Business plan — one package: website + POS + AI staff
Business plan — one package: website + POS + AI staff
Young, 2026-10-13: "Package 3 products into single package as sell services."
The product, in one line: a retailer gets a hosted online shop, in-store POS on their tablets, and an AI staff member that answers customers — one monthly price, we run all of it. The AI-staff part is what neither Medusa Cloud nor Shopify bundles.
This plan proposes; Young rules. Every price below is marked PROPOSED. Costs are measured (Tam's studies, live probes 2026-10-13); competitor prices are sourced (fetched 2026-10-13); margins follow by arithmetic.
1. What we sell (the package, from Tam's verified stack)
| Included | Built on | Licence |
|---|---|---|
| Hosted online store (Next.js storefront + Medusa backend) | Medusa v2.21.2, dtc-starter | MIT |
| In-store POS on client tablets (cart → draft orders, barcode via camera) | Agilo medusa-pos-starter (fork) | MIT |
| AI staff (storefront chat + support/sales, per-tenant memory) | EVE agent, self-hosted | Apache-2.0 |
| Backups, updates, TLS, domain hookup, admin onboarding | our runbooks | — |
Licence profile: clean end-to-end. MIT + Apache-2.0 means we can repackage, rebrand, and sell the bundle with no competing-use restriction and no per-seat metering. (Contrast: Saleor's Paper storefront is FSL — fine to run, not to resell as a product. The Medusa choice is what makes the package itself sellable.)
Known gaps, stated not smoothed: POS has no integrated payments yet (manual-payment path now; terminal roadmap later) and no offline mode. EVE is beta (API churn possible — pin versions in the tenant template).
2. Who buys it
Target customer: single-location AU retailers doing online + in-store who currently pay Shopify Basic ($29–39/mo) + POS Pro ($89/mo/location) = $118–128/mo before processing. Shopify ships Sidekick (a merchant-facing AI assistant that helps the owner sell) — but no customer-facing AI staff answering shoppers after hours. We sell the latter. Cafés, boutiques, trade counters, clinics selling products — anywhere one owner runs the till and the website.
The pain we solve: two bills (website + POS), two dashboards, no after-hours coverage. A customer who walks in at 9pm or messages the website at midnight gets no answer until morning. The AI receptionist closes that gap.
The alternative today: Shopify (closed ecosystem, $128/mo, no AI staff), Medusa Cloud (developer-focused, $359–849/mo real-world, no POS bundle), or Saleor Cloud ($1,599/mo, overkill for a single-location retailer). None bundles all three layers at a small-business price.
"Sell services" question OPEN: Young's brief says "sell services." Min's deep dive (§3.1) flags that services ≠ physical goods in the data model — a plumber selling "2-hour repair" needs booking/scheduling, not a shopping cart. Recommendation: pilot with a physical-goods retailer (simplest data model), then build the services catalogue module (custom Medusa module) as the differentiator for services businesses. This sequencing is assumed throughout the plan below.
3. Pricing PROPOSED
USD (competitors quote USD; convert at sale). Billed monthly, cancel anytime.
| Tier | Monthly | Gets |
|---|---|---|
| Solo | $79 | 1 location · website + checkout · POS on 2 tablets · AI receptionist (1,000 conversations/mo included) · backups |
| Shop | $149 | Solo + AI sales + support staff (3,000 conversations/mo) · staging environment · priority support |
| Extra location | +$49 each | same backend, second POS site |
How the tiers were set: Solo undercuts the Shopify equivalent ($118–128) by ~$40–50/mo while adding the customer-facing AI staff Shopify doesn't sell. Shop prices the staging env + support that Medusa Cloud sells as add-ons. Extra-location at $49 vs Shopify's $89.
Not in the price (pass-through, stated upfront): card processing (Stripe 2.9% + $0.30 online; ~2.6% + 10¢ in person) and POS hardware (tenant buys tablets/ readers). We take no cut of GMV — unlike Saleor Cloud's 0.8% overage. That is a selling point, not a gap.
AI overage PROPOSED: included conversations, then tenant supplies their own model key or pays metered overage at cost + 20%. Uncapped bundled AI would be an open tab — the same defect as an unmetered allowance.
4. Revenue model
Recurring monthly, tiered by tenant size. No setup fee at pilot (absorb onboarding cost to land reference customers). At commercial launch, a one-time setup fee of $199 PROPOSED covers:
- Domain + TLS hookup
- Admin user creation + onboarding session
- Storefront branding (logo, colours, hero copy)
- POS tablet setup + staff training (1 hour)
Why a setup fee at launch: it filters serious tenants, covers the ~2 hours of onboarding labour, and creates a small non-recurring revenue buffer. Waived for pilot tenants who agree to a case study / testimonial.
Revenue mix at steady state (10 tenants):
- 6 Solo × $79 = $474/mo
- 3 Shop × $149 = $447/mo
- 1 Shop + 2 extra locations = $149 + $98 = $247/mo
- Total: $1,168/mo recurring + occasional setup fees
5. Cost structure — real COGS per tenant
Infra per tenant (Tam, live-measured 2026-10-13): full stack ≈ 4.8 GB steady; Oracle Ampere A1 node (4 OCPUs, 23 Gi, 180 GB free) holds ~4 full or ~6 small stacks. That shape matches Oracle's always-free Ampere tier.
| Cost line | Per tenant/mo | Notes |
|---|---|---|
| Infra (node free-tier) | ≈$0 | Oracle ARM always-free; 4–6 tenants/node |
| Infra (node paid equiv) | ≈$10–13 | If free tier reclaimed; ~$50/mo node ÷ 4 tenants |
| Backup storage | ≈$1–2 | Per-tenant pg_dump → offsite (object storage) |
| Ops labour | ≈$15–25 | Onboarding amortised + support + updates (part-time, spread across tenants) |
| AI model calls | ≈$3–8 | Bundled conversations on free/long-context models; overage backstop |
| Total COGS/tenant | ≈$19–48 | Free-tier node: ~$19–35; paid node: ~$29–48 |
The margin is labour, not infra. Onboarding is scripted (compose project +
secrets + domain + predeploy + /health gate); support per tenant is the real
cost. At pilot scale (2–3 tenants) one person's part-time covers it.
Break-even:
| Scenario | Tenants | Revenue | COGS | Margin | Break-even? |
|---|---|---|---|---|---|
| Free-tier node, Solo only | 1 | $79 | ~$19 | ~$60 (76%) | ✅ Month 1 |
| Free-tier node, Solo only | 4 | $316 | ~$76 | ~$240 (76%) | ✅ Month 1 |
| Paid node (~$50/mo), Solo only | 1 | $79 | ~$29 | ~$50 (63%) | ✅ Month 1 |
| Paid node, Solo only | 4 | $316 | ~$116 + $50 node = $166 | ~$150 (47%) | ✅ Month 1 |
| Mixed tiers, paid node | 10 | $1,168 | ~$350 + $50 = $400 | ~$768 (66%) | ✅ Month 1 |
Break-even is immediate at any tier mix because the Oracle ARM free tier eliminates infra COGS. The paid-node case still clears 47–66% margin. The binding constraint is ops labour, not infrastructure.
6. Competitor bench (sourced 2026-10-13)
| Entry | Real-world prod | POS | AI staff | GMV cut | |
|---|---|---|---|---|---|
| Us (Solo) | $79 | $79 | included | included (customer-facing) | none |
| Shopify | $29–39 | ~$118–128 + $89/addl loc | +$89/location | merchant-facing only (Sidekick) | 2% 3rd-party |
| Medusa Cloud | $29 develop | ~$99–299+ (usage) | separate product | — | none stated |
| Saleor Cloud | $1,599 | $1,599 + 0.8% overage | via apps | — | 0.8% |
Our window: ~35% under Shopify's equivalent bundle, 1.3–4× under Medusa Cloud's real-world prod cost, with the one feature nobody else bundles (customer-facing AI staff + POS in one bill).
7. Go-to-market — first 3 pilot tenants
The goal: 3 paying pilot tenants within 60 days of the demo tenant going live.
Where they come from:
- Young's network — the first pilot tenant should be someone Young knows who runs a small retail business. Warm intro, zero sales friction.
- Local AU retail communities — Facebook groups for small business owners, local chambers of commerce, "Shop Local" networks. The pitch: "Your website, POS, and AI receptionist in one $79/mo bill — cheaper than Shopify alone."
- Medusa community — Medusa Discord (14k members), Medusa GitHub discussions, r/medusajs. Developers who know Medusa but can't get POS + AI staff from Medusa Cloud. The pitch: "What Medusa Cloud doesn't bundle."
The pilot offer: Solo tier at $49/mo for 3 months PROPOSED (33% discount) in exchange for:
- A case study / testimonial (written by us, approved by them)
- Permission to use their name + logo on the product site
- A 30-min feedback call at the end of month 1
Why $49: it's below the psychological $50/mo barrier, still cash-positive (COGS ~$19–25/tenant), and the discount is temporary — pilot tenants convert to $79/mo at month 4 or leave.
Pilot success criteria:
- All 3 tenants live for 30 consecutive days (health gates green)
- Zero unplanned outages > 1 hour
- At least 1 tenant uses the AI receptionist for real customer conversations
- At least 2 tenants renew at full $79/mo after the pilot discount ends
8. Risk register
| Risk | Cost if it fires | Likelihood | Mitigation in this plan |
|---|---|---|---|
| EVE beta churn breaks tenant agents | rebuild per tenant | Medium | version-pin @workflow/*; adopt for tenants first, fleet later |
| POS payments land late | manual path looks thin at sale | High | disclose at sale; price reflects it ($79, not $149) |
| Oracle free-tier reclaimed | infra jumps to ~$50/mo | Low–Med | still margin-positive at 2+ tenants; per-tenant stacks evacuate to paid node |
| AI overage abuse | unbounded model bill | Low | bundled cap + tenant-key-or-overage rule |
| One node = one failure domain | pilot outage | Medium | staging env on Shop tier; node 2 at pilot exit |
| "Sell services" module scope creep | pilot delayed | Medium | pilot with physical goods first; services module = Phase 2 |
| Tenant data loss (PG volume) | total loss for that tenant | Low | per-tenant pg_dump cron → offsite; tested restore |
| EVE agent gives wrong answer to customer | tenant reputation damage | Medium | HIL gates on refunds/bulk pricing; read-only storefront tools |
9. Open decisions for Young
- Tier prices ($79 / $149 / +$49) — approve or reset.
- Setup fee ($199 at commercial launch, waived for pilots) — approve or reset.
- Pilot discount ($49/mo for 3 months, 3 tenants) — approve or reset.
- AI overage model — tenant-key-or-metered, as proposed?
- First pilot tenant — name the shop and this moves from plan to build.
- AUD vs USD billing — proposal is USD; AU retailers may expect AUD.
- "Sell services" pilot vertical — physical goods first (recommended), or services first?
Sources & chain
- Product stack:
01-research/medusa-eve-integration-study.md(Tam, all claims live-verified 2026-10-13) - Sizing/capacity:
01-research/medusa-hosting-docker-study.md(node measured live) - Co-hosting/RAM:
01-research/hermes-eve-cohosting-study.md - Stack comparison/licences:
01-research/medusa-nextjs-commerce-study.md - Packaging mechanics:
01-research/unified-package-deepdive.md(Tam) - Gaps/risks/open questions:
01-research/3-package-deep-dive.md(Min + Tam) - TCO method + Stripe math:
01-research/saleor-commercial-review.md(Mira) - Competitor prices: medusajs.com/pricing + shopify.com/pricing (live-probed 2026-10-13 by Mira: Shopify Basic $29–39 + POS Pro $89/loc; Medusa tiers $29/$99/$299 + usage; prior draft's "$359–849" corrected) + Stripe standard 2.9% + $0.30
- Node specs re-verified live 2026-10-13 by Mira (ssh PROD): 4× aarch64 Neoverse-N1, 23Gi RAM (22 avail), 180G disk free — matches Tam's measurements
- Build sequencing: Tam's note §"Build sequencing" → Ace lane; design spec: Sol
- Architecture:
02-architecture/SOLOS-V3-PRODUCTION.md(Sol's production doc — tenant template, onboarding script, health gates)
Yukon IT admin backend — Option 1 vs Option 2, priced
Young, 2026-10-17: Yukon IT (public website selling services) needs a backend admin, EVE reception, storefront + POS, landing + domain + O365.
Business plan v2 — one package, AUD, honest infra cost
Young's three rulings, 2026-10-14: Oracle ARM shape confirmed as ours (4vCPU/24GB/200GB) · POS gaps + EVE beta are build work (Sol specs, Min+Tam Docker design) · prices in AUD.