COMMENT "HIRE"Posted Oct 19, 2026
The AI agent that replaced a $24k/mo hire
A founder says one AI agent does the job of a $24k/month hire: it finds UGC creators, negotiates rates, and handles outreach. Here's how to build or sell it.
01Quick steps
- 1What it does: sources creators, drafts the offer, sends and tracks outreach.
- 2Real tools that do pieces of this today: Upfluence (Jaice), GRIN (Gia), Agentio.
- 3Build your own: use Claude + n8n. Step 1, scrape creators from a source. Step 2, have Claude draft a personalized offer. Step 3, send the DMs/emails and log replies.
- 4Sell it as a service: set it up for brands and charge a setup fee + monthly, or charge per creator booked.
02Your ready to use prompt
Hi [name], I run creator outreach for brands with an AI agent, it finds UGC creators, pitches them, and books them for you. Want me to find your first 10 and show you?
03Full tutorial
Before any tooling, write the job down as four stages: source creators, qualify them, draft a personalized offer, then send and track outreach. Keep two things human on purpose. A person approves every outbound message before it sends, and a person signs off on any rate or contract. The founder claim that one agent replaces a $24k per month hire is his own unverified number, so do not sell that promise. Sell a system that does the repetitive 80 percent and leaves judgment to a human. That framing is both more honest and easier to deliver.
Three real products already do pieces of this. Upfluence has an agentic assistant called Jaice that drafts briefs, contacts creators and negotiates (upfluence.com); public reviews put it around 478 dollars a month and demo-gated. GRIN has an assistant called Gia (grin.co). Agentio (agentio.com) is a YouTube and Meta marketplace where brands buy creator ad reads through automated bidding and it takes roughly a 20 percent transaction fee. Your lane is small and mid brands that cannot afford enterprise seats and want someone to just run it for them. You are selling the service, not out-engineering a funded platform.
Use n8n for the workflow (self-host free with Docker from docs.n8n.io, or n8n.io cloud). Use Claude as the drafting brain through an Anthropic API key from console.anthropic.com. n8n ships a native Anthropic credential and an Anthropic Chat Model node that attaches to its AI Agent node, so you do not need custom code to call Claude. Use one Google Sheet (or Airtable) as the shared database every creator record flows through. Keep the whole thing in one n8n workflow you can show a client on a screen share.
Add a trigger (manual or schedule) and a step that pulls candidate creators for a niche. The honest, terms-of-service-safe route is a legitimate data source rather than scraping a platform that forbids it: an influencer-data API, a public newsletter or directory, or a creator marketplace export. In n8n use the HTTP Request node to call that source's API, then a Google Sheets node (Append operation) to write one row per creator with handle, platform, follower count, contact and a blank status column. Respect each platform's terms and rate limits; aggressive scraping gets accounts banned and can carry legal risk.
Add an n8n Loop Over Items node to walk the sheet rows, then an AI Agent node with the Anthropic Chat Model attached. Prompt Claude with the brand brief and each creator's stats and ask for a fit score from 1 to 10 plus one sentence of reasoning, returned as JSON. Write the score back to the sheet. This is where you save real hours: a human now only reviews the top of a ranked list instead of reading every profile. Spot-check Claude's scores early, because a model will confidently rate a creator it knows nothing real about.
For qualified rows, call Claude again to draft a short outreach message that names something specific about that creator's content, states the deliverable, and gives a rate range you set (never let the model invent a budget). Keep messages honest: disclose it is a paid partnership and do not fabricate praise. Write the draft into a 'message' column. Tell Claude to output plain text with no emoji and no dashes so the copy reads like a person wrote it.
Do not auto-send. Add an n8n step that pauses for review: either a simple 'approved?' column a human sets to yes in the sheet, or n8n's Wait node with an approval link. Only rows marked approved move to sending. This gate is the single most important design choice. It keeps a human legally and ethically responsible for every message, stops the agent from spamming at scale, and is exactly what protects the client's brand and sender reputation.
Send by email, not by automating DMs on platforms that ban automation. Use an email node (Gmail or an SMTP or sending-API node) so you keep deliverability and an unsubscribe path, which also keeps you on the right side of anti-spam rules like CAN-SPAM. Throttle sends (a few per minute, small daily caps) so you look human and protect the domain. Only email creators who have a public business or booking address, and honor opt-outs immediately.
Add an inbound step (an email trigger or a scheduled inbox read) that matches replies back to the creator row and updates status to replied, interested or declined. Add one polite follow-up after a set number of days for no-reply rows, then stop. Now the sheet is a live CRM: a human opens it, sees who is warm, and takes over the actual negotiation and contracting. Resist the urge to let the agent negotiate money on its own.
Run the full workflow end to end targeting a tiny test list, with send addresses you control. Confirm scoring looks sane, messages read naturally, the approval gate actually blocks unapproved rows, and replies update correctly. Fix the prompts and the throttling before a real brand's name is ever on an outgoing email. A broken first campaign burns a client relationship and a sending domain at the same time.
Sell outcomes, not software. A common shape: a setup fee of roughly 500 to 1,500 dollars to configure the workflow and write the brand's templates, then a monthly retainer of about 300 to 800 dollars to run it and report results, or a performance model of a flat fee per creator actually booked. Put real limits in writing: how many creators sourced per month, that a human approves every message, and that you do not guarantee bookings. Start with one friendly brand at a low rate to get a real case study before you raise prices.
Keep API keys and the client's data out of the workflow's shared logs, and give each client their own sheet. Make sure outreach includes required disclosures and an opt-out, and never buy or scrape email lists. Be clear in your contract that the client owns their creator relationships and data and can export it any time. The platforms in this space (Upfluence, GRIN, Agentio) change features and pricing often, so re-check their current terms before you position against them to a prospect.
His $24k and $30M figures are his own claims. The build is real.