BusinessRnD · draft v2 · built 2026-10-21 06:15 UTC

Round 3: AI-delivered services, sold by outreach

Tap a card for the details. On the decision cards, tap Approve, Not sure or No, then Send feedback.

Promising, not yet proven: a small catalogue of $250–1,000 services, delivered by agents and sold by agents' outreach.
At ~$600 a job, $10k/month is roughly 20 jobs. The one unknown no one has published: does a free, finished sample convert cold businesses? A 4-week test answers it.
Focus now
Video ads: the complete report is here. Angles, not just videos, for apps, games, gadget brands and crowdfunding, with real-company example pitches.AngleLab concept · Angle Sprint $390 · Angle Engine $490/mo · repriced after checking Ava Studio

Jobs and contacts needed for $10k/month

01020304050paid jobs per month (estimate)$250 job~50 · ~7.5–12.5k contacts$500 job~24 · ~3.6–6k contacts$750 job~16 · ~2.4–4k contacts$1,000 job~12 · ~1.8–3k contacts
ComfortableDoableGrind

Estimate: net per job after ~6% tax, ~4% payment fees and $10–20 compute, plus ~$300–600/month tools. Contacts assume 150–250 contacts per sale (round-3 outreach notes; unproven until the test). Monthly plans ($99–400) cut the jobs needed further.

Your four services (go-to-market in each tab)

Price check: what buyers can do instead

Service Cheapest credible alternative Does our price hold?
S1 App Store Fiverr gig ~$30–95; tools $9–180/mo ❌ Not alone. Door-opener plus bundle
S2 Accessibility Overlay widget $490–3,990/yr, which doesn't fix the code (1 in 4 sued sites had one) ✅ Best. Real code fixes plus legal fear
S3 Video ads Human UGC ~$99–167 per ad; agencies $2–10k/mo ✅ Holds. Video from real screens isn't a prompt
S5 Podcast clips DIY tools $14–65/mo; done-for-you $495 for 16 shorts ⚠️ Medium. Only for busy B2B teams

How it works

flowchart TD
  A["Agent detects a visible problem"] --> B["Agent makes a free sample fix"]
  B --> C["Email with the sample"]
  C --> D["Reply: pay link $250–1,000"]
  D --> E["Harness delivers"]
  E --> F["QA gate, 10–15 min human check"]
  F --> G["Delivered, 1 revision"]
  G --> H["Monthly plan upsell"]

The hard truths

Agents fail broad jobsBest agent: 21% acceptable on real paid projects (2026). Only narrow, machine-checkable services work. Gate: 70% ship-as-is before selling.
Sample-to-sale is unprovenNo public data on free-sample cold outreach. The test must measure it.
Where you can emailUS businesses and UK companies, with opt-out. Not Germany or Poland.
Compute and paymentsSubscription for building and QA; production on per-use non-Anthropic models (~$4–15/job). Payoneer pay links; read its dispute terms first.

Decisions for you

D15

Run the 4-week test for S1, S2, S3 and S5

I build each harness and prove 70% ship-as-is on 30 real targets. Then about 1,000 contacts per service from warmed inboxes (US and UK only). Each tab has its own kill thresholds.

D16

Approve per-use AI providers for production

Non-Anthropic models for the delivery steps, about $4–15 per job, paid from revenue. Your Claude subscription stays for building and quality checks.

D17

Approve outreach tooling, ~$300–600/month

Separate sending domains and inboxes, warmed for 2–3 weeks, US and UK targets only.

D18

Read Payoneer's card-dispute terms (only you can)

They're visible only inside your account. Needed before selling at volume. Fallback: invoices by wire.

D19

Keep blocked Google Play data out of S1

Your ASO tool's suggest, search and review calls hit addresses Google blocks for automated access. S1 uses only public Play app pages and Apple's official search API.

Your feedback

What you marked so far (all tabs)

Nothing yet.

Feedback is stored on this device and sent to the report's private storage. If sending fails, copy it and paste it into the chat.