Claude for Small Business and the Boring Stuff

Anthropic just announced Claude for Small Business, a packaged version of Claude aimed at the part of the market that has been quietly running spreadsheets and email threads for the last twenty years. The press coverage is what you would expect: marketing copy, content drafts, support replies, sales emails. Useful work. Real ROI.

But that is not where most small businesses actually lose time.

Adobe puts document challenges at 21.3% of productivity loss, costing roughly $19,732 per information worker per year. The SBE Council's 2026 survey found that 82% of small business employers have already invested in AI tools. Almost none of that AI spend is touching the most expensive problem they have.

The boring stuff. Intake. Documents. Signatures. Follow-ups. The work between "we won the client" and "we can actually start the work."

This is the part Claude for Small Business is going to be very good at, but only if you connect it to the right tools.

What Claude for Small Business actually plugs into

Out of the box, Claude for Small Business meets you in the apps you already use. Calendar. Email. Files. Generally useful. Generally also not the bottleneck.

For most small businesses, the bottleneck is not "draft a follow-up email." The bottleneck is the entire chain after the customer says yes. Send the agreement. Ask for the documents you need before you can start the work. Chase the missing ones. Get the signature. Move the file forward.

Claude does not solve that on its own. It solves it when you point it at a tool that can actually do the work.

The shift from suggesting to doing

The interesting thing about agentic AI is the difference between an assistant and an actor. Claude in the chat window suggesting an email is helpful. Claude with a connector that can create an intake request, attach the right document template, dispatch it to the right client, and watch for the upload is something else. That is when the boring stuff starts disappearing from your week.

This is the gap most "AI for small business" coverage misses. The model is not the constraint. The connectors are. If the AI cannot reach into the system where your clients live, your documents live, and your tasks live, it can only offer you words about the work. It cannot do the work.

We made the longer version of this argument in The Agentic Approach to Business Automation. The short version: traditional automation is brittle because someone has to keep wiring every new branch. Agents handle the branching themselves, as long as the tools underneath them can actually do things.

What this looks like for a real small business

Take a mom-and-pop landlord with a couple of rental properties. A prospective tenant emails on a Monday saying they are interested in the Elm Street unit. You want a completed rental application, a photo of their ID, the last two pay stubs, and references from their previous landlord. Most people in this position keep a Word doc with the list. Most people in this position still email it, file by file, every time a unit turns.

With Claude wired into Intake, the whole thing is one prompt. "Send Jamie Reyes the rental application packet for Elm Street." Claude pulls the right checklist, dispatches the request, sets the reminder cadence, and reports back when Jamie uploads. You did not click anything. You did not write a single email.

The pattern repeats across almost every small business. An auto shop wants a signed work authorization and a photo of the insurance card before it cracks open a transmission. A wedding photographer needs a signed contract, a model release, and a planning questionnaire from every couple they book. A dog walker wants vet records, vaccination proof, and an emergency contact form before the first walk. A personal trainer needs a waiver, a health questionnaire, and a payment authorization before the first session.

Every one of these is a templated checklist trapped in a Word doc somewhere. Every one of them is a perfect job for an agent. And none of them get done by Claude alone, because Claude alone has nowhere to put the request, no way to send the link, and no way to know when the file lands.

The threshold question

A reasonable thing to ask is: do I really want an AI sending real emails to real clients? That hesitation is correct. We wrote a longer take on it in Trust but Verify, but the short version is that the answer changes the moment you have a way to watch the agent work before anything ships.

This is why every Intake account ships with sandbox mode. Two sets of API keys, one live and one test. You hand Claude the test key, let it run the full workflow against fictional clients, and inspect every email it tries to send in the Sandbox Inbox. Once the boring version is reliable, you swap dk_test_ for dk_live_ and it is real.

Trust, in other words, is not a feeling you have about Claude. It is a track record you build with Claude.

The practical first step

If you are running a small professional services business in 2026, the move is not to wait and see what Claude for Small Business turns into next quarter. The move is to find the one workflow you are most tired of running by hand and put an agent on it this week.

For us, the obvious candidate is client document collection, because it is where every professional services business loses the most time to coordination. But the principle applies anywhere. Pick a workflow. Find a connector. Run it in a sandbox. Ship it.

Here is what that looks like in practice:

1. Pick the workflow you would not miss. Intake, document chasing, engagement letters, anything templated.
2. Generate a test API key in Settings → API Keys.
3. Connect Claude to Intake using the MCP integration in the Intake dashboard.
4. Give Claude a real instruction with a fictional client. Watch what it does.
5. Open the Sandbox Inbox and read every email, every reminder, every follow-up it tried to send.
6. Iterate until the boring version is reliable. Then flip to a live key and stop doing that work yourself.

The thing that makes Claude for Small Business interesting is not that it is a smarter chatbot. It is that small businesses finally have a serious, well-resourced AI vendor pointing directly at them. The work that pays off is not in the chat. It is in the connectors you wire underneath it.

Create your Intake account if you do not already have one, and start with a single sandbox key. The boring stuff is more replaceable than it looks.