SB StartupBasket
All ideas
74 /100 GO Low complexity

CodigoCerto — NFS-e code classifier for Brazil's Simples Nacional service providers

Type what you did, get the exact national service code — before the wrong one rejects your invoice or double-taxes you.

Compliance SaaS LATAM BrazilTax-techAI-agentSMBSolo-builderCompliance-drivenPortuguese-first
— views
Evaluation Scores
74/100

GO

Overall Score

15
Problem
12
Demand
13
Build
11
Distrib.
11
Revenue
8
Time
4
Defense

CodigoCerto — NFS-e code classifier for Brazil’s Simples Nacional service providers

1. One-liner

Type what you did, get the exact national service code — before the wrong one rejects your invoice or double-taxes you.

2. Trend signal — why now?

Brazil forced ~every service-issuing small business onto one national e-invoice standard, on a hard 2026 clock, and the single hardest field — the service classification code — has no good tooling.

  • Resolução CGSN nº 189/2026: from 1 September 2026, the NFS-e Padrão Nacional becomes mandatory for every ME and EPP under Simples Nacional that provides ISS services. Municipal portals that migrated have already deactivated their own emitters — “tentativas de novos faturamentos nos portais antigos serão rejeitadas automaticamente.” Over 3,400 municipalities (80%+ of urban population) were already on the national standard by Oct 2025.
  • MEI is already live and erroring. Accounting forums (contabeis.com.br) are full of MEIs who “sempre emiti nota corretamente… agora apareceu esta mensagem” — CNPJ/município mismatches, blank-vs-filled código de tributação municipal, browser workarounds. FENACON reported accountants hitting “instabilidade para emissão de NFS-e nacional.”
  • The classification field is the load-bearing risk. Pick the wrong código de tributação nacional and the client’s system can auto-withhold ISS, so the MEI pays DAS and eats a retention = double taxation. Worse: “uma classificação técnica errada neste ano resultará em cálculos incorretos de IBS/CBS quando a transição terminar.” The reform’s rejection-for-missing-IBS/CBS goes into production 3 August 2026 for regular-regime firms; Simples/MEI follow in 2027 — so the codes chosen now set the future tax.

This is a dated, hard-deadline migration where the government supplies a free emitter but nobody supplies confidence that the code is right. That gap is the product.

Provenance:

3. The opportunity

The incumbents (Notaas, Focus NFe, eNotas, TecnoSpeed, Conta Azul, and the free gov.br Emissor Nacional) all solve transmission — they turn a filled-in form into a valid NFS-e over the REST API. None of them solve “what do I put in the form.” They hand the MEI a search box for the código de tributação nacional and let them guess. Guessing wrong is silent: the note often goes through, then a client withholds tax, or an auditor flags a CNAE/code mismatch months later, or the reform recomputes IBS/CBS on a bad base.

The 10× wedge: a focused tool that takes a plain-Portuguese description of the service (“dei aula particular de inglês,” “consertei um vazamento,” “fiz o social media de uma loja”) and returns the single correct código de tributação nacional, its LC 116 item, the mapped NBS, whether ISS is due or withheld under SIMEI, and a one-line reason — with a flag when the description doesn’t match the CNPJ’s CNAE. Emission is a commodity; being right is not. We ride on top of the free emitter, we don’t compete with it.

4. Target market

  • Primary customer: Two wallets, one problem. (a) MEIs and micro service providers (~15M+ MEIs in Brazil, a large share ISS-service) who self-issue on gov.br and have no accountant. (b) Small contabilidade offices (contadores) with 50–500 MEI/ME/EPP clients each, drowning in Sept-2026 cutover tickets — the higher-value buyer.
  • Why they buy (in their words): MEI — “sempre deu certo, agora dá erro e não sei qual código usar.” Contador — clients flood WhatsApp asking which code to pick; a wrong pick creates a withholding dispute the accountant has to unwind for free.
  • Rough TAM reasoning: Brazil has 15M+ MEIs and ~5M+ ME/EPP; a conservative service-provider slice is millions of businesses and tens of thousands of small accounting firms. We need a few thousand paying seats/offices — a rounding error against the base.
  • Why now for them: The 01/09/2026 deadline is a forcing function. Before it, municipal portals hid this behind familiar local code lists; after it, everyone faces the same unfamiliar national taxonomy at once.

5. Product sketch (MVP)

  • Describe-in-plain-Portuguese → correct code. Free-text service description in, código de tributação nacional + LC 116 item + NBS + confidence + one-sentence rationale out.
  • CNAE cross-check. Paste/lookup the CNPJ; flag when the suggested service code is inconsistent with the registered CNAE (the #1 rejection/withholding trigger).
  • ISS / SIMEI treatment note. Tell the user plainly: is ISS due, withheld, or covered by the DAS — so they don’t get double-taxed.
  • Município readiness check. Look up whether the user’s município is already on the Padrão Nacional and when its old portal shuts, so they know their personal deadline.
  • Batch mode for accountants. Upload a client list (service description per client) → get a mapped code table back, ready to configure in whatever emitter they use.
  • Reform-forward note. Show which IBS/CBS classification the chosen code implies, so today’s pick is future-proof.
  • Copy-out, not lock-in. One-click copy the exact values to paste into gov.br or any provider — we’re the brain, not the pipe.

6. AI angle — what’s load-bearing

Remove the AI and this is a static lookup table nobody can navigate — which is exactly the failed status quo. The value is turning a messy, colloquial description of work into the one correct code out of hundreds, disambiguating near-synonyms (“consultoria” vs “assessoria” vs “treinamento” have different LC 116 items and ISS treatment), and reconciling it against the CNAE. That’s a constrained classification-with-reasoning task over a fixed taxonomy — LLMs are strong at it, and the fixed taxonomy means we can ground every answer and cite the rule, keeping hallucination in check. The AI is the product; the emitter is a paste target.

7. Localization angle

This is the localization play — it’s unbuildable without Brazil-specific fiscal knowledge: LC 116 service list, NBS nomenclature, código de tributação nacional, SIMEI/ISS interplay, per-município cutover schedules, and the reform’s IBS/CBS mapping. Portuguese-first, priced in reais (a ₹499-equivalent tier works where a US$49 tier wouldn’t — MEIs pay ~R$70/mo DAS total). No global competitor can parachute in; the moat and the market are both the local rulebook.

8. Business model — path to $1M–$5M ARR

  • Pricing:
    • MEI/self-serve: R$19–29/mo (or R$9 per one-off resolved code) — impulse-priced against the pain of one withholding dispute.
    • Accountant office: R$149–399/mo per office for batch mode + client cross-checks (they bill it into advisory fees).
  • ACV: ~R$2,000–4,000/yr for accountant offices; ~R$250/yr for self-serve MEIs. Offices carry the revenue.
  • Rough math to $1M ARR (~R$5.5M): ~1,200 accountant offices × R$250/mo × 12 ≈ R$3.6M, plus ~7,000 self-serve MEIs × R$24/mo × 12 ≈ R$2M. Both counts are trivial against the base.
  • Rough math to $5M ARR: become the default “code check” embedded in a couple of mid-size emitter/ERP partners (revenue share on their MEI base) + 5,000 offices. Requires one distribution partnership landing.
  • Expansion path: add NF-e (goods) classification, reform-transition advisory (IBS/CBS impact per client), and a monitored “your codes vs. new rules” alert as the 2027–2033 transition rolls — recurring reason to stay subscribed past the deadline spike.

9. Go-to-market wedge — first 100 customers

  • Ride the deadline in the accountant forums. contabeis.com.br and FENACON channels already have active threads of contadores hitting NFS-e errors. Answer them with the tool’s exact output + a link. Warm, self-selecting, dated demand.
  • Contador WhatsApp/Telegram groups. Accounting offices cluster in regional WhatsApp groups; a batch-mode demo that clears a Sept-cutover backlog in one upload is a forwardable artifact. Seed 20 offices, let the batch table spread.
  • “Qual meu código?” SEO + free single-lookup. People are actively Googling “código de tributação nacional MEI.” A free single-code answer captures them at the moment of pain; paywall batch + CNAE cross-check + reform note. Ship a free tool now, before Sept, when search spikes.
  • maisMEI / MEI-influencer partnerships. Portuguese MEI-help creators and portals (maismei.com.br audience) already produce this content; co-branded lookup widget = distribution to millions of MEIs.

10. Build complexity — justification

Low. Off-the-shelf LLM + the published LC 116 ↔ NBS ↔ código de tributação nacional correlation tables (public, finite) as a grounded retrieval base, plus the free CNPJ/CNAE lookup and the município-habilitação list from nfse.gov.br. No emitter integration required for v1 (copy-out). The genuine work is curating the taxonomy + ISS/SIMEI rules correctly and tuning the classifier — domain, not engineering, is the hard part. Solo builder with a Brazilian accountant advisor ships a credible v1 in 4–6 weeks, well inside the Sept-2026 window.

11. Gating checklist

GatePass?Note
Legal in target market✅Advisory/classification tool; not issuing fiscal docs or acting as fiscal authority. Add “confirm with your accountant” disclaimer.
Ethical — no harm / dark patterns✅Reduces mis-taxation; helps the smallest businesses. Must be honest about confidence.
Market exists (evidence above)✅Hard deadline + live forum complaints + millions of obligated businesses.
1–5 person team can build this✅Solo + domain advisor, 4–6 weeks.
Launchable with <$50K / ₹40L✅LLM API + web app + taxonomy curation. Low four figures to launch.

All five pass.

12. Feasibility score

AxisWeightScoreNotes
Problem intensity2015/20Real and dated — wrong code = rejection, double taxation, future IBS/CBS error. But per-note pain is intermittent, not hourly hair-on-fire for a solo MEI.
Demand evidence1512/15Live forum complaints, hard 01/09/2026 mandate, millions obligated, active search intent. Not yet proven anyone pays specifically for classification vs. emission.
Build feasibility1513/15Bounded taxonomy + LLM; no emitter integration in v1. Domain curation is the effort, not infra.
Distribution clarity1511/15Clear warm channels (accountant forums, WhatsApp groups, MEI portals) but conversion from free-lookup to paid is unproven.
Revenue mechanics1511/15Office tier carries it and is benchmarkable; self-serve ARPU is thin and low-reais pricing needs volume.
Time to first revenue108/10Accountants buy fast pre-deadline; a batch demo can close in days.
Defensibility104/10Taxonomy is public; emitters could bolt this on. Moat is curation quality, accountant trust, and speed — copyable in ~12 months.
Total10074/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required — you need someone who can build a grounded classifier and a Brazilian accountant who lives in LC 116/NBS/SIMEI. Without the domain half, the answers will be plausibly wrong, which is fatal for a tax tool.

Key assumptions to validate (3–5)

  1. Assumption: Accountants will pay R$149–399/mo for a classification/cross-check tool separate from their emitter. How to test: Pitch batch mode to 30 small offices from contabeis threads; pre-sell 10 at the office price before building past MVP.
  2. Assumption: The classifier hits high accuracy on real, messy descriptions (needs to be right enough to trust). How to test: Build a 300-item gold set from real MEI service descriptions + verified codes; require ≥95% top-1 on common services before launch.
  3. Assumption: Self-serve MEIs will convert from a free single lookup to a paid tier. How to test: Ship the free lookup now, measure paid-conversion on the CNAE cross-check + reform note behind the wall.
  4. Assumption: Emitters won’t ship “smart code suggestion” as a free feature before we build trust. How to test: Audit current Notaas/Focus/Conta Azul UIs monthly; move fast on the accountant relationship (the sticky part they can’t copy).

Risk flags

  1. Defensibility risk: The taxonomy is public and emitters own the emission surface. If Conta Azul/Notaas add a good AI code-picker, our standalone value erodes. Mitigation: go deep on the accountant workflow + reform advisory, and pursue an embed/partnership before they build.
  2. Accuracy/liability risk: A confidently-wrong code that causes double taxation destroys trust and invites blame. Mitigation: ground every answer, show confidence + rationale, disclaim, and lean on the accountant-in-the-loop for the office tier.
  3. Timing risk: Demand spikes hard around 01/09/2026 then flattens. Mitigation: convert the deadline spike into recurring value via reform-transition monitoring (2027–2033) so it isn’t a one-shot event product.
  4. Platform dependency: Relies on nfse.gov.br/CNPJ lookups and published correlation tables staying available; a format change breaks ingestion. Manageable but real.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical solo founder + Brazilian accountant (contador) domain advisor
Time to revenue:        4–8 weeks (pre-deadline accountant pre-sales)
Capital to launch:      R$15–30k (~$3–6K) — LLM API, web app, taxonomy curation, advisor
Top 3 assumptions to validate first:
  1. Accountants pay office-tier for classification+cross-check — pre-sell 10 offices before full build
  2. Classifier ≥95% top-1 on a 300-item real-description gold set — build the set and measure
  3. Free single-lookup converts to paid on CNAE cross-check + reform note — ship free tool, measure
Kill criteria:
  - Abandon if <5 of 30 pitched accountant offices pre-commit at the office price
  - Abandon if classifier can't clear 90% top-1 on the gold set after two curation passes
  - Abandon if a major emitter ships a free, accurate AI code-picker before our v1 and accountants shrug at the standalone

15. Next step — 1-week validation sprint

  • Day 1–2: Build a 300-item gold set from real MEI/ME service descriptions (pulled from forum posts + an accountant advisor) mapped to verified código de tributação nacional / LC 116 / NBS. This is the falsifiable core.
  • Day 3–4: Wire a grounded LLM classifier over the correlation tables; measure top-1 accuracy against the gold set. Fix the obvious failure classes.
  • Day 5: Take the working single-lookup + a batch demo to 30 small accounting offices from contabeis/FENACON threads and WhatsApp groups; pitch the office tier.
  • Decision (go/no-go): GO only if (a) classifier clears ≥90% top-1 on the gold set and (b) ≥5 of 30 offices pre-commit at R$149+/mo. Both are hard numbers — no “they seemed interested.”

Interested in a detailed proposal?

Get a deep-dive with market research, competitive analysis, and implementation roadmap.

Contact us

info@startupbasket.ai