how to make $2K/month selling an AI agent you can build this weekend.

97% of consumers read online reviews before choosing a local business, but 87% of businesses never respond to negative reviews.
a one-star improvement on Google = 5-9% revenue increase. for a restaurant doing $500K/year, that's $25-45K in lost revenue.
this is a big gap. and you can fill it with a simple AI agent.
the agent flow would be:
1/ monitors Google Business Profile for new reviews (polling every 15 min)
2/ for positive reviews (4-5 stars): generates a personalized thank-you reply and posts it automatically
3/ for negative reviews (1-3 stars): generates a draft reply, sends it to the owner via Slack/email for approval before posting
4/ logs everything to a Google Sheet with sentiment tags
5/ sends a weekly digest: new reviews, avg rating trend, response time, sentiment breakdown
the magic is in the prompt engineering -> the replies don't sound like a bot. they reference specific details from the review, match the business's tone, and handle complaints with empathy.
the stack (total cost: ~$30/month)
> n8n (self-hosted) -> workflow automation (free)
> Google Business Profile API -> free, native n8n node available
> OpenAI API (gpt-4o-mini) -> ~$5/month for most volumes
> Google Sheets -> logging + dashboard, free
> Slack or Email -> notifications, free
> VPS -> $5-10/month (Hetzner or Railway for n8n hosting)
total operational cost per client: under $15/month. you charge $500.
step 1: set up n8n
spin up n8n on a VPS or use n8n cloud ($24/month if you don't want to self-host).
step 2: connect Google Business Profile
n8n has a native Google Business Profile node. here's the setup:
> go to Google Cloud Console -> create a new project
> enable the "Google My Business API" (now called Business Profile API)
> create OAuth 2.0 credentials (Web Application type)
> in n8n, add a Google Business Profile credential using OAuth2
> authorize access to your client's Google Business account
important: the client needs to add your Google account as a manager on their Business Profile. takes 30 seconds on their end.
step 3: build the review monitoring flow
create a workflow with these nodes:
> Schedule Trigger (every 15 minutes)
> Google Business Profile node (operation: Get Many Reviews)
> IF node (filter: only reviews without a reply + created in last 24h)
> Switch node (split by star rating: 4-5 stars vs 1-3 stars)
step 4: positive review path (auto-reply)
for 4-5 star reviews, wire it to:
> OpenAI node with this system prompt: "You are a friendly review response writer for [Business Name], a [business type] in [city]. Write a short, warm reply to this positive Google review. Reference specific details the customer mentioned. Keep it under 3 sentences. Sound human, not corporate. Never use exclamation marks more than once. Never start with 'Thank you for your kind words.' Be genuine."
> Google Business Profile node (operation: Reply to Review)
step 5: negative review path (human-in-the-loop) for 1-3 star reviews, wire it to:
> OpenAI node with a different prompt: "You are responding to a negative review for [Business Name]. Acknowledge the issue without being defensive. Offer to make it right. Include a specific next step (call this number / email this address). Keep it professional but warm. Never blame the customer. Never say 'we take all feedback seriously', that's corporate speak. Be real."
> Slack/Email node -> sends the draft reply to the owner with one-click approve/reject buttons
> on approval -> Google Business Profile node (Reply to Review)
this is critical. negative reviews need human oversight. one bad auto-reply can go viral for the wrong reasons.
step 6: logging + weekly digest
> Google Sheets node -> append every review with: date, rating, reviewer name, review text, AI reply, status (auto-replied / pending / approved)
> separate workflow: runs every Monday, reads the sheet, generates a summary via OpenAI, emails it to the client.
weekly digest includes:
- total new reviews this week
- average rating trend (vs last week/month)
- response time average
- top complaint themes
- suggested improvements based on negative review patterns
this digest alone is worth the retainer. it turns raw reviews into business intelligence.