Anthropic has shipped a genuine SMB stack: a Teams plan with shared Projects and admin controls, native integrations with Google Drive and Slack, MCP for connecting Claude to your own tools, and a new generation of models that cost far less to run at scale. If you're still treating Claude as a chat window, you're missing most of what it can now do.
For the first two years, Claude was a great model living inside a chat interface. Useful, but not operationally interesting for small businesses. You couldn't share context across a team, you couldn't connect it to your actual tools, and the pricing model didn't fit how small teams work.
That changed. Anthropic has quietly built out a feature set that closes most of the gaps. Here's a plain-English breakdown of what's now available and why it matters if you run a business with fewer than 50 people.
Projects: shared context that actually persists
Projects is the feature that finally makes Claude useful as a team tool rather than a personal one. You create a Project, upload the files and instructions that define your business context — brand voice guidelines, SOPs, pricing sheets, product documentation, past proposals — and every conversation inside that Project starts with all of that already loaded.
The context window is 200,000 tokens, which is roughly 150,000 words. That's enough to hold a complete business playbook, a full product catalog, or several months of customer conversation history, all simultaneously in context.
The practical result: instead of pasting your brand guide into every conversation, you set it once and it's always there. Instead of explaining your pricing to Claude every time you want a quote drafted, the pricing is already loaded. That's not a marginal improvement. It changes how you actually integrate AI into daily work.
Clients use Projects to house their brand voice, FAQ documents, and sales scripts. Every team member who opens Claude to draft a proposal or reply to a lead pulls from the same loaded context. It's the closest thing to "AI that already knows your business" that currently exists without building custom infrastructure.
The Teams plan: Claude for a whole business, not just one person
Anthropic's Teams plan runs at $30 per user per month (minimum five seats). That's the same underlying model access as Claude Pro, plus the things that actually matter for a business: shared Projects visible to everyone on the team, central admin controls, usage visibility, and the ability to manage who has access to what.
For a team that's been handing Claude Pro subscriptions to individual employees, the switch to Teams usually costs about the same per seat but adds the infrastructure that makes AI actually work as a shared resource instead of a set of disconnected individual chat windows.
The admin controls aren't glamorous but they're important. When someone leaves the company, you revoke their access from a central dashboard. When you want to set guardrails on what the team builds inside Claude, you can. When you want to see whether the team is actually using the tools you're paying for, the usage data is there.
Native integrations: Claude connects to where your work actually lives
The integrations that matter most for small businesses are now live. Claude connects directly to:
- Google Drive — grant Claude access to specific folders and it can read, summarize, and work with your documents, spreadsheets, and slides without you copying and pasting anything
- Slack — Claude can participate in channels, answer questions from your team, and surface information from your message history
- GitHub — useful for any team with developers; Claude can review code, answer questions about a codebase, and help with pull requests directly
The framing that helps people understand this is: Claude is no longer just a chat window you visit. It's becoming connective tissue between the tools your team already uses. You stop going to Claude with context. Claude starts being where the context already lives.
MCP: the thing that makes custom connections possible
MCP stands for Model Context Protocol. It's an open standard Anthropic published that defines how external tools connect to Claude. Think of it as a USB standard for AI — any tool that builds an MCP connector can now talk to Claude natively, without custom integration work on your end.
For small businesses, the immediate implication is this: if your CRM, your project management tool, your calendar, or any other software you use publishes an MCP server, Claude can read from and write to it using plain language. You describe what you want — "pull every deal that hasn't had contact in 30 days and draft a follow-up for each one" — and Claude has the access to actually do it.
The MCP ecosystem is growing fast. Hundreds of tools have already published MCP servers, and the number is accelerating. This is the infrastructure layer that will eventually let small businesses build AI workflows without writing code.
If you want a deeper look at MCP, we wrote a full breakdown: What is MCP and why does it matter for your business?
Cheaper, faster models: Claude Haiku 3.5
Most coverage of Anthropic focuses on the top-of-the-line models — Opus and Sonnet. But the model that matters most for high-volume small business use is Haiku.
Claude Haiku 3.5 is fast, inexpensive, and smart enough to handle the majority of repetitive business tasks: drafting replies, categorizing inputs, extracting data from documents, generating short content, routing inquiries. It costs roughly 80% less per token than Sonnet and responds in fractions of a second.
For a small business running any kind of automated workflow — lead enrichment, support triage, daily reports — Haiku is what you actually build on. You use Sonnet or Opus when a task genuinely needs deep reasoning. You use Haiku for everything else. Getting that split right is often the difference between an AI workflow that costs $40 a month to run and one that costs $400.
What this looks like in practice for a 5–20 person business
Put it together and the picture looks something like this for a small team that's serious about using AI operationally:
- Teams plan for the 3–5 people who use Claude daily (founders, ops, marketing, sales)
- Projects loaded with your brand guide, product info, SOPs, and any reference material Claude needs to work without briefing each time
- Google Drive integration turned on so Claude can read proposals, contracts, and meeting notes directly
- API + Haiku running background workflows: lead scoring, email drafting, report generation, support ticket categorization
- MCP connections to your CRM and calendar as those become available for your tools
Total monthly cost for a team of 10 with this setup: roughly $200–$400, depending on API volume. That's three or four Pro seats plus API usage. Not a massive budget line for what it replaces in human hours.
The gap that still exists
The honest answer is that accessing all of this still requires some setup and judgment that most small business owners don't have time to figure out solo. Which model for which task. How to structure Projects for maximum reuse. Which workflows are worth automating versus which ones are too variable to hand to AI reliably.
That's the work we do with clients. Not selling them software — helping them figure out where the leverage actually is and building the systems to capture it.
If you want to understand where your business specifically sits, the fastest path is a 30-minute call. We'll tell you what's worth your attention and what isn't.