AG1 runs affiliate-creator recruiting across 40 Gmail inboxes with Cheerful's AI reply and routing agent

L2 · Supervised Verified · Production Consumer Health Marketing

AG1 used Cheerful for an affiliate recruiting campaign: 20,521 creators contacted across 40 Gmail inboxes, about 2,460 replies handled with AI-drafted, context-aware responses ready for one-click send (tuned by 80 custom AG1 rules) in multiple languages, 261 FAQs answered automatically, 123 paid requests redirected to the affiliate program, 45 edge cases escalated, and 2,030 creator opt-ins written to Google Sheets over about seven weeks. Cheerful projects a 24x ROI ($36k value on about $1.5k spend) and estimates about 90% of the workload removed (reported by Cheerful; projections, campaign still active)

Problem

26,000+ creator email threads across 40 inboxes needed replies in many languages and data capture for 2,000+ creators, which Cheerful estimates would take 15-20 staff by hand (reported by Cheerful)

Workflow

  1. Unify 40 Gmail outreach inboxes into one workspace with AI status tracking
  2. Agent pre-drafts context-aware replies in the creator's language for one-click send, following 80 AG1 rules
  3. Agent routes paid-collaboration requests to the affiliate program and answers FAQs automatically
  4. Edge cases are escalated to the AG1 team
  5. Agent extracts names, handles and opt-in status into Google Sheets

Human-in-the-loop: Most replies are pre-drafted for a one-click human send; FAQs are answered automatically; edge cases go to humans.

Tools & stack

Tools

  • Cheerful

Integrations

  • Gmail (40 inboxes)
  • Google Sheets

Frameworks

None listed

Prerequisites

  • Connected outreach inboxes
  • Brand rules for replies and routing
  • An affiliate program to route paid requests to

Risks & limitations

Risks

  • Off-brand or inaccurate replies to creators at volume
  • Mis-routing paid partnership requests

Limitations

  • Vendor-published with no named AG1 source
  • ROI and workload figures are projections or estimates; method not disclosed
  • Campaign was still active when published, so the window is about seven weeks
  • Instagram view counts on the page are engagement figures and are not used here

Outcomes & evidence

Reported outcomes

  • 2,030 creator opt-ins captured from 20,521 creators contacted in about 7 weeks
    Reported by Cheerful · Source (cheerful.ai)
  • Projected 24x ROI: an estimated $36k value on about $1.5k spend (projection reported by Cheerful)
    Source: cheerful.ai · Source (cheerful.ai)
  • About 90% of the workload estimated as removed from the team (estimate reported by Cheerful)
    Source: cheerful.ai · Source (cheerful.ai)

Evidence

Vendors & implementers

Review history

Last reviewed by Research Chief (agent).

  • : Added in batch-15; source fetched 2026-10-19 and key figures confirmed in page text; figures attributed.

More marketing use cases

Marketing

260 Sample Sale Loomi marketing agent for weekly Last Chance targeting (Bloomreach)

260 Sample Sale has Bloomreach's Loomi marketing agent rescan its 900K+ contacts for each week's sale theme and build the "Last Chance" send list, shrinking it from 800K+ recipients to 36K high-intent shoppers; conversion rose 2.4x, from 1.44% to 3.45% (reported by Bloomreach)

L3 · Conditional Verified · Production

Retail — Marketing

Marketing

360 MMS compliant multi-brand campaign building with Agentforce Marketing

Australian non-bank lender 360 MMS runs marketing for four brands on Agentforce Marketing, with pre-approved brand templates that carry Spam Act and credit-licensing disclosures into every send; campaign build time fell from up to four hours to 15 minutes (reported by Salesforce / 360 MMS)

L2 · Supervised Verified · Production

Financial Services — Marketing

Marketing

7-Eleven multi-agent creative assistant

LangGraph multi-agent system on Databricks generates campaign concepts and copy in brand voice.

L2 · Supervised Verified · Production

Retail — Marketing

Marketing

Adore Me's merchandising and copy teams built WRITER agents that rewrite product descriptions to each marketplace's rules, batch-localized 2,900 descriptions for its Mexico launch and draft stylist notes

Adore Me, a roughly 500-person intimate-apparel retailer, uses WRITER AI Studio, with a non-engineer 'prompt coordinator' role, to build content agents. A marketplace agent takes a product ID, writes descriptions that meet each channel's rules (Macy's, Target, Walmart, Amazon, Victoria's Secret) and can publish to the Adore Me site. For Adore Me Mexico, a prompt coordinator, a merchandiser and a native speaker built a two-step translate-then-localize prompt and ran 2,900 product descriptions as a batch through the API. Earlier agents write SEO product descriptions and draft the 400-500-word notes stylists write more than 1,000 times a day for its Dailylook try-on service. Reported results: the marketplace process went from 20 hours a month to 20 minutes, the Mexico content took ten days instead of an estimated months, stylist note-writing time fell 36%, and non-branded search rose 40% (reported by WRITER, citing Adore Me's SVP of Strategy; self-reported, 2024 interview)

L2 · Supervised Verified · Production

E-commerce — Marketing

Marketing

Algolia's marketing team builds a pricing-page A/B test with Amplitude's Website Conversion Agent instead of waiting for developers

Algolia's website optimization manager needed a pricing-page A/B test that required custom JavaScript, CSS and HTML, which could have taken months through a busy dev team. She described it to Amplitude's Website Conversion Agent, which built several working variants, iterated with her on look and function, and configured a two-variant experiment ready to launch within hours. The test drove a 15% increase in its target metric and an estimated six-figure pipeline impact (reported by Amplitude; one experiment, pipeline impact is an estimate)

L2 · Supervised Verified · Production

SaaS — Marketing

Marketing

Anthropic's marketing operations team runs scheduled Claude Cowork skills for the weekly metrics report and event builds

Two Anthropic marketing-ops staff moved manual work into Claude Cowork. A Sunday-evening scheduled task reads last week's review, the latest meeting transcript and Slack, queries the data warehouse and leaves draft metrics and suggested headlines; a proofreading skill checks every number against a verified source. For campaigns, an hourly dispatcher skill picks the most urgent Slack intake request and hands it to specialist skills that build the event across Salesforce, HubSpot, Swoogo and email tools, then a fresh audit agent submits a test registration before a human reviews it. The weekly report went from up to two days of work to up to two hours (vendor's own product; reported by Anthropic); event-build savings are not quantified

L2 · Supervised Verified · Production

Technology — Marketing

See all AI agent use cases for marketing

Newsletter

Get new agent use cases in your inbox

A short roundup of newly added and newly verified use cases. Unsubscribe any time.

By subscribing you agree to receive our newsletter by email. We'll send a confirmation link first. We store your email address and signup date only to send it, and you can unsubscribe at any time. See our privacy policy.