# How to launch an AI automation agency offering voice AI agents for local businesses > Source: > Published: 2026-08-22 00:32:35+00:00 You'll build a repeatable service that lets plumbers, dentists, and other service-business owners answer calls with a natural-sounding, AI-driven voice that schedules appointments, qualifies leads, and captures payments. The result is a hands-free phone front-desk that you can sell as a monthly subscription and use to acquire new clients for your agency. **What you'll get:** a working n8n workflow that wires Anthropic's Claude, ElevenLabs text-to-speech, and Twilio Programmable Voice together, plus a go-to client-acquisition script that turns the service into a scalable AI automation agency. | Tool | Plan / Price* | Role | |---|---|---| | n8n (self-hosted Docker) | Free (self-hosted) - see Docker Hub for latest image | Orchestrates API calls, stores conversation state | | Twilio Programmable Voice | Pay-as-you-go - check Twilio pricing page | Provides inbound phone numbers and SIP bridge | | Anthropic Claude API | Usage-based - check Anthropic pricing page | Generates conversational replies | | ElevenLabs TTS API | Usage-based - check ElevenLabs pricing page | Turns Claude's text into a lifelike voice | | Cloudflare DNS + SSL | Free tier available - verify limits | Publishes a secure webhook for Twilio | | Git (optional) | Free | Version-controls workflow JSON | *We avoid stating exact free-tier caps; always verify the current provider pricing. **Estimated time-to-build:** 12-16 hours total (including testing and client-onboarding script). **Voice AI** is the combination of speech-to-text, natural-language generation, and text-to-speech that lets a computer hold a phone conversation. In this guide we skip the speech-to-text step by letting Twilio forward the caller's audio to our n8n webhook; the rest happens via APIs. Key insight:The biggest revenue lever for an AI automation agency is the repeatable client-acquisition funnel, not the underlying technology. Below is a step-by-step walkthrough. Every step mentions the exact UI field, API endpoint, or n8n node name so you can copy-paste without guessing. ``` docker run -d \ --name n8n \ -p 5678:5678 \ -e N8N_BASIC_AUTH_ACTIVE=true \ -e N8N_BASIC_AUTH_USER=admin \ -e N8N_BASIC_AUTH_PASSWORD=CHANGE_ME \ n8nio/n8n:latest ``` `voice.youragency.com` ) to your server's IP in Cloudflare DNS. `https://voice.youragency.com/webhook` to your n8n port `5678` . When you browse to `https://voice.youragency.com/webhook` , you should see a JSON "Welcome" response - that tells you the endpoint is reachable. `https://voice.youragency.com/webhook` . `CallSid` and raw audio stream to your n8n webhook every time the number is dialed. Open n8n, click **New Workflow**, and add the following nodes in order: | Node | Purpose | |---|---| Webhook | Receives Twilio's inbound call payload | Set (named Extract Speech) | Pulls `CallSid` and `From` fields into variables | HTTP Request (Anthropic) | Sends the conversation transcript to Claude and gets a text reply | HTTP Request (ElevenLabs) | Sends Claude's reply text to ElevenLabs TTS and receives an audio URL | Twilio (Make Call) | Plays the TTS audio back to the caller | NoOp (End) | Returns a `200 OK` to Twilio | *What this does:* Calls Claude with a prompt that includes the latest caller input and a short service-specific script. ``` { "name": "Anthropic Claude", "type": "n8n-nodes-base.httpRequest", "position": [600,200], "parameters": { "url": "https://api.anthropic.com/v1/complete", "method": "POST", "jsonParameters": true, "options": { "bodyContentType": "json" }, "bodyParametersJson": { "model": "claude-3-sonnet-20240229", "max_tokens": 150, "temperature": 0.3, "prompt": "You are a friendly office assistant for a local {{ $json.service_type }}. The caller says: \"{{ $json.caller_text }}\". Respond with a concise answer, ask for the needed information (e.g., appointment date), and keep the tone professional." }, "authentication": "headerAuth", "headerAuth": { "name": "Authorization", "value": "Bearer {{ $env.ANTHROPIC_API_KEY }}" } } } ``` **Why it matters:** The prompt is service-specific (`{{ $json.service_type }}` ) so the same workflow can serve plumbers, dentists, or any other *service business automation* client. ``` { "name": "ElevenLabs TTS", "type": "n8n-nodes-base.httpRequest", "position": [900,200], "parameters": { "url": "https://api.elevenlabs.io/v1/text-to-speech/{{ $env.ELEVENLABS_VOICE_ID }}/stream", "method": "POST", "jsonParameters": false, "options": { "bodyContentType": "raw", "responseFormat": "json" }, "bodyParametersJson": { "text": "{{ $node['Anthropic Claude'].json.completion }}" }, "authentication": "headerAuth", "headerAuth": { "name": "xi-api-key", "value": "{{ $env.ELEVENLABS_API_KEY }}" } } } ``` The node streams the generated audio back to Twilio in the next step. Add a **Twilio** node (type *Make Call*) and set: `{{ $json.From }}` (the original caller) ``` {{ $node['ElevenLabs TTS'].json.audio_url }} ``` When the workflow finishes, the caller hears the AI-generated response as if a human were on the line. Create a second **HTTP Request** node that POSTs the call details to a Google Sheet via Zapier or directly to HubSpot. This gives you an automatic **client acquisition** log for each conversation, letting you track lead quality and invoice per-call usage. `https://getaab.com/free` and link it from your agency landing page - this is the `https://getaab.com/ai-automations-to-sell` ) to outline pricing tiers and upsell options (e.g., multilingual agents, call analytics). | Failure mode | Symptom | Fix | |---|---|---| Twilio webhook not reachable | 11200 "HTTP retrieval failure" in Twilio console | Verify Cloudflare DNS, ensure port 5678 is open, and that the HTTPS cert is valid. | Anthropic rate limit | 429 error in n8n logs, no reply text | Implement a n8n Retry node with exponential back-off; consider upgrading to a higher usage plan. | ElevenLabs voice ID change | 404 from ElevenLabs, empty `audio_url` | Store the voice ID as an environment variable and update it whenever you create a new custom voice. | Expired API keys | 401 Unauthorized from either API | Rotate keys regularly; add a Cron node that emails you when a request returns 401. | Cost blow-out | Monthly invoice > forecast | Set usage alerts in Twilio, Anthropic, and ElevenLabs dashboards; cap the number of calls per client in the n8n workflow with a IF node. | Caller hangs up before TTS finishes | Twilio logs "Call ended" with no audio played | Reduce `max_tokens` in the Claude prompt and enable EleventLabs streaming to lower latency. | Bottom line:Most breakages are network or quota-related; proactive monitoring stops surprise bills before they happen. For a deeper technical reference, see [n8n's documentation](https://docs.n8n.io/). Pick a voice that matches the brand tone: a warm, confident male voice for a plumber, a friendly female voice for a dentist. ElevenLabs lets you audition voices in the dashboard; select the `voice_id` and store it in the `ELEVENLABS_VOICE_ID` env var. Yes. Replace the **Set** node with a Twilio **Gather** action that records the caller's speech, then send the audio file to a STT service such as Google Speech-to-Text. The transcript becomes `caller_text` for the Claude prompt. `https://getaab.com/free` ). You'll pay per-minute for Twilio inbound minutes, per-token for Anthropic, and per-character for ElevenLabs. The exact amount varies with call length and prompt complexity; consult each provider's pricing page for up-to-date rates. No. You can run n8n in the cloud (e.g., n8n.cloud) if you prefer not to maintain a server, but self-hosting eliminates recurring SaaS fees and gives you full control over data privacy - critical for many *service business automation* clients. Ready to start building your **voice ai agents for local businesses** service? Grab the free guide, set up the workflow, and begin acquiring plumbers, dentists, and other local pros today. *Internal resources:*