01-projects/agentify-3000

archetypes

2026-10-04·analysis·status: draft for founder review·source: data/businesses-scored.csv; scheme derived 2026-10-04 14:06 EDT from a 146-business stratified sample before classification (full rubric and tie-breakers in IMPLEMENTATION-NOTES)

Archetypes: how 2,566 [[2026-10-04-starter-story-pat-walls-assessment|Starter Story]] businesses split, and where the work is agent-shaped

Do not publish; internal analysis (Starter Story ToS: non-commercial reuse with link back).

Unit = business (3,261 stories deduped by website to 2,566). A fixed-prompt model assigned each an archetype, B2B/B2C, and a 0-3 score per core job for agent carry today (0 cannot, 1 assists, 2 does most, 3 runs it). Aggregates exclude 66 revenue-suspect businesses (2,500 remain; 2,052 with revenue). Revenue is self-reported.

The scheme

Per archetype (clean rows)

Archetype n Median $/mo Median $/mo per person Median people Median agent-carryable B2B share Most agent-shaped job Least
saas 513 42k 6.0k 7 58% 79% admin 2.3 find 2.0
dtc_brand 425 75k 13.0k 5 40% 1% admin 2.1 deliver 0.8
agency 357 46k 5.3k 8 48% 95% admin 2.2 deliver 1.5
media 210 8k 3.4k 2 55% 2% admin 2.3 support 1.6
consumer_app 200 19k 4.5k 3 60% 0% deliver 2.3 find 1.9
local 142 44k 7.6k 5.5 25% 11% admin 2.0 deliver 0.03
coaching 138 22k 8.2k 2 32% 48% admin 2.1 deliver 0.7
marketplace 131 80k 8.3k 8 50% 36% admin 2.2 deliver 1.5
edu 99 18k 7.5k 2 55% 12% admin 2.2 deliver 1.7
ops 85 200k 10.8k 14 30% 76% admin 2.0 deliver 0.4
retail 84 140k 18.0k 5 45% 6% admin 2.1 deliver 1.0
maker 79 6k 3.0k 2 30% 1% admin 2.1 deliver 0.1

Job scores are 0-3 means. "Other" (37) omitted. Overall median agent-carryable share 45%.

What the table says

  1. Admin is the universally agent-shaped job. Top job in 11 of the 12 real archetypes (consumer_app's top job is deliver, 2.3), never below 1.9. Bookkeeping, invoicing, scheduling, reporting go first.
  2. Deliver is the moat for anything physical. Local 0.03, maker 0.1, ops 0.4, coaching 0.7, dtc_brand 0.8. These five hold 869 businesses (35% of clean rows). Revenue per person is mixed: dtc_brand ($13.0k) and ops ($10.8k) rank second and third, maker ($3.0k) ranks last, and retail ($18.0k), outside the five, ranks first. For factory goods the money is where the agent is not.
  3. Software and content are already half-agent. consumer_app 60%, saas 58%, media and edu 55%. For saas and consumer_app the weak job is finding customers (1.9-2.0); for media it is support (1.6) and for edu it is deliver (1.7). Distribution is still bought with a face or a decade of SEO.
  4. Agencies are the odd one. 95% B2B, 48% carryable, deliver only 1.5: clients buy judgment. Lowest B2B revenue per person (5.3k): the opportunities thesis.
  5. Small teams dominate the agent-shaped archetypes (median 2-3 people). The overhang here is a solo-founder story.

One named example per archetype (monthly revenue)

Caveat: scores come from a 300-character public description per business (spot-check: 28 of 30 archetypes agreed, all 30 within one point on a 0-3 scale, judged by the producer, not an independent reader). The medians describe this dataset, not the economy. Bear case: Starter Story is a self-selected sample of founder-reported successes whose header figures can be peak, stale or wrong. The carryable share is a model grading its own capability from a blurb, and incumbents can adopt agents too.