CreditSentry

1. One-liner

Flags which suppliers are about to cost a Brazilian buyer its IBS/CBS credit, before the payment goes out.

2. Trend signal — why now?

Brazil just rewired the single most important mechanic in its tax system, and almost nobody has priced the consequence.

Under Article 47 of Complementary Law 214/2025, a buyer’s right to an IBS/CBS credit no longer comes from the purchase. It comes from the extinction of the supplier’s tax debt. Plainly: you can do everything right, pay your supplier in full, file perfectly — and still lose the credit because the guy who sold to you didn’t settle with the tax authority. As one tax firm put it, the credit “deixou de ser automático e passou a depender do cumprimento fiscal do prestador” — it stopped being automatic and now depends on the supplier’s compliance.

Three things landed in the last 90 days that turn this from a legal footnote into an operational emergency:

1. Split payment went to production on May 4, 2026 (Phase 1: Pix and boleto). Tax is now withheld at the moment of financial settlement. The system queries in real time whether the seller holds enough input credits before withholding. Companies that used to sit on tax money for 30–45 days as free working capital lost that float overnight.
2. August 3, 2026 — invoice rejection went live. NF-e and NFC-e without correctly populated IBS/CBS fields are now rejected outright by SEFAZ for Lucro Real and Lucro Presumido companies. No invoice means goods don’t leave stock and the service isn’t official. Billing simply stops.
3. The compliance gap is concentrated in exactly the wrong place. Federal Revenue data from June 2026: 75% of invoices from Normal Regime companies were already compliant — but only 8% from Simples Nacional companies. Simples Nacional firms are the small suppliers that mid-market buyers depend on, and their mandate doesn’t bite until January 2027.

The economics of this are brutal and specific. Lucas Madureira of Gedanken gave the cleanest version: a supplier offering a 10% discount can trigger losses reaching roughly 27% in tax credits under the new model. The cheapest supplier is now frequently the most expensive one. Meanwhile the penalty for appropriating a credit you weren’t entitled to runs 75% of the improper credit, with 66% for failure to reverse.

And the direction of risk is getting worse, not better: fiscal irregularity among small companies rose from 9.6% to 11.4%, while medium and large companies moved the opposite way, falling from 10.4% to 7.2%. The risk is migrating down-market into the supplier tail.

Every serious advisory in Brazil is now telling companies the same five things: identify critical suppliers by credit exposure, formalize vendor approval policies, automate fiscal verification, integrate procurement with tax, and track supplier fiscal health on a dashboard. That advice is everywhere. The tool that actually does it for a 40-person company is nowhere.

Provenance:
  - Signal 1 (Demand): Art. 47 LC 214/2025 ties buyer credit to supplier debt extinction; advisories broadly warn of the "invisible risk" and recommend continuous supplier fiscal monitoring — https://decisionit.com.br/blog/cbs-e-ibs-credito-depende-do-fornecedor/ — 2026-08-04
  - Signal 2 (Feasibility): Split payment in production since 2026-05-04; NF-e IBS/CBS rejection enforced 2026-08-03; RFB June data shows 75% Normal Regime vs 8% Simples Nacional compliance — https://exame.com/economia/ibs-e-cbs-na-nota-fiscal-erro-podera-travar-faturamento-a-partir-de-3-de-agosto/ — 2026-08-04
  - Signal 3 (Economic): Supplier fiscal health moving to the center of purchasing decisions; 10% supplier discount can produce ~27% credit loss; SMB irregularity rising 9.6%→11.4% — https://brazileconomy.com.br/economia/2026/03/reforma-tributaria-amplia-riscos-na-cadeia-de-fornecedores-e-deve-elevar-custos/ — 2026-08-04
  Category: Regulatory arbitrage

3. The opportunity

There are two markets in Brazilian tax-reform software right now. One is crowded. One is empty.

The crowded one is outbound classification. ClassTax, FiscoPro, ClassTrib, Contabil Descomplica, SOS Reforma, Tax Radar — a dozen products racing to map your NCM catalog to cClassTrib codes so your invoices don’t get rejected. That fight is over; the market is saturated and commoditizing toward per-item pricing. I’d stay out.

The empty one is inbound credit risk. Nobody is protecting the buyer. Every one of those classification tools makes sure your invoice is valid. None of them tell you that the supplier you’re about to pay R$180K has a lapsed CND, sits in Simples Nacional with a broken invoice history, and is about to vaporize R$24K of your credit.

The incumbents who could serve this are all pointed at the enterprise. Gedanken, Neoway, and Serasa sell supplier homologation and due-diligence platforms — 400+ databases, ESG criteria, PEP and sanctions screening, negative media, four-week structured onboarding. That’s a procurement-department product with a procurement-department price, sold to companies with a procurement department. It is magnificently wrong for a R$40M-revenue distributor in Campinas whose “procurement department” is one finance manager and a spreadsheet.

The gap is a narrow, cheap, fast product that answers exactly one question, continuously: which of my suppliers is going to cost me credit, and how much?

This is the classic enterprise-tool-that-SMBs-can’t-afford arbitrage, except with a hard regulatory forcing function and a dated deadline. The buyer doesn’t need due diligence. They need a sentry.

4. Target market

• Primary customer: Finance manager / controller (gerente financeiro, controller) at a Brazilian company on Lucro Presumido or Lucro Real, R$10M–R$150M annual revenue, 20–200 employees, with 50–600 active suppliers. Concentrated in distribution, light industry, construction materials, food service supply, and B2B services. São Paulo, Minas, Paraná, Rio Grande do Sul first.

• Why they buy: Their credit position just became hostage to other people’s tax behavior, and they have no visibility into it. Advisories are explicit that inconsistencies “na outra ponta da cadeia” — at the other end of the chain — can block credits for a company that did everything right. The controller’s actual fear is concrete: signing off on a payment run, then discovering three months later that a chunk of credit never materialized, with a 75% penalty if they claimed it anyway. Right now their entire control is asking for a CND once a year at onboarding — and the advisory consensus is that relying on negative-debt certificates alone is no longer sufficient.

• Rough TAM reasoning: Brazil has roughly 1.3M companies outside Simples Nacional in the Lucro Real / Lucro Presumido regimes — the cohort already under the August 2026 mandate. Cut to the segment with meaningful supplier counts and a person whose job it is to care: call it 120K–200K realistic accounts. Capture 1,000 of them at R$490/mo and that is ~R$5.9M ARR (~US$1.1M). I do not need a large share of this.

• Why now for them: Split payment is live. Invoice rejection is live as of yesterday. January 2027 brings Simples Nacional suppliers into the regime and CBS collection begins in earnest. Their supplier base is at its least compliant precisely when the penalty for supplier non-compliance transfers to them. The window where this is an urgent, board-visible problem is roughly the next 18 months.

5. Product sketch (MVP)

• Supplier base import in one step — upload XMLs of received NF-e (companies already keep these), or connect the ERP. The product reconstructs who you actually buy from, how much, and how often. No manual supplier registry to build.
• Credit exposure ranking — every supplier scored not by generic risk but by reais of IBS/CBS credit at risk, based on your real purchase volume with them. The controller sees the top 20 suppliers that matter and ignores the tail.
• Continuous fiscal status monitoring — CND, PGFN standing, CNPJ registration status, regime (Simples vs Normal), and SEFAZ/Sintegra state registration, re-checked on a schedule rather than once at onboarding.
• Invoice-health check on inbound NF-e — flags received invoices whose IBS/CBS fields, CST, or cClassTrib look wrong or missing, which is the leading indicator that this supplier’s credit will not survive.
• Pre-payment alert — before a payment run, a plain-language list: “These 4 suppliers, R$312K in payments, R$41K of credit at risk. Here’s why for each.”
• Supplier nudge letters — one-click Portuguese email/WhatsApp to the offending supplier explaining exactly what’s wrong with their invoices and what to fix. Most small suppliers aren’t malicious; they’re at that 8% compliance rate and don’t know it.
• Audit trail — a timestamped record showing you performed reasonable verification before appropriating each credit. This is the artifact that matters when the assessment arrives.
• Monthly credit reconciliation — credits you expected vs. credits that actually survived, with the delta attributed by supplier.

6. AI angle — what’s load-bearing

Remove the AI and this collapses into a CND-checking cron job — which, to be blunt, is what a competitor could build in a weekend and why I’ve scored defensibility low.

The load-bearing AI work is in three places:

Invoice-health judgment. The classification problem is genuinely not a lookup. As one vendor analysis puts it directly, mapping NCM to cClassTrib “requires analysis of product description, not just the NCM code” and it is “not possible to create a simple from/to between NCM and cClassTrib” because classification depends on contextual reading of the operation and possible exceptions or reductions. Judging whether a supplier’s invoice to you is correctly classified requires reading the item description against the operation context. That’s an LLM task, not a rules table.

Risk synthesis in plain Portuguese. The raw inputs are a mess of CND statuses, PGFN entries, regime flags, and invoice anomalies across hundreds of suppliers. Turning that into “Fornecedor X: CND vencida há 40 dias, 12 notas com cClassTrib inconsistente, R$18.400 de crédito em risco neste trimestre” is summarization and prioritization work that no dashboard does well.

The nudge letters. Writing a correct, specific, non-insulting Portuguese message to a small supplier explaining which field on which invoice is wrong — at scale, personalized per supplier and per error — is exactly what generative models are good at and what nobody does manually.

The AI is doing the judgment and the writing. The data plumbing is boring and off-the-shelf.

7. Localization angle

This is a Brazil-only product and that is a feature, not a limitation. The entire opportunity is the localization.

• Regulatory specificity: LC 214/2025, Art. 27 and Art. 47, the CST/cClassTrib tables from Informe Técnico RT 2025.002, the split-payment standard vs. simplified procedures. No global tax product will model these because the rules are brand new, Brazil-specific, and changing quarterly through the 2026–2033 transition.
• Data rails: NF-e XML is a universal, government-mandated, machine-readable artifact every Brazilian company already possesses. That is a gift — it means zero-friction onboarding without an ERP integration. There is no equivalent in most markets.
• Cheap verification infrastructure: CNPJ/CND/PGFN/Sintegra query APIs are a commodity in Brazil, with plans from around R$149/mo and per-query pricing that drops with volume.
• Pricing: R$390–990/mo lands correctly for a mid-market Brazilian controller. The enterprise homologation platforms are priced far above this and sold through a sales motion these companies never enter.
• Distribution: the contador (accountant) is the trusted advisor in Brazilian SMB finance in a way that has no US analog. Accounting offices each hold 50–300 client companies. That’s the channel.

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

• Pricing (direct): R$390/mo up to 100 suppliers · R$690/mo up to 300 · R$1,290/mo up to 800 + ERP connector. Annual prepay at 2 months free, which Brazilian SMBs take readily.
• Pricing (accounting-office channel): R$1,900/mo for an office managing up to 25 client companies, R$4,500/mo up to 75. Offices resell or bundle into their advisory fee. This is the higher-ACV motion and where I’d push.
• ACV: ~R$8,300 (~US$1,550) blended direct. ~R$34,000 (~US$6,300) for accounting offices.
• Math to $1M ARR (~R$5.4M): 400 direct accounts (R$3.3M) + 60 accounting offices (R$2.1M). Roughly 460 logos. Achievable in 12–18 months if the accountant channel works at all.
• Math to $5M ARR (~R$27M): needs ~250 accounting offices averaging 40 client companies each, plus ~1,200 direct accounts. What must be true: the accountant channel compounds (each office brings clients without per-client selling), and the January 2027 Simples Nacional deadline creates a second, larger buying wave. That second wave is the real unlock — it triples the addressable base.
• Expansion path: supplier count tiers → ERP connector → multi-CNPJ groups (very common in Brazil; companies routinely operate 3–8 CNPJs) → the supplier-side product. That last one is interesting: once buyers are nudging suppliers, you have a warm path to selling the supplier a R$99/mo “keep your invoices clean so your customers keep buying from you” tier. That’s a genuine second act, not a hypothetical one.

Gross margin is comfortable — the variable cost is per-CNPJ verification queries and LLM inference on invoice batches. Call it R$40–90/mo per account at the mid tier.

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

The unusual advantage here: I can compute a prospect’s exposure before they’ve ever heard of me, because a large share of the inputs are public.

1. The exposure report cold open (customers 1–40). Pull companies by CNAE and size from public CNPJ data in 3 target states. For each, check the fiscal standing of identifiable counterparties and build a one-page PDF: “We looked at your supplier base. Here are 6 suppliers whose fiscal status puts an estimated R$X of your IBS/CBS credit at risk this quarter.” Send to the gerente financeiro by email and LinkedIn. This isn’t a pitch, it’s a finding — and it’s specific enough to be uncomfortable. Send 600, expect 12–18% reply on a document that names their actual suppliers, close 40.

2. Accounting offices via CRC and the contábil media (customers 40–75). The Brazilian accounting press is saturated with tax-reform anxiety content right now — Contábeis, Jornal Contábil, and CRC state councils are publishing on cClassTrib errors and supplier credit risk weekly. Offer free co-branded webinars: “Art. 47 — como proteger o crédito do seu cliente.” One 45-minute webinar to 200 accountants converts a handful of offices, and each office is worth 25–75 client companies. Target the 500 mid-size offices in SP/MG/PR. This is the compounding channel; I’d start it in week one and let it cook.

3. ERP and NF-e integrator partnerships (customers 75–100). Omie, Contmatic, Sankhya, and TecnoSpeed all serve exactly this mid-market and are all publicly publishing tax-reform content — they have the customer relationship and no inbound-credit product. Offer revenue share as an add-on module. Two partnerships get you to 100 and past it.

4. The rejection wave itself. As of August 3 invoices are being rejected. Companies are actively searching, in panic, right now. A tight Portuguese content play targeting “rejeição NF-e 1024”, “cClassTrib incompatível”, and “crédito IBS fornecedor” catches people at the exact moment of pain. Cheap, and it compounds into 2027.

The honest weak point: this is a sales-led motion in a market where I’d want a Portuguese-native co-founder or first hire. That’s priced into the score.

10. Build complexity — justification

Medium. Roughly 10–14 weeks to a sellable v1 for two people.

Off-the-shelf: NF-e XML parsing is a solved problem with mature Brazilian libraries; CNPJ/CND/PGFN/Sintegra queries are commodity paid APIs; LLM calls for classification judgment and letter generation are standard.

The real work is threefold: encoding the CST/cClassTrib compatibility rules from Informe Técnico RT 2025.002 well enough to spot supplier invoice errors reliably; building the exposure model that converts “supplier has a problem” into “R$X of your credit is at risk”; and keeping current as the transition rules change through 2027. That last one is ongoing operational cost, not a one-time build — and it’s also part of the moat.

No ML training, no novel infrastructure, no hardware. It’s integration work plus domain encoding, which is the sweet spot.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Uses public fiscal registries and the customer’s own invoice data, with their authorization. Standard practice.
Ethical — no harm / dark patterns	✅	One caution: this could be used to blacklist small suppliers unfairly. Mitigated by designing the nudge-letter flow as the default action — help the supplier fix it before dropping them. Worth holding to deliberately.
Market exists (evidence above)	✅	Live regulation, live deadlines, enterprise incumbents already monetizing the adjacent enterprise version.
1–5 person team can build this	✅	Two people, 10–14 weeks.
Launchable with <$50K / ₹40L	✅	~US$12–18K: API credits, LLM inference, a fiscal consultant on retainer for rule verification.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Money, monthly, with a 75% penalty attached. A 10% supplier discount producing ~27% credit loss is the kind of number that gets a controller’s attention immediately. Not 19–20 only because the pain is partly anticipated — the largest losses land in 2027, and some controllers will procrastinate accordingly.
Demand evidence	15	12/15	Strong indirect evidence: live regulation, hard dates, RFB compliance statistics, enterprise incumbents charging real money for the adjacent product, advisory firms uniformly recommending exactly this monitoring. Docked 3 because I found no verbatim SMB customer complaints — the Brazilian mid-market controller doesn’t post on Reddit. That’s a real gap in my evidence and the first thing I’d close.
Build feasibility	15	11/15	Standard stack, commodity APIs, well-defined data. The cClassTrib rule encoding and the moving-target transition rules are genuine work. Two people, 10–14 weeks.
Distribution clarity	15	12/15	The exposure-report cold open is specific and unusually strong — I can name their suppliers before they reply. The accountant channel is the proven route into Brazilian SMB finance. Docked for needing Portuguese-native sales capability.
Revenue mechanics	15	12/15	Pricing benchmarked below enterprise homologation platforms and above commodity API tools; ~460 logos to $1M ARR is realistic. The $5M path leans on the accountant channel compounding and the 2027 wave — one solid assumption and one hopeful one.
Time to first revenue	10	7/10	10–14 weeks to v1, then a 4–8 week B2B cycle. Pre-selling on exposure reports during the build could pull this in, but I won’t score on the optimistic case.
Defensibility	10	5/10	Honest assessment: this is execution-and-timing, not moat. The classification market saturated in months and this could too. Real assets accrue with time — the supplier fiscal-history database compounds, buyer-side workflow lock-in is sticky once it’s in the payment-run process, and the accountant relationships are hard to dislodge. But at month 3 there’s nothing stopping a fast follower.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

This needs someone who can build integrations and someone who genuinely understands LC 214/2025. A Brazilian tax professional as co-founder or first hire isn’t optional — the rule encoding is the product and getting it wrong destroys trust immediately. Note this is not primarily a sales-heavy play, because the accountant channel does the selling once seeded.

Key assumptions to validate

1. Assumption: Mid-market controllers perceive supplier-driven credit loss as their problem today, not a 2027 problem to defer. How to test: 25 interviews with gerentes financeiros at R$10–150M companies. Ask what they did differently after August 3. If the answer is “nothing yet, our accountant handles it,” the urgency is a year early and I’d delay or lead with the accountant.
2. Assumption: Enough of the credit-risk signal is computable from public registries plus the buyer’s own received NF-e XMLs to produce an exposure number a controller believes. How to test: build the exposure report for 10 real companies using only obtainable data; show it to 3 tax professionals and ask whether the number is defensible. This is the technical crux — if the number is mush, there’s no product.
3. Assumption: Accounting offices will distribute this rather than treat it as a threat to their advisory billings. How to test: pitch 15 offices in SP/MG. Measure how many will co-host a webinar. Offices that see it as competition will say so fast.
4. Assumption: R$690/mo clears the bar without a procurement process. How to test: put real pricing in front of the first 25 prospects. If it routes to a committee, the ACV is wrong for the motion.

Risk flags

1. Fast-follower risk (highest). The cClassTrib classification market went from empty to a dozen vendors in under a year. The same could happen here, and ClassTax or FiscoPro extending from outbound classification to inbound credit risk is the obvious move — they already have the rule engine and the customers. Speed and the accountant channel are the only real defenses.
2. Regulatory churn. The 2026–2033 transition means rules keep moving; the JPC-style amendments and RFB technical notes arrive continuously. This is simultaneously the moat and a permanent maintenance tax. Underestimate it and the product silently goes wrong.
3. Market timing. Split payment and credit consequences bite hardest in 2027–2028. Launch too early and you’re selling insurance against a fire nobody has smelled yet. I think August 2026 is right — the invoice rejections make it visceral — but this is the assumption I’m least sure of.
4. Attribution risk. When a credit is denied, proving which supplier caused it, and that the tool would have caught it, is messy. Overclaiming precision here is the fastest route to churn. The product should promise prioritized visibility, not guarantees.
5. Ethical drift. A tool that scores small suppliers on fiscal health can quietly become a blacklisting machine that squeezes exactly the Simples Nacional firms that are 8% compliant through no real fault. Lead with the nudge letters. If the product’s main output becomes “drop these vendors,” it’s doing harm and will also attract the wrong kind of attention.

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + Brazilian tax professional (contador or
                        tax lawyer) as co-founder. Portuguese-native. São Paulo based.
Time to revenue:        4–5 months (10–14 weeks to v1, 4–8 week sales cycle)
Capital to launch:      R$70–100K (US$12–18K)
Top 3 assumptions to validate first:
  1. Controllers feel this as a 2026 problem, not a 2027 one — 25 interviews with
     gerentes financeiros, asking what changed after August 3
  2. A defensible exposure number is computable from public data + received NF-e XML —
     build 10 real reports, have 3 tax professionals judge them
  3. Accounting offices distribute rather than compete — pitch 15 offices, count
     webinar commitments
Kill criteria:
  - Abandon if <5 of 25 interviewed controllers can name a specific supplier they
    worry about, or if the modal answer is "our accountant handles that"
  - Abandon if the exposure calculation can't be made defensible from obtainable
    data — if tax professionals call the number speculative, there is no product
  - Abandon if an established classification vendor (ClassTax, FiscoPro, Contmatic,
    Omie) ships inbound supplier credit monitoring before your v1 — they have the
    rule engine and the customer base and you will not win that race from behind
  - Abandon if fewer than 2 of 15 accounting offices will co-host a webinar

15. Next step — 1-week validation sprint

• Day 1–2 — Prove the number. Get received-NF-e XML sets from 3 friendly mid-market companies (a contador can broker this in an afternoon). Build the exposure report by hand: parse suppliers, check fiscal standing, flag invoice anomalies, compute reais at risk. The only question that matters: does a real, non-trivial number fall out, or is it noise?
• Day 3–4 — Sell the report. Take those 3 reports to the companies’ finance managers. Don’t demo anything; hand them the PDF and shut up. Watch what they do. Then run 12 more calls with controllers sourced through accounting contacts, and ask a single closing question: “If this arrived monthly, would you pay R$690?”
• Day 5 — Test the channel. Pitch 8 accounting offices in São Paulo on co-hosting a webinar about Art. 47 supplier credit risk. Measure commitments, not compliments.

Go/no-go: proceed only if (a) the hand-built exposure report surfaces ≥R$15K of plausible at-risk credit for at least 2 of the 3 companies, (b) ≥5 of 15 controllers say yes to R$690/mo unprompted, and (c) ≥2 of 8 accounting offices commit to a date for the webinar.

Any two of those failing and this is a 2027 idea, not a 2026 one — worth shelving and revisiting in six months when the credit denials start landing and the pain stops being theoretical.