Skip to content
ACUSE Investor pitch

Acuse · investor pitch · fall 2026

Finds what a SAT audit finds. Before SAT does.

Every Mexican company runs on CFDI, government-stamped XML invoices. Acuse reads them and answers, in pesos and in seconds, the question every CFO and accountant faces monthly: what would the tax authority find if it audited us today?

Live demo Product site

Screened against SAT’s live 69-B list

12,097
0 86% flagged as risk 14,054

11,138 Definitivo plus 959 Presunto, out of the 14,054 RFCs Acuse parses from the official list.

Live product 100% client-side Monterrey, MX Solo founder

01The problem

Mexico’s tax authority automated enforcement. Companies did not automate defense.

  • SAT’s electronic audits are automated cross-checks of invoices, payment complements and bank flows, the exact data Acuse analyzes.
  • Article 69-B lets SAT retroactively void deductions from blacklisted suppliers. The blacklist is public, but checking it is manual, so it happens late and never covers every invoice.
  • Credit invoices without a payment complement (REP) mean non-creditable IVA, including partially-paid installments.
  • Today this checking is junior accountants, a government CSV, and Excel. Every month, for every client, at every accounting firm.

02The product

Upload the month’s XMLs. Get the audit verdict in pesos.

Live today. Runs 100% in the browser, so tax data never leaves the machine, which is the whole objection-killer for accounting firms handling client data. Six findings, each quantified, each with the exact invoices listed.

  • 69-B blacklisted suppliers Pesos
  • PPD without REP (installment-aware) Pesos
  • Invoices with no bank charge Pesos
  • Charges with no invoice Pesos
  • Duplicate folios Control
  • Malformed RFCs Control
Live demo, sample data, full report Product site (Spanish)

03Validated with real data

Not a mockup. The engine already ran against the real files.

Four readings, taken with the files themselves: the government’s own blacklist, public stamped invoices, an end-to-end run and a two-implementation parity check.

14,054

RFCs on the live 69-B list

11,138 Definitivo plus 959 Presunto flagged as risk. Desvirtuado and Sentencia Favorable correctly excluded.

Source: SAT, listado completo 69-B CFF.

9/9

Reference CFDIs, first try

Real stamped 3.3 and 4.0 invoices, Pagos 2.0 payment complements with installments, discounts, namespace edge cases.

Source: public CFDI ecosystem fixtures.

1 hit Definitivo

Caught end to end

Sample portfolio against the live official 69-B list: a genuinely blacklisted supplier RFC, with the exposure quantified. The exact report a despacho would hand its client.

Run: real 69-B list against sample portfolio.

2 engines

Python and browser, in parity

Same analysis in Python (reference CLI) and dependency-free client-side JavaScript. On every load the browser replays a seeded case battery pinned by the Python reference CLI; a mismatch blocks the demo.

Parity self-check on every load.

04Market

Sold through accounting firms. Every sale multiplies.

MX$100–300 per client company per month. A despacho manages dozens to a few hundred client companies, and the day-17 filing deadline returns every month, built-in recurrence.

ARR by number of accounting firms

Illustrative: 100 client RFCs per firm, MX$200 per RFC per month.

  1. 10 firms MX$2.4M ≈ US$140K
  2. 50 firms MX$12M ≈ US$700K
  3. 200 firms MX$48M ≈ US$2.8M

Arithmetic: firms × 100 RFCs × MX$200 × 12, USD at ~17 MXN. US$50M ARR ≈ 350K client companies via ~3,500 firms, in a country with millions of active CFDI-issuing taxpayers. Same playbook extends to Colombia, Chile, Peru and Brazil, all with mandatory e-invoicing.

05Model and plan

Free report first. Paid monthly close after.

Wedge → close

Free risk report on the firm’s own client data. No signup friction, nothing to trust, runs the same day. Convert to the paid monthly close: SAT mass-download, bank reconciliation, draft filings (DIOT) delivered before day 17.

Inertia, not competition, is the honest obstacle. The wedge fights it with pesos at risk on the firm’s own clients, and the report lands as the firm’s own catch: “we found this before SAT did.” The moat is the workflow, not the parser. Every month a firm closes with Acuse, switching costs compound. And the core report stays client-side permanently; AI features are opt-in, per document.

Next 6 months

  • M1–2 · 10 Monterrey despachos running the free report; SAT mass-download with client e.firma
  • M3–4 · per-bank statement adapters; DIOT draft
  • M5–6 · multi-client firm portal; recurring billing per RFC

06Founder

Marcel Lara. Automates hostile schemas for a living.

  • Data engineer at DeAcero, one of Mexico’s largest steelmakers: SAP to BigQuery ingestion, dbt models. Hostile-schema reconciliation as a day job.
  • Shipped an FDA ELT pipeline over 108,265 adverse-event reports (Airbyte · dbt · BigQuery · 27 automated tests · CI) and an applied ML risk model on the same data.
  • Previous startup APOLO (robotic surgical-lamp automation): patents filed at IMPI and USPTO, first author of a Springer publication, GSEA recognition.
  • Wrote 100% of Acuse’s code, solo, using Claude Code. Idea to live, real-data-validated product in a week, not months.
  • Also: published four books before writing reconciliation engines. Both jobs are keeping a thousand small facts consistent.
108,265

FDA adverse-event reports ingested

0.799
0 1.0

ROC-AUC, applied ML risk model

07The ask

Applying to YC Fall 2026

The RFS calls for AI-native compliance infrastructure. Acuse is the deterministic substrate an AI compliance agent needs, auditable math in the core, Claude at the messy edges, for the largest U.S. trading partner.

Acceptance is my resignation letter: full-time from day one. What YC compresses for a solo founder in Monterrey: the first 10 firms → 100 firms playbook, U.S. incorporation and the LatAm expansion.

Next milestone: 10 accounting firms running the report monthly