Small businesses don’t seem to want an AI that runs the company for them.
They want one that gets Tuesday afternoon’s admin finished faster.
That is one of the clearest lessons from a new Anthropic field report based on working directly with more than 1,000 small business owners across the United States.
Over six weeks, Anthropic ran free half-day AI workshops in ten US cities. Most participants came from businesses with between 5 and 50 employees, including construction companies, manufacturers, logistics businesses, trades and professional services.
The results offer a useful look at what SMBs actually want from AI once you move beyond demos and hype.
Roughly two-thirds of the requests Anthropic encountered were about running the business more efficiently, while around one-third focused on growth.
One of the most common requirements, after marketing, was remarkably simple: take information scattered across several business systems and turn it into one useful report.
That could mean combining accounting data, CRM activity and spreadsheets into a weekly management report without someone spending Monday morning copying numbers between applications.
Claude for Small Business is built around these everyday jobs
Anthropic’s answer is Claude for Small Business, which connects Claude to tools companies already use, including QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace and Microsoft 365.
Instead of simply providing another chatbot window, the system is designed around recognizable business workflows such as chasing unpaid invoices, preparing month-end reports, qualifying leads, reviewing contracts, drafting marketing campaigns and planning payroll.
Importantly, existing permissions carry across. An employee who cannot access a particular QuickBooks report should not suddenly gain access to it through Claude.
Anthropic also says conversations on Team and Enterprise plans are not used for model training by default.
What are small businesses actually building with AI?
Some of the examples from Anthropic’s workshops show how practical these applications can be.
A cleaning company replaced paper scheduling processes with a live dashboard. A technician used photographs to identify unlabeled parts. A manufacturer created a reconciliation tool during a guided workshop. A trucking compliance company built a workflow to reduce fuel-tax errors after an existing software vendor gave it just 30 days’ notice.
None of these businesses needed a futuristic autonomous AI company.
They needed a specific annoying job fixed.
That distinction matters.
Anthropic also found that owners were more concerned about accuracy and privacy than about AI replacing them or their employees.
Businesses getting useful results tended to treat Claude more like a new employee than an all-knowing machine. They asked it to explain its work, identify assumptions and request approval before sending, publishing or paying anything.
The hardest part was often building the first useful workflow. Once owners saw one repetitive process working reliably, they quickly started identifying others.
What should an SMB automate first?
For a small business already using tools such as QuickBooks, HubSpot, Google Workspace or Microsoft 365, there is no need to start with an ambitious “AI agent that runs my business.”
Pick one job you repeat every week.
For example, connect accounting and CRM information and automatically prepare Monday’s management report. Turn a discovery-call transcript into the first draft of a proposal. Find overdue invoices and prepare personalized follow-up emails. Summarize new leads and identify which ones deserve attention first.
Keep sending emails, making payments and other consequential actions behind human approval initially.
If that workflow saves two hours every week, you have something measurable. Then automate the next one.
This is also a useful lesson for agencies and consultants
For agencies, automation specialists and consultants, Anthropic’s report is effectively a list of things SMB clients are willing to solve.
Reporting. Proposals. Scheduling. Lead qualification. Invoice follow-up. Marketing production. Data reconciliation.
The specific AI platform matters less than the workflow.
You could implement these processes with Claude, ChatGPT Work, automation platforms or a combination of tools. The commercial opportunity is identifying a painful recurring process, connecting the systems the client already uses and proving that the automation saves measurable time or money.
That is a much easier AI proposition for an SMB to understand than “multi-agent orchestration.”
Don’t sell them an AI transformation.
Fix the job they hate doing every Friday.
Product background: Anthropic – Claude for Small Business
