Demo deployment — a fresh offline cycle, rebuilt on every deploy. Edits don't persist.

Circle

Design the ambassador program mechanics needed to reach 25 ambassadors: who qualifies, what they receive, the activation ladder from first yes to active advocate, the definition of "active", and the first outreach message. Buyers are growth and marketing leaders at startups and AI-forward brands. Anchor the work to the "Agents are doing the marketing work now" narrative where it fits honestly.

run_f905351a02664e21996fAmbassadorssucceeded
Created
Aug 5, 2026 20:53 UTC
Duration
138ms
Cost
Tokens (in / out)

Brief

what the engine asked for
Objective

Design the ambassador program mechanics needed to reach 25 ambassadors: who qualifies, what they receive, the activation ladder from first yes to active advocate, the definition of "active", and the first outreach message. Buyers are growth and marketing leaders at startups and AI-forward brands. Anchor the work to the "Agents are doing the marketing work now" narrative where it fits honestly.

Goal: 25 active ambassadors (active_ambassadors)

Constraints
  • no fabricated metrics or customers
  • no disparaging named competitors
  • Every claim must be defensible — prefer concrete mechanisms and numbers over superlatives.
  • Produce work product for human review; nothing is published to a live channel by the agent.
  • Ambassadors are practitioners, not resellers — incentives must never create pressure to overstate results.
  • Every individual ask of an ambassador should cost them under 15 minutes.
Market context (Adveron)

BRAND Brand: Waldo (waldo.fyi) — An AI startup building the data and orchestration layer for marketing agents. We run our own marketing on our own products, in public. Category: AI marketing tools Products: Adveron — Brand and audience intelligence data layer for marketing agents.; mos — Marketing OS that sets goals, orchestrates marketing agents, and learns from what they ship. Audience: growth and marketing leaders at startups and AI-forward brands Voice: sharp, technical but human, no hype Competitors: Brandwatch, Similarweb, SparkToro Constraints: no fabricated metrics or customers; no disparaging named competitors Brand perception — sentiment 0.59, share of voice 6%. Associated with: AI agents that do marketing work, answer-engine optimization (AEO), building in public, developer-grade marketing stack, small team, big surface area. Waldo over the last 90 days: Waldo is read as an agent-native challenger: people file it next to the AEO/LLM-visibility wave rather than next to legacy listening suites. The two-product story (Adveron as the data layer, mos as the orchestration layer) is understood by practitioners but blurs for non-operators. Strongest association is "AI agents that do marketing work"; the loudest objection is proof rather than concept. Trending narratives in AI marketing tools: "Agents are doing the marketing work now" (momentum 0.99) — The conversation moved from 'AI writes drafts' to 'AI runs the channel'. Practitioners are comparing agent output to junior-hire output and publishing the receipts. | "AEO is eating SEO" (momentum 0.90) — Answer-engine optimization has crossed from niche to default vocabulary. Teams are re-scoping SEO roles around being cited by LLMs rather than ranking on page one. | "Brand presence inside LLM answers is the new share of voice" (momentum 0.87) — Brands are discovering they are absent, mis-described, or beaten by a competitor inside ChatGPT/Claude/Perplexity answers, and treating that as a measurable brand-equity problem. Channel opportunities: ambassadors: Formalise the people already recommending you into an ambassador roster with early access, a private channel, and shareable experiment results. — why now: Advocacy already exists but is unstructured; the category's buyers trust practitioner recommendation far more than vendor content, and the roster compounds. (category: AI marketing tools; brand: Waldo) (effort low, impact high) | social: Founder-led build-in-public cadence on LinkedIn and X, shipping one head-to-head agent experiment result per week. — why now: Creator-led B2B is at peak momentum and experiment results are inherently screenshot-shaped — the format the feed rewards. (category: AI marketing tools; brand: Waldo) (effort low, impact high) | owned_media: Publish a recurring 'what AI marketing can't answer' data-gap report built from mos's own Adveron gap log. — why now: Nobody else can publish it — the artifact only exists because you instrument every query. It doubles as proof …

Performance context (connectors)

Performance summary — latest period ending 2026-08-05T20:53:31.354Z (1 period(s) on file, 62 snapshots scanned). Owned Media - engagement_rate: 3.54% [social_owned] - follower_growth: 495 followers [social_owned] - impressions: 135,393 impressions [social_owned] - platform_impressions: 135,393 impressions [social_owned] top: linkedin 59,347 impressions · x 51,543 impressions · youtube 24,503 impressions - posts_published: 9 posts [social_owned] - viral_posts: 1 posts [social_owned] SEO / AEO - avg_search_rank: 12.3 [search_console] - clicks: 298 clicks [search_console] - impressions: 21,362 impressions [search_console] - indexed_queries: 424 queries [search_console] - query_avg_rank: avg 10.4 [search_console] top: adveron 1.5 · brand intelligence api 6.8 · marketing os 8.1 · answer engine optimization 10.9 - query_clicks: 287 clicks [search_console] top: marketing os 77 clicks · answer engine optimization 59 clicks · adveron 57 clicks · brand intelligence api 39 clicks Paid Ads - cac_usd: avg $45 [google_ads, meta_ads] top: meta_ads $38 · google_ads $52 - campaign_spend_usd: $1,666 [google_ads] top: aeo-intent-terms $797 · competitor-conquest $503 · brand-defense $366 - clicks: 3,892 clicks [google_ads, meta_ads] top: meta_ads 2,036 clicks · google_ads 1,856 clicks - conversions: 51 conversions [google_ads, meta_ads] top: google_ads 32 conversions · meta_ads 19 conversions - impressions: 265,912 impressions [google_ads, meta_ads] top: meta_ads 197,665 impressions · google_ads 68,247 impress…

Result

what the agent produced
Summary

Ambassadors: 2 deliverable(s) drafted for "Design the ambassador program mechanics needed to reach 25 ambassadors: who qualifies, what they receive, the activation ladder from first yes to active advocate, the definition of "active", and the first outreach message. Buyers are growth and marketing leaders at startups and AI-forward brands. Anchor the work to the "Agents are doing the marketing work now" narrative where it fits honestly." from 4 Adveron intelligence section(s).

Metrics to watch
active_ambassadorsreferrals_generatedadvocacy_mentionsreferral_signupsambassador_retention_rate
Deliverables (2)

Ambassador program design: three tiers, one loop

program_design
Objective: Design the ambassador program mechanics needed to reach 25 ambassadors: who qualifies, what they receive, the activation ladder from first yes to active advocate, the definition of "active", and the first outreach message. Buyers are growth and marketing leaders at startups and AI-forward brands. Anchor the work to the "Agents are doing the marketing work now" narrative where it fits honestly.
Advocate pool: Growth & marketing leaders at seed–Series B startups (~142k globally; ~31k reachable in US/EU)
Perception to lean on: people already associate us with "brand intelligence as an API".
TIERS (each entered by doing something, not by applying)
  1. Signal — entered by: ran three planning cycles, or posted publicly about a result.
     Gets: private Slack channel with the team, early roadmap visibility, name on the site.
     Time cost: ~15 min/month.
  2. Operator — entered by: shipped a real outcome and is willing to show the numbers.
     Gets: unreleased Adveron intelligence for their category, direct line to the founders, co-authored teardown on our blog with their byline.
     Time cost: ~1 hour/month.
  3. Council — entered by: sustained Operator activity for a quarter, by invitation only.
     Gets: a vote on roadmap sequencing, a stage at our events, first access to the agent-experiment results before publication.
     Time cost: ~2 hours/month, quarterly call.
ACTIVATION LOOP
  trigger: user hits a real result in mos (goal met, experiment decided)
  → prompt: in-product nudge offering to turn the run into a shareable teardown, pre-drafted from their own data
  → act: they publish it with disclosure
  → reward: we amplify it from the founder account and credit them by name; the publish event advances their tier
  → repeat: next result re-enters the loop; dormancy after 90 days moves them down a tier quietly, no email.
INCENTIVES WE ARE DELIBERATELY NOT USING: discount codes, revenue share, merch as a primary reward. This audience trades on credibility; paying them for enthusiasm devalues the endorsement and creates disclosure problems.
MEASUREMENT AND KILL CRITERIA: track active ambassadors, referral signups, and advocacy mentions monthly. If after one quarter fewer than 8 people reach Operator, or referral signups stay under 5% of total, the loop is broken — fix the trigger before adding incentives.
Missing inputs: no share-of-voice data on named existing customers, and no category advocacy benchmarks (data gaps recorded). Tier thresholds are therefore judgment calls, not calibrated numbers.
{
  "offline": true,
  "tiers": 3
}

Ambassador playbook: profiles, scripts, cadence

ambassador_playbook
CANDIDATE PROFILES (in priority order)
  1. Growth & marketing leaders at seed–Series B startups — pain we solved: Market research is a two-week project they never start. Look for: has run cycles in mos, posts publicly, answers questions in Lenny's Newsletter + community.
  2. Agency operators & fractional CMOs — pain we solved: Junior output quality varies more than clients tolerate. Look for: has run cycles in mos, posts publicly, answers questions in X (agency + fractional CMO circles).
  Red flags: has never used the product beyond signup; asks about compensation before access; audience is other vendors rather than practitioners.
RECRUITING SCRIPTS (send unedited, one-to-one, from a founder)
  Script A — someone who posted about a result:
    "Saw your post about brand intelligence as an API — the part about what didn't work is the part I care about. We're putting together a small group of people who use mos for real and will say so honestly, including when it's wrong. You'd get the unreleased intelligence for your category and a direct line to us. No contract, no approval over what you write. Want in?"
  Script B — a quiet heavy user:
    "You've run more cycles than almost anyone and I've never heard you say a word about it publicly, which I respect. Standing offer: I'll turn any of your runs into a draft teardown using your own numbers, you edit or bin it. Interested?"
  Reference line available from perception data: ""mos is what I keep trying to build in Notion and failing at: goals, agents, and results in one loop." — fractional CMO"
ACTIVATION CADENCE
  Week 0: personal invite, access granted same day.
  Week 1: onboarding call — 20 minutes, we ask what they want to prove, not what they'll post.
  Week 2–3: first co-created artifact (teardown, thread, or talk), drafted by us from their data, published by them.
  Monthly: one intelligence drop for their category, one ask (max), one amplification of something they made.
  Quarterly: Council call — roadmap sequencing and what the agent experiments found.
  Dormancy: no activity in 90 days → quiet tier move, no reactivation email campaign.
Objective this serves: Design the ambassador program mechanics needed to reach 25 ambassadors: who qualifies, what they receive, the activation ladder from first yes to active advocate, the definition of "active", and the first outreach message. Buyers are growth and marketing leaders at startups and AI-forward brands. Anchor the work to the "Agents are doing the marketing work now" narrative where it fits honestly.
{
  "offline": true,
  "scripts": 2,
  "profileCount": 2
}
Reasoning notes

[offline draft — no ANTHROPIC_API_KEY] Circle ran its full Adveron gathering and data-gap probes, then produced template deliverables instead of Claude output. Intelligence used: Audience segments (advocate pool); Brand perception; Wishlist — share-of-voice and reach signals for named existing customers/community members, so advocates can be ranked rather than guessed at; Wishlist — advocacy program benchmarks in category: participation rate, incentive mix, referral yield, retention. Re-run with ANTHROPIC_API_KEY set for the real work product.