Internal Document
Aronlight v4 — A2 Build Strategy + Cost Model (v4 - April 2026)
For Giuseppe + Raffaello only. Companion to Aronlight_AI_Proposal_v4.html
Product
What We Are Building
One AI-powered workflow module connected to Aronlight's Odoo ERP. A2 (Quote Generation) is the sole focus for now. A5 (RMA) is deferred until A2 is live and validated - see the LATER section at the bottom of this document.
Current problem
Back-office / commercial team handles ~3,000 proposals/month manually (~120 requests/day): distributor emails back-office → back-office looks up product in Odoo → copies pricing → formats reply → sends. No automation, no review queue.
✅ CONFIRMED — Apr 8 Requests arrive 90% by email, some WhatsApp/phone, directly to back-office. 2,000 active SKUs. ~20 items/proposal. €4K–€19K average order range. Odoo is the working catalogue and pricing system (90% accurate — exchange rate issues for USD imports require manual overrides). Confirmed goal: Odoo as single source of truth.
❓ STILL OPEN Are inbound requests structured product codes or free-text? Structured = simpler parsing. Free-text = NLP extraction layer needed, adds complexity and cost. Confirm in scope call.
What it does
Receives distributor product requests via a messaging channel, matches to Odoo catalogue, generates and sends a formatted proposal automatically. Complex cases route to back-office with a pre-filled draft.
Input channel
TBD with Manel — WhatsApp, Telegram, or another channel they already use. We are channel-agnostic. For Phase 1 fallback: back-office can paste requests manually into a simple form (no channel integration needed to demo the core logic).
Output
Auto-sent proposal OR pre-filled draft for back-office review
Odoo connector
✅ CONFIRMED — Apr 8 Odoo is live and stable. Staging environment available (separate DB from production). A third-party integration company manages the instance and holds the code — API access requires their authorization. Manel is negotiating access. Phase 1 path: start with weekly CSV export (safe — catalogue changes every 1.5–2 years, new products added monthly). Upgrade to full REST connector when API access is granted. See Architecture Decision section.
Complexity
Medium (6–8 weeks production)
Repeatability
MEDIUM — meaning we see this same problem in 2–3 other clients in our current pipeline (MAGG: 3,000+ distributors; Eurostar: likely receives quote requests; AutoalDia: adjacent WhatsApp dynamic). Once the core architecture is built for Aronlight, the catalogue lookup engine and proposal template engine are reusable across future clients with minimal rework. This is why we can price it below a dev shop — we amortize the build.
Full product specs
To be built after scope confirmed. See module-registry.md > A2 for current definition.
Product specs gap: We have a high-level definition in the module registry but no full product spec document for A2 yet. This should be built once scope is confirmed in the next call with Manel. Assign to: Giuseppe (A2).
Architecture Decision
Odoo Connector — Must-Have or Workaround?
Three options exist for connecting to Odoo. The key decision: MVP does NOT depend on Odoo API negotiation. This de-risks the timeline completely.
MVP (v1)
Option A - CSV/Excel Export from Odoo (no API)
Manel exports catalogue + pricing from Odoo as a CSV/Excel weekly. Our system reads the file. No API needed. Email ingestion + proposal generation works against the static catalogue.
Gets to demo in: 4 weeks. No dependency on third-party integration company or API authorization.
Catalogue changes every 1.5-2 years, new products added monthly. Weekly CSV sync is safe - no risk of stale proposals at this update frequency. This is the v1 path. Period.
v2 Upgrade
Option B or C - Odoo REST API (after MVP validated)
Once MVP is running and Manel has closed the API authorization with the third-party integration company, upgrade to live Odoo REST connector. Real-time catalogue, pricing. Also required foundation for A5 (RMA) when that activates later.
Timeline: 4-6 weeks after MVP, when API access is granted. Not blocking MVP delivery.
Requires Odoo API documentation and IT cooperation. Manel is negotiating access. We proceed with CSV until access is granted - zero hard dependency on their timeline for MVP.
Decision (final): MVP uses Option A (CSV/Excel export). No API negotiation dependency. This de-risks the entire timeline. v2 adds the REST connector once Manel closes the third-party authorization. The catalogue update frequency (1.5-2 year cycles, monthly additions) makes weekly CSV perfectly safe for production use.
Open questions that determine the connector decision
✅ Apr 8
Catalogue changes every 1.5–2 years. New products added monthly. Weekly CSV export is safe for Phase 1 — stable enough that a once-a-week sync won't generate stale proposals.
Decision: proceed with Option 1 (CSV export) for A2 Phase 1 MVP. Move to REST connector when integration company grants API access.
✅ Apr 8
Odoo is fully live. Staging environment available — separate DB from production. We can build and test without touching live data.
Path: build on staging, integration company deploys to prod. No risk of disrupting live operations during build.
⚠️ Partial
A third-party integration company manages Odoo and holds the code. Aronlight pays a monthly retainer. Any API access or code change requires their authorization. Manel is actively negotiating access.
Blocker for moving from CSV to REST connector. Action owner: Manel. We proceed with CSV until access is granted — no hard dependency on their timeline for Phase 1.
Factor 1
Cost of Building and Maintaining
New cost model. Partners are "vibecoders" using Claude Code (Opus) - they direct AI to build. 1 hour of partner vibecoding = what used to take 3-4 dev hours traditionally. Claude Code does 65-70% of the actual coding. Dev talent (LatAm) at $40/hr handles deployment, infra, code review, and edge cases. Partner time ($200/hr combined) is the expensive input - minimize it by having Claude Code do the heavy lifting.
Rates: Partners (Giuseppe + Raffaello combined vibecoding) = $200/h | Dev (LatAm) = $40/h | Claude Code = 2× Max $200/mo + API overflow during sprints
MVP Build Cost (4 weeks)
| Role | Hours | Rate | Cost |
|---|
| Partner vibecoding (with Claude Code) | 25h | $200/h | $5,000 |
| Dev (LatAm) | 35h | $40/h | $1,400 |
| Claude Code (Max $200/mo × 50% + API overflow) | - | - | ~$600 |
| Infrastructure setup | - | - | ~$200 |
| Total MVP build | | | ~$7,200 |
MVP Weekly Breakdown (4 weeks)
Week 1
Email ingestion + parser prototype + catalogue import
Partner 8h, Dev 8h
Week 2
Pricing logic + proposal generator
Partner 6h, Dev 8h
Week 3
Review queue + dashboard
Partner 6h, Dev 10h
Week 4
Test with real emails + deploy
Partner 5h, Dev 9h
v2 Build Cost (4-6 weeks after MVP)
| Role | Hours | Rate | Cost |
|---|
| Partner vibecoding | 20h | $200/h | $4,000 |
| Dev (Odoo experience needed) | 40h | $40/h | $1,600 |
| Claude Code (Max + API overflow) | - | - | ~$500 |
| Total v2 build | | | ~$6,100 |
v2 dev note: v2 needs a dev with Odoo REST API experience. This might be $40-60/hr depending on availability. Budget the upper range until sourced.
v3 Build Cost (4 weeks after v2)
| Role | Hours | Rate | Cost |
|---|
| Partner vibecoding | 15h | $200/h | $3,000 |
| Dev | 30h | $40/h | $1,200 |
| Claude Code (Max + API overflow) | - | - | ~$450 |
| Total v3 build | | | ~$4,650 |
MVP build (4 weeks)~$7,200
v2 build (4-6 weeks after MVP)~$6,100
v3 build (4 weeks after v2)~$4,650
Total build cost (MVP through v3)~$17,950
Old estimate (A2+A5 combined)€26,800-€33,400
New A2-only total~$17,950 - still roughly half the old estimate
Monthly Running Cost (post-MVP)
| Cost item | Monthly | Notes |
|---|
| LLM API (3K proposals x ~$0.05-0.09/req) | ~$150-270 | Drops as requests become more structured over time |
| Infrastructure (hosting, email service) | ~$100-150 | Fly.io or Railway |
| Dev maintenance (5-8h/month) | ~$200-320 | Bug fixes, catalogue sync, minor adjustments |
| Partner check-in (1h/month) | $200 | Monthly review with Manel |
| Total monthly | ~$650-940 | |
Margin Analysis at Different Price Points
| Monthly price to client | Monthly cost | Monthly margin | Margin % |
|---|
| €1,500/mo | ~$850 (~€780) | €720/mo | 48% |
| €2,000/mo | ~$850 (~€780) | €1,220/mo | 61% |
| €2,500/mo | ~$850 (~€780) | €1,720/mo | 69% |
Key insight: The vibecoding model dramatically reduces build cost. Partner time is the expensive input, but Claude Code (2× Max $200/mo + API overflow) multiplies each partner hour by 3-4x in output. Dev at $40/hr handles the residual work. Total MVP build at ~$7,200 is less than half the old Phase 1 estimate alone. Monthly running costs of $650-940 leave healthy margins at any price point above €1,500/mo.
Team + Tooling
Who Builds This — Team Profile + Tools
Likely dev team composition
| Role | Count | Profile | Used for |
|---|
| Backend dev | 1 | Python (primary). FastAPI or Flask for API layer. REST API integration experience (Odoo a plus). LatAm or Eastern Europe for cost efficiency. | Odoo connector, message parser, catalogue lookup, quote engine, RMA workflow logic |
| Frontend dev | 1 | React or Vue. Can be the same person as backend if fullstack. Simple UI — back-office review queue and director dashboard are not complex. | Back-office review queue UI, director dashboard, customer-facing RMA intake form |
| Partner (Raffaello) | 0.5 | Scoping, client-side QA, A5 domain expertise (prior RMA experience). Not writing code. | A5 scoping call, back-office workflow design, testing against real cases |
| Partner (Giuseppe) | 0.5 | Client management, A2 architecture decisions, proposal templates, delivery coordination. | A2 logic design, channel integration decisions, Manel communication |
Could one fullstack dev handle both? Probably yes for the optimistic scenario. For the realistic scenario (more back-and-forth on Odoo connector, more UI iterations), two devs in parallel would compress the timeline significantly. Recommend: one primary backend dev + one frontend dev, both part-time if budget is tight.
Expected tools + stack
| Tool / Layer | What it does | Impact on time/cost |
|---|
| Python | Backend logic, Odoo API calls, message parsing, quote generation | Standard. Most LatAm devs know it. |
| Odoo REST API (or CSV fallback) | Catalogue data, pricing, stock levels, write-back for A5 | If API: 2–3 weeks to build. If CSV: 3–5 days. See connector decision section. |
| Claude API | NLP for parsing free-text product requests (e.g. "I need the 50W warm-white recessed light for outdoor"). Extracts intent and maps to SKUs. Also useful for summarising RMA case context. | Reduces dev time significantly. Without Claude: custom NLP logic = 2–3 extra weeks. With Claude: parsing layer = 2–3 days. Partners spend less time on edge-case logic. Recommend using Claude API for the message parser from day 1. |
| WhatsApp Business API (or Telegram / other) | Receiving and sending messages. Channel integration layer. | Adds 3–7 days depending on which channel and whether API access is already approved. WhatsApp Business API requires Meta approval (~1–2 weeks if not already active). |
| React / Vue (frontend) | Back-office review queue, director dashboard, RMA customer form | Standard. Could use a lightweight admin template to save 1–2 weeks of frontend work. |
| PostgreSQL / Supabase | Store cases, proposals, logs, audit trail | Simple. A few tables. Not a complexity driver. |
Claude API = lever on partner + dev time. Using Claude API for the message parsing layer (free-text → SKU extraction) removes the hardest NLP problem from our dev scope. This reduces both partner hours (less edge-case design) and dev hours (less custom logic). Rough estimate: saves 10–15 partner hours and 20–30 dev hours. At our rates: ~€2,800–4,300 in cost reduction, or equivalent as timeline compression. We should price this in as an API cost (~€45–175/mo depending on volume) and flag it to Manel as part of the infrastructure.
Factor 2
Client Value
A2 — Hidden labor cost (120 requests/day)
| Variable | Estimate | Source |
|---|
| Routine proposals/week | 600 | From discovery: 700 total, ~600 routine catalogue items. From v2 proposal + pain section. |
| Requests/day (catalogue) | 120/day | 600/week ÷ 5 working days. |
| Time per proposal | 15 min | Look up in Odoo, format, send. Catalogue-only = straightforward. |
| Total hours/day | 30h/day | 120 × 15min / 60. Requires 3–4 back-office staff full time. |
| Total hours/month | ~600h/mo | 30h/day × 20 working days |
| Back-office salary (Portugal) | ~€2,500/mo | Including employer taxes + benefits. Gross salary alone is ~€1,800–2,000 but total employer cost is ~€2,500. Verify with Raffaello. |
| Hourly cost | €15.60/h | €2,500 ÷ 160h |
| Hidden monthly labor cost | ~€9,375/mo | 600h × €15.60 |
| Hidden annual labor cost | ~€112,500/yr | ~€107,000/yr (conversion note: original €112,500 EUR base). Implied: ~3–4 FTEs at 100% dedication, OR 6–8 people spending ~50% of their time on proposals. ❓ Unconfirmed — ask Manel how many people generate proposals and roughly what % of their day it takes. |
ROI math (updated with single-price model). At €107K/yr hidden cost and €2,000/mo (€24K/yr): ROI = (€107K − €24K) / €24K = 3.5x. Comfortably above the 1.5x target. Annual at €1,500/mo (€18K/yr, 25% discount): 4.9x. Both pass easily. And this is only the labor cost side — revenue upside from faster proposals (avg order value unknown) could push this higher.
Unconfirmed — revenue upside (ask Manel). Faster proposals = higher distributor conversion. Even a 3% uplift on 120 requests/day at an unknown average order value could be worth far more than labor savings. We need avg order value to calculate this. See "Questions to Ask Manel" section.
A5 — RMA value (DEFERRED — expansion opportunity)
A5 is parked. This data is preserved for the expansion pitch after A2 MVP is live.
| Value driver | Estimate | Note |
|---|
| Direct labor (RMA handling) | ~€2,600/yr | ~10 cases/mo × 90min × $15.60/h × 12. Low on its own — the urgency story is stronger than the volume story. |
| Business continuity premium | €12,900–17,200/yr | Replaces 0.25 FTE that just left. They need a system NOW regardless of volume. |
| Carrier claim recovery | ~€3,400/yr | If 20% of cases are carrier responsibility at avg €200 claim × 10 cases/mo × 12. Currently untracked. |
| Distributor relationship risk | €8,600–43,000/yr | 1–2 distributor relationships at risk if returns are mishandled. LTV of a mid-size distributor likely €10K–50K. Unquantifiable without data. |
A2 hidden labor cost (incl. employer taxes)~€107,000/yr
ROI at €2,000/mo (€24K/yr)~4.5x — passes 1.5x target
Annual at €1,500/mo (€18K/yr)~5.9x — still well above target
If revenue upside confirmedPotentially 5x+ (depends on avg order value)
A5 expansion value (later)~€19,000–23,300/yr — use for upsell pitch post-MVP
Internal — AI Economics
What Does It Cost Us to Run This?
Monthly support cost covers human time only: ~2h/mo partner oversight ($200/hr) + ~4h/mo dev maintenance ($40/hr) = ~$560/mo (~€520). AI processing and infrastructure are additional. This section gives the full picture.
AI costs now incorporated: All P&L tables below use ~€650–940/mo as total ongoing cost (support ~$560 + AI ~$90–330 + infra ~$100). Pricing: €2,000/mo (monthly) or €1,500/mo (annual, 25% discount). Implementation fee: €5,000 one-time.
LLM Processing Cost — Per Request
Model: Claude Sonnet 4.6 ($3/$15 per MTok input/output). Prices in USD. Estimates based on typical call counts per request type.
| Request type | LLM calls | Approx tokens | Cost/request | Notes |
|---|
| A2 — Routine quote | 4–5 | ~9–11K | ~$0.054 | Parse request + catalogue lookup + format response + sanity check. Fully automatic, no review queue. |
| A2 — Complex quote | 8–12 | ~30–50K | ~$0.24 | Ambiguous product refs, multi-SKU bundles, or edge cases routed to review queue with pre-filled draft. Estimated 20–30% of requests. |
| A2 — Blended average | — | — | ~$0.11/quote | 70% routine × $0.054 + 30% complex × $0.24. Conservative blend. Improves as the model learns Aronlight's catalogue patterns over time. |
| A5 — RMA case (DEFERRED) | 3–4 | ~8K | ~$0.048 | Deferred. Preserved for expansion pricing. |
Infrastructure — Monthly Fixed Costs
| Service | Monthly | What it covers |
|---|
| Hosting (Fly.io or Railway) | ~$40–60 | API server + async workers. Scales automatically with volume spikes. |
| Database (Supabase) | ~$25 | Request log, approval states, Odoo sync cache. |
| Cloudflare Workers / Pages | ~$5 | Webhook endpoints, edge routing. Negligible at this scale. |
| Monitoring + logs (BetterStack or Sentry) | ~$20–25 | Error alerts, uptime checks, performance dashboards. Required for SLA compliance. |
| Infrastructure total | ~$90–115/mo | ~€83–106 at current exchange rate. Stable regardless of request volume within the cap. |
Full Monthly Cost — Three Volume Scenarios (A2 only)
Confirmed volume: 3,000 quotes/month (A2). A5 RMA deferred. Support = 2h partner ($200/hr) + 4h dev ($40/hr) = $560/mo.
| Scenario | A2 LLM cost | Infra | Support (human) | Total monthly |
|---|
| Confirmed volume (3,000 quotes) | ~$330 | ~$100 | $560 | ~$990 (~€920/mo) |
| Volume cap (3,600 quotes) | ~$396 | ~$100 | $560 | ~$1,056 (~€980/mo) |
| Spike scenario (+50% above cap) | ~$495 | ~$115 | $560 | ~$1,170 (~€1,080/mo) — overage revenue covers it |
Note: These detailed per-request LLM costs ($0.11 blended) are higher than the simplified ~$150–270 range in the "Cost of Building" section. That section estimates MVP-level processing (fewer LLM calls per quote). As the system matures to v2/v3, per-request costs may increase with complexity. Use this table as the conservative ceiling.
A2 LLM detail (confirmed volume): 3,000 quotes/month × $0.11 blended = $330 total. But 70% routine ($0.054) = $113 and 30% complex ($0.24) = $216. If Manel's catalogue is more structured than average, complex rate drops to 15-20% and LLM cost drops to ~$200/month. We should instrument this from day 1 — track routine vs. complex ratio and surface it in the monthly performance report.
Volume Caps + Overage Pricing
The monthly subscription covers processing up to the cap. Requests above the cap are billed at a per-unit overage rate. This protects our margin if usage spikes above what we priced for.
| Module | Confirmed volume | Cap included in subscription | Overage rate | Basis |
|---|
| A2 — Quote Generation | ~3,000/mo | 3,600/mo | €0.30/quote above cap | 3,000 × 1.2 = 3,600. Buffer absorbs normal growth + seasonal peaks without triggering overage. |
How to present this to Manel: "The monthly fee covers everything up to 3,600 proposals per month — that's 20% above your current volume. If you grow beyond that, there's a small per-unit fee so the system scales with you. You only pay more when you get more." Do not mention the word "cap" or "overage" proactively. Only explain if he asks about what happens at high growth.
Margin at Monthly vs Annual Billing — Floor Check
Annual billing = 25% discount off monthly rate, paid as lump sum at Mes 3. Our cost structure (~€780/mo) does not change with billing frequency — the discount comes entirely out of margin.
| Price point + billing | Monthly revenue | Monthly cost | Monthly margin | Margin % |
|---|
| €2,000/mo — monthly | €2,000 | ~€780 | €1,220 | 61% |
| €2,000/mo — annual (€1,500/mo) | €1,500 | ~€780 | €720 | 48% |
Floor check: Annual at €1,500/mo = €18K/yr (above €10K floor), 48% margin. This is the walk-away floor. Lead with €2,000/mo monthly (61%). Offer annual (25% off = €1,500/mo, €18K lump sum) only if Manel pushes for a commitment discount.
Business Justifications for the Monthly Fee
These arguments are written for internal alignment but are safe to surface in the external proposal or verbally with Manel if he pushes back on the monthly cost.
Argument 1 — Headcount, not software
The back-office team processes ~700 proposals/week. At €2,500/mo all-in per person, a single full-time equivalent costs €30,000/year. The system replaces the most repetitive part of 2–3 people's jobs. At €2,000/mo, Aronlight pays less than a junior employee — and the system runs 24/7 without sick days or training overhead. The monthly fee is the cost of keeping those people doing higher-value work instead of copy-paste proposals.
Argument 2 — The system keeps learning your business
Every month, the system gets better: it learns Aronlight's catalogue edge cases, improves product matching accuracy, and adapts to any Odoo catalogue changes. Cancelling isn't just turning off a tool — it means losing accumulated context and starting from zero. The monthly fee covers the monthly tuning cycle, compatibility monitoring, and performance reviews that make the system improve over time. A one-time build without ongoing service degrades within 3–6 months as the catalogue evolves.
Argument 3 — Volume is included; you only pay more when you grow
The subscription covers 3,600 proposals/month — 20% above current confirmed volume. Most months, Aronlight pays a flat fee regardless of actual usage. If volume grows above the cap, there's a small per-unit rate (<€0.30/proposal) that scales the cost with the benefit. Compare this to a dev agency retainer ($10,000–15,000/month) or an internal hire — both charge a fixed cost with no scaling benefit.
Our P&L
When Do We Break Even?
Billing model (V3, Apr 2026): Months 1-2 = implementation (consulting, development, testing) · Mes 3+ = monthly subscription. Phased build: MVP first, then v2 and v3 funded from subscription revenue.
Impl fee sizing: MVP build cost is ~€6,650 ($7,200). At €5,000 impl fee we absorb ~€1,650 as design partner trade. Impl fee as % of annual: €5,000 / €24K = 21% (within 20–30% Medium band).
Founder draw: Not included in P&L below. At €2,000/mo sub with €1,220/mo net margin: a combined draw of €1,200/mo (€600 each) consumes nearly all margin. Second client is the real unlock for compensation.
Scenario A — MVP Only (close fast, validate)
Aligned with First Workflow Rule: close the MVP, deliver it, then expand.
| Period | Cash in | Cost out | Cumulative | Margin |
|---|
| Mes 1-2 (build + test) | +€5,000 impl | €7,150 | −€2,150 | Design partner gap |
| Mes 3+ (subscription at €2,000/mo) | +€2,000/mo | ~€780/mo | +€1,220/mo net | 61% |
| Breakeven | €2,150 ÷ €1,220 = 1.8 months from Mes 3 | Month 5 from signing |
Scenario B — Full Product (MVP → v2 → v3)
v2 and v3 builds happen while subscription is running. Creates temporary negative cash flow months 3–6.
| Period | Cash in | Cost out | Cumulative | Note |
|---|
| Mes 1-2 (build + test) | +€5,000 impl | €7,150 | −€2,150 | Design partner gap |
| Mes 3–4 (sub + v2 build) | +€2,000/mo | €780 + ~€2,830/mo (v2 build) | −€5,370 | v2 investment period |
| Mes 5–6 (sub + v3 build) | +€2,000/mo | €780 + ~€2,160/mo (v3 build) | −€7,250 | v3 investment period |
| Mes 7+ (sub only, full product) | +€2,000/mo | ~€780/mo | +€1,220/mo net | 61% margin |
| Breakeven | €7,250 ÷ €1,220 = ~5.9 months from Mes 7 | Month 13 from signing |
Year 1 net (MVP-only at €2,000/mo, no draw)
+€9,550
€5,000 impl + 10mo × €1,220 net − €7,150 build+test = +€10,050. With €1,200/mo draw: −€1,950.
Year 1 net (full product at €2,000/mo, no draw)
−€1,050
€5,000 impl + 10mo sub − full build (~€16,600) − running costs. Slightly negative. With draw: deep negative. Full product needs a second client to fund.
Year 2 net (full product, no draw)
+€14,640
12 × €1,220/mo net. All builds amortized. With €1,200/mo draw: +€240/mo real profit. Second client at same stack = near-zero extra build cost.
Bottom line: At €5,000 impl we absorb ~€2,150 as design partner trade. Breakeven is Month 5 for MVP-only, Month 13 for full product. Strategy: close the MVP fast at €2,000/mo, prove value, then upsell v2/v3. Annual commitment (€1,500/mo, €18K lump) is the concession if Manel pushes back.
Factor 3
Replicability — Do Other Clients Have This Problem?
A2 — Quote Generation: other clients in our pipeline
| Client | Similar pain? | Stage |
|---|
| Aronlight | Confirmed — WhatsApp quote bot, 700 proposals/week | Active — this proposal |
| MAGG | Likely — 3,000+ distributors, must receive quote requests. Not yet confirmed on call. | Proposal sent |
| AutoalDia | Adjacent — A4 (WhatsApp lead capture) is not quote gen but same channel + same back-office dynamic | In progress |
| Lamosa | Different — A1 is about discount approval, not quote generation. Different workflow. | In progress |
| Eurostar | Possible — machinery distributor, likely gets quote requests. Not explored yet. | Pre-prop sent |
Mercura (YC W25, $2.1M raised) — are they building A2?
Yes, partially. Mercura automates inbound B2B quote requests (RFQ — Request for Quotation) for distributors, integrated with ERPs. This is the same workflow as A2 but from a SaaS product angle. Their raise validates the market.
How we differ: (1) We integrate directly with Aronlight's Odoo instance — Mercura is a standalone product that may or may not connect cleanly. (2) We add RMA (A5) which Mercura does not offer. (3) We are boutique — we configure for their exact workflow, not a generic template. (4) We are cheaper for a single company than a Mercura SaaS subscription at €26,000–86,000/yr.
Risk: If Manel finds Mercura before signing, he may compare. Our positioning: same outcome, better fit, plus the RMA gap which Mercura does not solve.
A5 — Returns Merchandise Authorization (RMA): other clients
| Client | Similar pain? | Stage |
|---|
| Aronlight | Confirmed — urgent, dedicated person just left | Active — this proposal |
| Eurostar | Adjacent — B4 (After-Sales Case Management) is similar but focused on machine repair cases, not product returns. Could evolve into RMA. | Pre-prop sent |
| MAGG | ❓ Not yet checked. 1,500+ SKUs, 3,000+ distributors — almost certainly has product returns. Nobody has asked. Raffaello: add to next MAGG call agenda. | Proposal sent |
| Lamosa / AutoalDia | Unlikely — different product/channel profile. Not a priority. | — |
Action for Raffaello — MAGG: Check whether A2 (quote automation) and A5 (RMA) are applicable to MAGG before their next call. They have 3,000+ distributors, a large catalogue, and a proposal already sent. If they have the same quote generation and returns problems, we should be building toward a reusable architecture that works for both Aronlight and MAGG from the start — not building Aronlight-specific code we'll have to rewrite.
A2 replicability ratingMEDIUM — 1 confirmed, 2–3 pipeline candidates, Mercura validates market
A5 replicability ratingMEDIUM — 1 confirmed (urgent), 2 pipeline adjacents, universal workflow
TogetherPotential "Distributor Ops Suite" product for mid-market distributors LatAm + EMEA
Outputs — Updated Apr 11, 2026
A2 Phased Pricing — MVP First
Single workflow (A2 Quote Generation), phased delivery. Billing follows Mes 1 / Mes 2 / Mes 3 model. MVP goes live first; v2 and v3 are subscription-funded expansions.
Billing structure: Months 1-2 cover implementation (consulting, development, testing). Mes 3+ = monthly subscription. Client message: "Months 1 and 2 are for the build. You only pay the monthly fee once it's running in production." v2/v3 upgrades included in the subscription, plus updates to technology, models, or infrastructure.
Pricing (as sent in v5)
Quote Automation — Phased
A2 full product (MVP → v2 → v3)
Month 1 — impl fee
€5,000
Month 2
€0
Month 3+ monthly
€2,000/mo
61% margin at monthly. Annual: €1,500/mo (25% discount, €18K lump sum).
Strategy note: Single price in v5: €5,000 impl + €2,000/mo. Annual option at 25% discount (€1,500/mo, €18K lump sum at Month 3). If Manel pushes back on monthly, offer annual as the concession. Walk-away floor: €5,000 impl + €1,500/mo annual. Proposal expires April 17.
Build Cost vs Revenue (vibecoding model)
Partner vibecoding ($200/hr) + LatAm dev ($40/hr) + Claude Code API. Details in "Cost of Building and Maintaining" section.
| Monthly | Annual (25% off) |
|---|
| Impl fee | €5,000 | €5,000 |
| MVP build cost | ~€6,650 | ~€6,650 |
| Impl coverage | 75% | 75% |
| Subscription yr 1 (10 mo) | €20,000 | €18,000 |
| Running cost yr 1 (10 mo) | €7,800 | €7,800 |
| Year 1 total revenue | €25,000 | €23,000 |
| Year 1 net (MVP-only build) | +€10,550 | +€8,550 |
| Year 1 net (full product build) | +€1,050 | -€950 |
Floor: €5,000 impl + €1,500/mo annual is the absolute floor (€18K/yr, 48% margin at monthly equivalent). Impl fee at €5,000 covers 75% of MVP build cost — we absorb ~€1,650. Acceptable as design partner trade. Impl fee as % of annual: €5,000 / €24,000 = 21% (within 20–30% Medium band).
Sales Talking Points
Why MVP First — Arguments for Manel
Aligned with First Workflow Rule. Close the MVP, deliver it, prove value, then expand to v2/v3 and later A5. These talking points are for when Manel asks "why not build the full product from day one?" or "can't you just do everything at once?"
Five reasons why starting with the MVP is better for Aronlight:
| # | Reason | Why it matters |
|---|
| 1 |
Live in 4 weeks, not 4 months |
The MVP handles catalogue quotes from email using a CSV export. No Odoo API negotiation, no IT involvement. The back-office starts seeing value in one month instead of waiting for a full integration. |
| 2 |
Real data before big investment |
The MVP generates real accuracy metrics on day one. Instead of committing €24K+ on faith, Manel sees the system process actual proposals before deciding on v2 (Odoo integration) and v3 (auto-send, WhatsApp). |
| 3 |
Team adoption without shock |
The back-office reviews AI-generated drafts in a simple queue. No workflow change beyond "check the draft before sending." v2 adds the Odoo connection; v3 adds auto-send. Each step is a small change, not a big-bang rollout. |
| 4 |
Lower upfront commitment |
€6,000 impl fee vs €16,000 in the old model. The monthly subscription is the same whether we start with MVP or full product. Manel pays less today and upgrades are included. |
| 5 |
The system learns before it connects |
Running the MVP on real proposals for 4-6 weeks trains the model on Aronlight's specific catalogue patterns, edge cases, and formatting preferences. When v2 connects to Odoo, the AI already knows the business. This cuts v2 build time significantly. |
Platform
Can We Resell These Modules?
A2 is one of the most resellable modules we have. A5 (RMA) is deferred for Aronlight but remains a strong expansion play for the portfolio.
A2 as standalone product"Quote Automation" — email/WhatsApp to structured proposal, any ERP
Target segmentMid-market distributors, $10M–$500M, LatAm + EMEA, Odoo or SAP, WhatsApp-heavy sales
Pipeline candidates alreadyMAGG (A2 likely), Eurostar (adjacent), AutoalDia (A2 adjacent)
Second client price€2,500–3,500/mo (market rate, no design partner discount)
Bundle (A2+A5) for later"Distributor Ops Suite" — expansion play once A2 is proven
Long-term goal: SaaS like Mercura. We are starting as a boutique implementation shop (one client, configured for their exact setup). But the ambition is to productize — turn A2 into a configurable SaaS that any distributor can activate without a full custom build. Mercura raised $2.1M doing exactly this for the inbound RFQ side. Our differentiation: Odoo-first, LatAm/EMEA distribution segment, WhatsApp-native. Every Aronlight build decision should ask: "Can this be made configurable for the next client, or are we hardcoding something Aronlight-specific?" Build for reuse from day one.
Strategy
Deal Strategy — Notes for Raffaello
Decisions made — updated Apr 12
Lead with
A2 MVP — close one workflow fast, deliver it well, then expand. €5,000 impl + €2,000/mo subscription.
Entry point
€5,000 impl + €2,000/mo. Single price in v5 (no tiers). Design partner rate. Second client pays €2,500–3,500/mo. Annual: 25% discount (€1,500/mo, €18K lump sum at Month 3).
Design partner trade
Lower price in exchange for: reference case, testimonial, feedback during build, case study. Must be made explicit to Manel.
Walk-away floor
€5,000 impl + €1,500/mo annual (€18K/yr). Below this: margin doesn't work.
Proposal expiry
April 17, 2026. Pricing and timeline held until then. After that, scope and availability reassessed.
If price pushback
Don't discount silently. Two structures that work:
Structure 1 — Annual commitment (most likely). Offer annual billing: €1,500/mo (25% off), €18K paid as lump sum at Month 3. We get cash certainty, Manel gets a lower effective rate.
Structure 2 — Milestone impl payments. If upfront cash is the issue: €2,500 at signing + €2,500 when MVP goes live. Monthly subscription unchanged. We carry €2,500 of build cost risk for 4 weeks.
Floor: Never go below €5,000 impl + €1,500/mo annual. Below this the math doesn't work even with vibecoding.
A5 (RMA)
Deferred. Expansion play after A2 MVP is live and proving value. Keep the pain data, don't pitch it now.
A3 (spec matching)
Deferred by Manel. Revisit after A2 full product is running.
Contera
Out of scope. Manel confirmed. Do not bring up.
Manel's signal
Personal relationship with Giuseppe. Price-checking before committing. The new lower entry point (€6K vs €16K impl) removes the biggest friction. One more nudge to close.
Next Call Prep
Questions to Ask Manel
Updated Apr 9 — after Apr 8 alignment call. Answered questions are closed. Open items are the agenda for the next scope call.
❓ Still Open — Agenda for Next Scope Call
A2
Are inbound requests structured product codes or free-text descriptions?
Structured codes = simpler parsing. Free-text = NLP extraction layer needed, adds complexity and cost. This one question changes Phase 1 effort by 2–3 weeks.
A2
What format does the proposal output need to be? Email reply text, PDF, or Odoo document?
PDF = template engine + rendering layer. Email text = simpler. Affects Phase 1 scope significantly. Currently assumed email reply based on 90% email intake.
A5
Odoo stock movement write-back: required in Phase 1 or can it be Phase 2?
Phase 1 write-back forces full REST connector from day one (no CSV workaround). Deferring to Phase 2 = faster MVP, CSV path for A5 Phase 1 intake only. Raffaello to confirm.
A5
Customer intake: web form or WhatsApp/email bot? Note: "old-school customers batch returns in pallets every 2–3 months, won't use portal" (Apr 8). How does this affect the intake design?
If high-volume customers won't use a portal, we may need a hybrid: form for others + email/WhatsApp fallback with manual logging. Confirm with new RMA hire.
✅ Answered (Apr 8)
Revenue
Average order value: €4K–€19K range.
ROI confirmed at 2.9x on labor alone. Revenue upside from faster proposals could push this to 5x+. Use the range in the external doc.
A2
Requests arrive 90% by email, some WhatsApp/phone, directly to back-office.
User of A2 is back-office. No salesperson handoff to design around. Email is the primary integration surface for Phase 1.
A2
Catalogue changes every 1.5–2 years. New products added monthly.
Weekly CSV export is safe for Phase 1 MVP. No real-time API needed at launch.
A2
Odoo is live and stable. Staging environment available. Managed by third-party integration company.
Dev path: staging. Blocker for REST connector = API authorization from integration company (Manel negotiating).
A2
2,000 active SKUs in the Odoo catalogue.
Medium complexity for catalogue lookup. No exotic scaling required at this volume.
A2
Odoo is the single source for pricing (90% accurate). Manual overrides needed for USD/exchange rate cases.
Integration design: read Odoo pricing, flag USD-priced SKUs for back-office review. Confirmed goal: Odoo as 100% source of truth over time.
A5
~80 returns/month normal volume (Aronlight only). ⚠️ Verify unit — could be 80/week.
Primary framing: business continuity (owner left). Volume is secondary. Current 10x spike is temporary — Odoo migration.
A5
New hire started Apr 7 — 10 years RMA at Nvidia. She is the A5 champion.
Scoping call can proceed. Raffaello leads — he has prior RMA experience from a previous company.
Budget
Impl fee (€5K–€6K): significantly lower than previous €10K–€16K range. Monthly fee (€1,500–€2,000): more accessible entry point. "We are not far away."
New pricing removes the biggest friction. €6K impl is less than a month of a junior hire. Monthly fee framed as "less than one FTE." Signal: 🟡 Warm, trending green with lower entry point.
Actions
Open To-Dos — Before Next Call
Immediate (before Manel call)
Rafa
Confirm: does A5 (RMA) require the full REST connector for Phase 1, or can stock write-back be Phase 2 with a simpler intake flow first?
Affects whether we can demo A5 Phase 1 without the full connector. Changes timeline and risk.
Rafa
Check MAGG: do they have a quote generation problem (A2) and/or a returns problem (A5)? If yes, flag for shared architecture from the start.
If MAGG confirms both, we should build Aronlight with reusability in mind from day 1 — not Aronlight-specific code.
A2
Confirm with Manel: what messaging channel do their distributors use to send requests today?
Determines Phase 1 integration. If unclear, propose starting with Excel/form fallback and adding channel after demo.
Giuseppe
Decide on founder draw structure. Align with Raffaello before signing.
Go to P&L section ↓Affects real personal P&L from month 7 onward. Need agreement before committing to delivery timeline.
Medium-term (parallel to Aronlight build)
Giuseppe
Understand AdapttoAI product fit for AutoalDía (a.k.a. Aldea?). Confirm: is "Aldea" the same as AutoalDía? If yes: think through what A2 (quote bot for service requests) and A5 (returns / case management) would mean for their context. They have 300+ leads/month via WhatsApp + no CRM. What workflows overlap with the Aronlight scope?
Opens a potential second client conversation on the same modules. Could compress the time to the second deal significantly.
Both
Confirm with MAGG: are they going to use Returns Merchandise Authorization (A5)? Add to their next call agenda explicitly.
If MAGG confirms A5, it validates the module as a pipeline product, not just an Aronlight emergency fix.
A2
Get avg order value from Manel (catalogue routine order). Even a rough number unlocks the revenue upside calculation and can shift ROI from 2x to 5x+.
Most impactful question for the commercial case. If avg order is high, the revenue upside dwarfs the labor savings story.
Later
LATER - A5 Returns Management (RMA)
Deferred until A2 is live and validated. Builds on A2's Odoo connector. Estimated 5-7 weeks incremental build once A2 foundation is in place.
What A5 does: Replaces Excel + WhatsApp + phone returns handling with a structured digital flow. Customer submits return via guided form. Back-office processes through a defined 10-step workflow. All resolutions and stock movements tracked. Director dashboard for open cases, resolution rate, cost per incident.
Why deferred: A5 requires Odoo write-back (credit notes, resends, repairs) which needs the full REST connector - not available until v2 of A2 is complete. Also: new hire started Apr 7 with 10 years RMA experience at Nvidia, so the urgency gap is partially covered.
Volume: ~80 returns/month normal (verify: could be 80/week). Currently 10x spike due to Odoo migration - temporary.
Source material: Manel provided fluxo_devolucoes_rma_completo.pdf (7-step customer flow + 10-step back-office) + Global_Controlo_Incidencias.xlsx (current tracking spreadsheet).
Raffaello has prior RMA experience - assign lead when this activates.