Verdict: You can build a $20,000/month solo or lean AI-operations business by selling small businesses a single high-impact Claude workflow, then charging a monthly retainer to tune and expand it. The opportunity is not selling "AI" — it is selling the hours the owner and their team get back.
Small businesses already use AI, but most stop at the chat window. A 2025 U.S. Chamber of Commerce and Teneo survey found that 58% of small businesses use generative AI, up from 40% in 2024 and more than double the 23% rate in 2023 [U.S. Chamber, 2025]. Meanwhile, Salesforce's 2024 SMB research found that 91% of AI-enabled SMBs attribute revenue growth to the technology [Salesforce, 2024]. Yet few small businesses have connected AI to the repetitive tasks that actually drain time: lead follow-up, invoice chasing, review responses, missed-call recovery, and daily email triage.
Anthropic's recent tooling is designed exactly for this gap. Claude Skills (launched October 16, 2025) are reusable instruction documents that teach Claude a specific process [Anthropic, 2025]. Claude Code Routines (research preview, April 14, 2026) run saved Claude Code configurations automatically on a schedule, API call, or event [Anthropic, 2026]. Combined, they let a non-coder build a cloud-running, human-approved workflow that acts like a trained employee rather than a generic chatbot.
What Claude skills and routines actually do
A Skill is a markdown-style process document — often a SKILL.md file — that tells Claude exactly how to perform a repeatable task: the tone, format, decision logic, and edge cases. Skills load only when relevant, so they do not waste context tokens [Anthropic, 2025]. Anthropic ships pre-built skills for Excel, PowerPoint, Word, and PDF, and lets anyone create custom skills in Claude Code, Claude.ai, and the API [Anthropic, 2025].
A Routine is a saved Claude Code configuration — prompt + repo + connectors — that runs automatically. Triggers include [Anthropic, 2026]:
- Schedule: hourly, nightly, or weekly cadence.
- API call: each routine gets its own HTTP endpoint and token.
- GitHub event: new PRs, comments, CI failures.
Routines run on Anthropic-managed cloud infrastructure, so they do not depend on your laptop being on [Anthropic, 2026]. Daily limits are Pro 5/day, Max 15/day, Team/Enterprise 25/day [Anthropic, 2026].
The real leverage comes from the combination. A routine without a skill is a generic Claude session; a routine with a skill is an expert running the client's exact playbook. That is the difference between "a chatbot they tried" and "a worker they keep paying for."
The five workflows that convert
Pick one workflow for the first client, prove ROI quickly, then add more. These five patterns address pain most small-business owners recognize immediately:
| Workflow | Trigger | What Claude does | Approx. time saved |
|---|---|---|---|
| New lead follow-up | Form submission via webhook | Drafts a personalized first reply in the client's voice within minutes | 5–10 hours/week |
| Negative review response | Review-platform webhook | Drafts a thoughtful, brand-tone reply for owner approval | 2–4 hours/week |
| Overdue invoice nudger | Daily routine | Checks aging invoices and drafts escalation-aware follow-ups | 3–6 hours/week |
| Missed-call recovery | Phone-system webhook | Sends text/email follow-up before the lead goes cold | 3–5 hours/week |
| Email triage | Daily 7:00 a.m. routine | Classifies inbox, drafts replies, surfaces what needs the owner | 5–10 hours/week |
All of these use human-in-the-loop approval before anything sends, posts, or pays. That approval step is what makes the service sellable to risk-aware owners and keeps the agency out of liability.
How to price for $20K/month
The math is deliberately simple: one-time setup plus recurring retainer.
- Workflow audit: $2,000. Map the client's manual processes, interview the team, score workflows by hours lost, and pick the first automation.
- Workflow build: $5,000–$8,000. Build one skill + one routine connected to the client's existing systems.
- Monthly retainer: $150–$200+ per workflow. This is where the revenue compounds.
To reach $20,000/month, you need roughly 10–13 clients on retainer at $150–$200/workflow, or fewer clients with multiple workflows each. Each new client becomes cheaper to serve because the skill library grows: the intake skill from client one becomes the starting point for client two.
Position the retainer as workflow operations, not maintenance. Include three things every month:
- Tune the skill as the workflow shifts — new products, new objections, changed tone.
- Ship one new routine to compound value and lock in the relationship.
- Send a workflow impact report showing hours and money saved.
The impact report is the retention engine. When the owner sees quantified savings, the retainer becomes obvious. The skill library itself becomes a switching cost: it carries the owner's tone, decisions, and edge cases. Switching providers means rebuilding from scratch.
How to land the first clients
This is the hardest part and the part most builders skip. Start with your existing network, not cold outreach.
- List 10 businesses you have talked to in the last year — dentist, contractor, property manager, agency, retailer. Anyone counts.
- Pick the three with the most chaos. Look for the ones where you think, "These guys are making money, but their operations are held together with spreadsheets and memory."
- Send a two-sentence message: "I'm doing a 45-minute free workflow audit this month for businesses like yours. Happy to map the one workflow that is costing you the most money."
- Sit down with the owner and one employee. Watch the actual work being done. Write down every step, every exception, every "I hate this" moment.
- Score workflows by hours lost and pick one. Build the automation live or in a sandbox, then demo it.
Conversion is high when the pain is real. The goal is not to find problems; it is to find workflows the owner already knows are broken.
From service to SaaS — if you want to scale
Running a service first is a low-risk way to discover product-market fit. Run the service for six months. If three clients in the same industry want the same skill stack, you have found a repeatable product. Package that stack as a SaaS and sell it as a vertical solution.
If no pattern emerges after six months, you were paid to learn the niche was wrong — far cheaper than building a product nobody wants. Foundation Capital calls this the "services as software" thesis: AI startups target the services market rather than just selling software seats [Foundation Capital, 2026].
What this means for you
If you are a solo operator or a small agency, the playbook is: audit → build one workflow → charge a retainer → repeat. You do not need to be a developer to start, but you do need to talk to real business owners and watch real work. The technology — Claude Skills, Claude Code Routines, and the growing connector ecosystem — is now good enough that the bottleneck is customer discovery, not coding.
If you are a small-business owner, the takeaway is the inverse: someone in your network is probably offering this service soon. The businesses that win will be the ones that map their worst workflow first, measure the time saved, and keep tuning it monthly.
Related reading
FAQ
Q: Do I need to know how to code to sell Claude workflow automations? A: Not for every workflow. Many use cases can be built with Claude Code on the web, pre-built connectors, and no-code triggers like webhooks from Typeform, Google Forms, or review platforms. Complex integrations — ERPs, custom APIs — may need a technical partner.
Q: How is this different from hiring a Zapier consultant? A: Zapier moves data between apps on rigid if-this-then-that rules. Claude Skills + Routines add judgment: tone-matched replies, escalation tactics, inbox classification, and exception handling. You are selling a decision-making layer, not just data plumbing.
Q: What if Claude makes a mistake and sends the wrong thing? A: Build every workflow with a human approval gate before any external send, post, or payment. Routines can draft; the owner clicks approve. This keeps the service safe and the client in control.
Q: Which industries are easiest to start with? A: Service businesses with high inbound volume and repetitive follow-up: local services (dentists, contractors, cleaners), property managers, agencies, and professional services. They feel the pain daily and measure ROI in hours saved.
Q: How do I justify $5,000–$8,000 for one workflow? A: Tie it to a number the owner already tracks. If the workflow saves 10 hours/week at $25/hour fully loaded, that is $10,000/year. If it reduces lead drop-off or speeds up invoice collection, the payoff can be weeks, not months. Show the math in the audit.
Q: Can I run this as a side business? A: Yes. The audit and build can happen evenings and weekends; the retainer work is front-loaded in the first month and then becomes lighter as skills mature. Many solo operators run 5–10 clients on the side before going full time.
Discussion
0 comments