only. Claude gives administrators real enforcement over who has an account, which models and connectors are available, how long data is kept, and which networks can reach it, and Enterprise adds inference hooks, a beta feature that sends every prompt to a policy server you host and waits for an allow or deny before Claude responds. On Team you can set organization instructions and manage members, but you cannot inspect prompts, cannot set a data retention period, and cannot claim existing personal accounts on your domain. Everything else is policy and training, which is why a written acceptable use policy your staff have actually read carries most of the real weight.
We bought Claude seats and nobody uses it. What now?
Do not relaunch. Low usage after a Claude rollout almost always traces to a small number of causes: no named workflow anyone was asked to do differently, no owner accountable for a specific outcome, access that was never actually granted, unclear data rules that produce quiet avoidance, or a genuine mismatch with that role's work. Diagnose which one you have from your own Analytics page, which reports active members against assigned seats and the percentage of users who have run even one chat or project, then restart with one recurring deliverable and one team instead of a second announcement. If a role turns out to have no work Claude improves, reclaiming that seat is a correct answer rather than a failure.
We already have Microsoft Copilot. Why would we add Claude?
Often you should not, at least not yet. What a base Microsoft 365 subscription includes is Microsoft 365 Copilot Chat, which Microsoft's own documentation describes as grounded in data from the web only and not in your files, emails, or chats; grounding in your tenant data through Microsoft Graph requires the paid Microsoft 365 Copilot add-on license. Microsoft has also onboarded Anthropic as a Microsoft subprocessor and turns Anthropic models on by default for most commercial cloud tenants outside the EU, EFTA, and the UK, so Claude inside Copilot is often a setting in your admin center rather than a second purchase. The real case for buying Claude directly is a narrow set of jobs, mostly long-document reasoning and drafting that happens outside the Office apps, and if almost nobody at your company actually uses Copilot yet, fix that before you buy anything else.
What is Claude for Small Business, and do I need it?
Claude for Small Business is a toggle install inside Claude Cowork, not a separate subscription tier. Anthropic launched it on May 13, 2026 as a package of 15 ready-to-run agentic workflows and 15 skills wired to Intuit QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365, and Claude can now send, post, and pay through those tools with you approving first. It requires a paid Claude plan, because Cowork is available on Pro, Max, Team, and Enterprise only and never on Free. Whether you need it comes down to one question almost nobody asks: those pre-built workflows are written against a small-business finance stack, so if your real system of record is a practice management system, an EHR, a legal case management system, a field service platform, or a mid-market ERP, the workflows will not touch the software where your work actually lives.
Should I hire someone to implement Claude, or do it ourselves?
Most firms should do the first round themselves, and it takes about a week of part-time effort: pick one recurring written deliverable, write the prompt against your own documents, put it in a shared project, and have one person own it. Outside help earns its money in three situations, which are a rollout that already stalled once, a regulated workflow where a wrong answer creates real liability, and the case where the work lives in a system of record that no connector reaches. Ask anyone you are evaluating for a fixed-scope first engagement with a named deliverable rather than a monthly retainer, because a retainer before anything ships transfers all the risk to you. If a vendor leads with headcount reduction or a percentage saved, they have not seen your operation and are quoting a brochure.
What should our AI policy actually say?
Six things: which Claude plan and account staff may use, what may and may not be typed into it, who reviews output before it reaches a client, when you disclose AI involvement, who is accountable when the output is wrong, and what happens when someone breaks the rules. Two of those are not optional in a regulated service line, because Anthropic's Usage Policy already requires that a qualified professional review AI-assisted advice before it is finalized and that you disclose AI involvement to the person receiving it. Anchor the whole document to one named plan, because Anthropic's consumer terms and commercial terms handle your data differently, and a policy written against the wrong one governs nothing you actually run. Everything below is a starting point drafted for a US service business, not legal advice, and employment policy is jurisdictional enough that a lawyer where you operate should review it before you make it binding on employees.
Still not sure where Claude fits?
Answer four questions and Claude writes three automations for your specific business, with the prompts and an honest list of what to skip.
Build my planWant these answers in your inbox?
The questions that block people first, answered with the sources linked so you can forward them. Then four short emails on making Claude stick. Unsubscribe any time.