Businesses don't buy "automation for the sake of automation." They pay for concrete workflows that save time, cut costs, and unlock revenue. Below you'll see ten ai automations businesses pay for, the typical price per execution, and the buyer persona that values each. I then walk you through building one of them - an AI-enriched lead-to-CRM pipeline - using n8n, OpenAI, and a webhook. The guide is detailed enough for a seasoned builder, precise enough for Google's crawlers, and clear enough for answer engines.
Only the automations that directly impact cash flow or customer experience get budget approvals.
What is an ai automation?
An ai automation is a workflow that combines a trigger (e.g., a new web form submission) with one or more AI-powered actions (e.g., text classification, summarisation, or vector search) and finishes with a concrete business outcome (e.g., a record in a CRM, an invoice emailed to a client).
Which ai automations businesses pay for
| # | Automation | Typical price per execution* | Who buys it |
|---|---|---|---|
| 1 | Lead capture → enriched CRM record (OpenAI text enrichment) | $0.0006 per 1 k tokens (OpenAI-gpt-3.5-turbo) | SaaS founders, B2B marketers |
| 2 | Invoice generation from email (LLM-template fill) | $0.0012 per 1 k tokens | Accounting firms, freelancers |
| 3 | Customer-support ticket triage (sentiment + category) | $0.0008 per 1 k tokens | Support teams, SaaS platforms |
| 4 | Churn-risk scoring (historical data + vector similarity) | $0.003 per 1 k tokens + Pinecone query cost | Subscription services |
| 5 | Meeting-summary email (audio transcription + LLM summarisation) | $0.015 per minute (Whisper) + $0.001 per 1 k tokens | Sales teams, consultants |
| 6 | Product-feedback clustering (topic modelling) | $0.0009 per 1 k tokens | Product managers |
| 7 | Automated quote generation (price-rule engine + LLM) | $0.001 per 1 k tokens | B2B sales, procurement |
| 8 | Document extraction → searchable vector DB (OCR + embedding) | $0.002 per 1 k tokens + Pinecone storage | Legal firms, research groups |
| 9 | Dynamic FAQ chatbot (retrieval-augmented generation) | $0.0015 per 1 k tokens + Pinecone query | E-commerce sites |
| 10 | Weekly KPI dashboard update (data pull + LLM narrative) | $0.0005 per 1 k tokens | Executives, data teams |
*Prices are based on OpenAI token pricing (see https://openai.com/pricing) and typical third-party costs. Rates can vary with model choice and payload size.
What you need
| Tool | Plan / Price | Role |
|---|---|---|
| n8n (self-hosted Docker) | Free (self-hosted) - Cloud plan $20 / month | Orchestrator |
| Make (formerly Integromat) | Free tier up to 1 000 tasks/mo; paid plans start at $9 / mo | Alternative orchestrator |
| Zapier | Free tier 100 tasks/mo; paid plans start at $19.99 / mo | Quick-start prototyping |
| OpenAI API | Pay-as-you-go, $0.0006 per 1 k tokens for gpt-3.5-turbo | LLM engine |
| HubSpot CRM | Check HubSpot's current pricing for any free-tier or paid options | Customer data store |
| Pinecone | Check Pinecone's current pricing for vector storage and query costs | Vector similarity |
| AWS S3 (or any object storage) | Free tier 5 GB; standard storage $0.023 / GB / month | File storage |
| Git (for version control) | Free | Code management |
| Docker | Free | Container runtime |
Estimated build time: 4-6 hours for a complete end-to-end workflow, including testing and documentation.
Build example: AI-enriched lead capture to HubSpot CRM
Below is a step-by-step guide that you can copy-paste into an n8n workflow. The same logic applies in Make or Zapier with equivalent nodes.
1. Set up the n8n instance
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=supersecret \
n8nio/n8n
This launches a self-hosted n8n on port 5678 with basic auth. Verify it's running by visiting http://localhost:5678.
2. Create a Webhook node (trigger)
- Click + → Add Node → Webhook.
- Set HTTP Method to
POST. - Name the node
New Lead Webhook. - Copy the generated URL; you'll embed it in your website's lead form.
What this does: Receives raw lead data (name, email, company) from a web form.
3. Add a Function node to normalise payload
{
"name": "NormaliseLead",
"type": "n8n-nodes-base.function",
"position": [400, 200],
"parameters": {
"functionCode": "return [{\n name: $json.body.name,\n email: $json.body.email,\n company: $json.body.company,\n notes: $json.body.message || ''\n}];"
}
}
The node strips out any extra fields and guarantees a consistent schema.
4. Call OpenAI to enrich the lead description
Add an HTTP Request node:
-
Method:
POST -
URL:
https://api.openai.com/v1/chat/completions -
Authentication: Header
Authorization: Bearer <YOUR_OPENAI_API_KEY>(store the key in n8n's Credentials). -
Headers:
Content-Type: application/json - Body:
{
"model": "gpt-3.5-turbo",
"messages": [
{
"role": "system",
"content": "You are a concise business analyst."
},
{
"role": "user",
"content": "Summarise the following lead information for a sales team:\nName: {{$json.name}}\nEmail: {{$json.email}}\nCompany: {{$json.company}}\nNotes: {{$json.notes}}"
}
],
"max_tokens": 150
}
This generates a short, sales-ready summary of the raw lead.
5. Parse the OpenAI response
Add another Function node (named ParseSummary):
{
"name": "ParseSummary",
"type": "n8n-nodes-base.function",
"position": [800, 200],
"parameters": {
"functionCode": "const content = $json.choices[0].message.content;\nreturn [{ summary: content }];"
}
}
6. Push to HubSpot CRM via HTTP Request
- Create a HubSpot API Key (or OAuth token) in HubSpot settings.
- Add an HTTP Request node:
-
Method:
POST -
URL:
https://api.hubapi.com/crm/v3/objects/contacts -
Authentication: Header
Authorization: Bearer <YOUR_HUBSPOT_TOKEN> -
Headers:
Content-Type: application/json - Body:
{
"properties": {
"firstname": "{{$json.name}}",
"email": "{{$json.email}}",
"company": "{{$json.company}}",
"notes": "{{$node.ParseSummary.json.summary}}"
}
}
This creates a new contact in HubSpot with the AI-generated notes.
7. Optional: Send a Slack notification
Add a Slack node (or use a webhook) to alert the sales channel:
-
Channel:
#sales-leads -
Message:
New lead from {{ $json.name }} - {{ $node.ParseSummary.json.summary }}
8. Test end-to-end
- Submit a test lead via the webhook URL (e.g., using
curl). - Verify the contact appears in HubSpot with the summarised notes.
- Check the Slack channel for the notification.
curl -X POST -H "Content-Type: application/json" \
-d '{"name":"Jane Doe","email":"jane@example.com","company":"Acme Corp","message":"Looking for a SaaS solution"}' \
https://your-n8n-domain/webhook/new-lead-webhook
If the workflow runs without errors, you've built a production-ready AI-enriched lead capture pipeline.
Where this breaks
| Failure mode | Symptom | Fix |
|---|---|---|
| OpenAI rate limit (60 RPM on the free tier) | "429 Too Many Requests" response in the OpenAI node | Upgrade to a paid tier or add a Throttle node to cap calls at 50 RPM. |
| Expired HubSpot token | 401 Unauthorized in the HubSpot request | Store the token in n8n's Credential store with auto-refresh (OAuth) or schedule a token-renewal script. |
| Payload size > 2 MB (n8n webhook limit) | Webhook returns "Payload too large" | Trim fields in the Function node or use a signed URL to upload large files to S3 first. |
| Cost blow-up (high token usage) | Unexpected monthly bill from OpenAI | Log token count per execution ({{ $json.usage.total_tokens }}) and set alerts when daily usage exceeds a threshold. |
| Pinecone index quota exceeded (if added later) | 429 error from Pinecone query | Check Pinecone's current quota limits (see docs) and request a larger plan or implement query caching. |
| Slack webhook throttling | "rate_limited" error | Use Slack's "Retry-After" header to back-off or batch notifications into a digest. |
Never assume free tiers will stay free. Always monitor usage dashboards and set hard limits in your workflow.
For a deeper technical reference, see n8n's documentation.
FAQ
What price should I quote for building a lead-enrichment automation?
Check the guide on what to charge for automation services. A common model is a setup fee ($500-$2 000) plus a monthly maintenance fee (5 % of the client's projected token cost). Adjust based on complexity and ongoing support.
How do I find clients who will pay for these automations?
Read the article on finding automation clients. Target SaaS founders, e-commerce owners, and professional services firms that already spend on CRMs or support tools.
Can I replace n8n with a no-code platform like Zapier?
Yes, but Zapier's free tier caps at 100 tasks/month and its pricing jumps quickly for higher volumes. For production-grade workloads, the self-hosted n8n or Make plans are more cost-effective. Compare the trade-offs in the What you need table.
Are there hidden costs when using OpenAI's API?
OpenAI charges per 1 k tokens for both input and output. If you add temperature or max_tokens settings that increase response length, your cost grows linearly. Track usage with the usage.total_tokens field and set alerts in n8n's IF node.
How do I secure the webhook endpoint against spam?
Add a Header Validation node that checks a shared secret header (e.g., X-Auth-Token). Store the secret in n8n's Credentials and reject any request lacking the correct token.
Is there a free way to store lead data long-term?
You can use AWS S3's free tier (5 GB) or any self-hosted MinIO bucket. No additional cost until you exceed the free allocation.
Ready to start building? Grab the free starter kit and a checklist of the ten highest-value ai automations businesses pay for at https://getaab.com/free.
Top comments (0)