SB StartupBasket
All ideas
72 /100 GO Medium complexity

FechaDoc — document chaser for Brazil's accounting firms

AI WhatsApp agent that collects, checks, and reconciles every client's monthly docs so contadores stop nagging by hand.

ProfessionalServices SaaS LATAM WhatsApp-firstAI-agentPortugueseCompliance-drivenSMBWorkflow-automation
— views
Evaluation Scores
72/100

GO

Overall Score

16
Problem
12
Demand
11
Build
11
Distrib.
12
Revenue
7
Time
5
Defense

FechaDoc — document chaser for Brazil’s accounting firms

1. One-liner

AI WhatsApp agent that collects, checks, and reconciles every client’s monthly docs so contadores stop nagging by hand.

2. Trend signal — why now?

Three things are converging in Brazil right now.

First, the pain is loud and documented. Brazilian accounting blogs name document-chasing — the follow-up (they literally call it “FUP”) — as the task that “seems harmless but steals time and productivity from your day.” A firm with 80 clients burns 6–10 hours every month just sending reminders (Pasta Contábil). The relationship cost is worse than the time cost: “a cobrança manual desgasta a relação com o cliente… o contador vira cobrador enquanto o cliente começa a evitar o contato” — the accountant becomes a debt collector while the client starts dodging contact. On the client side, the #1 complaint about accounting firms is offices “that only function through constant billing follow-ups” (Nibo, Roberto Dias Duarte).

Second, the channel and the AI both arrived. WhatsApp is the business channel in Brazil, the Business API is mature, and a Portuguese-fluent LLM that can read an uploaded bank statement or nota fiscal and decide whether it’s complete now costs cents per client per month. The chasing-and-checking loop that used to need a human can now run agentically.

Third, money is moving and the deadline pressure is spiking. LATAM accounting software is growing ~10.9% CAGR (2026–2035, Expert Market Research); funded incumbents (Nibo, ContaAzul, Contabilizei) all sell into this base; and Brazil’s 2026 tax reform (CBS/IBS test rates begin 2026) raises the volume and stakes of monthly document/data collection for every one of the country’s 103,176 accounting firms (CFC, Mar 2026).

Provenance:

3. The opportunity

There is already a category here — and that’s the point. Tools like Pasta Contábil and Ottimizza sell “automatic document collection”: a client portal plus a scheduler that fires “please upload your docs” links on a calendar (day 1, day 5, day 10). Nibo Docs sells storage — a cloud folder. They’ve automated the reminder. They have not automated the judgment.

The judgment is the expensive part. When a client uploads something, a human at the firm still has to open it and ask: is this the bank statement for the whole month or just two weeks? Are all the notas fiscais here or is the firm short three? Is this even the right CNPJ? Did this client have payroll this month, or am I about to nag them for a folha that doesn’t exist? A dumb scheduler can’t answer any of that, so it either nags clients for things they already sent (the relationship-killer the blogs complain about) or lets gaps through to fechamento.

FechaDoc replaces the judgment, not just the calendar. The agent ingests whatever lands in the WhatsApp thread, reads it, checks it against what this specific client should owe this specific month, and only escalates a real gap — and when the client asks “which statement do you need?” it answers in the same thread. The incumbents are a megaphone; this is a junior assistant that actually reads.

4. Target market

  • Primary customer: Owner or operations lead of a small-to-mid Brazilian escritório de contabilidade — 30–300 client companies (CNPJs), 2–15 staff. Concentrated in São Paulo (28% of the 103,176 firms) but present in every state.
  • Why they buy: Their staff lose 6–10+ hours per 80 clients every month to manual chasing and document review, fechamento deadlines slip when clients deliver late or incomplete, and the nagging actively damages client relationships. They feel this every single month.
  • Rough TAM reasoning: 103,176 firms. Even a conservative serviceable slice — the ~30–40k firms organized enough to adopt SaaS and large enough to feel the pain (10+ clients) — at R$300–600/mo is a R$1B+/yr addressable line. Capturing 2,000 firms is a healthy bootstrapped business.
  • Why now for them: WhatsApp is where their clients already are; 2026 tax reform is increasing both document volume and deadline anxiety; and the “half an intern” ROI framing is now obvious enough that firms are actively shopping for collection tools (multiple competing products launched in 2025–2026).

5. Product sketch (MVP)

  • Per-client monthly checklist auto-built from the client’s regime (Simples / Lucro Presumido / Lucro Real), activities, and last month’s actuals — not a one-size template.
  • WhatsApp-native intake: client sends docs straight into the thread or a no-login link; agent accepts PDFs, photos, XML notas fiscais, OFX statements.
  • Read-and-verify: agent classifies each incoming file, checks the statement covers the full period and the CNPJ matches, and counts notas against what’s expected — flags partial or wrong uploads instead of silently accepting them.
  • Smart chasing: only nags for what’s actually still missing for that client this month; suppresses reminders the moment the item arrives; escalates tone on a configurable cadence (day 1 / 5 / 10).
  • Conversational replies: answers the client’s “which one do you need / for what month?” questions in Portuguese inside the same thread.
  • Firm dashboard: one screen showing, across all clients, exactly who is complete, who is partial, who is silent — ranked by deadline risk.
  • Handoff to the firm’s stack: verified docs exported/organized so the team can post them (integrations to ContaAzul/Nibo folders as a fast-follow, not v1).

6. AI angle — what’s load-bearing

Remove the AI and you have Pasta Contábil — a scheduler that fires reminders and a portal that holds files. The defensible work is reasoning: deciding the correct checklist for each client this month (regime + activity + history), reading a messy phone-photo of a bank statement or an XML batch and judging completeness/correctness, and conversing naturally enough that clients answer the agent instead of dodging it. That’s three LLM-load-bearing jobs the calendar-based incumbents structurally don’t do. Without the AI, there is no product — there’s just the competitor.

7. Localization angle

This is a Brazil-first play by construction, and the localization is the moat: Portuguese conversation, WhatsApp as the primary surface (not email), Brazilian document types (nota fiscal eletrônica XML, OFX, DAS/Simples docs), tax regimes (Simples Nacional / Lucro Presumido / Lucro Real), and the 2026 CBS/IBS transition. A generic global “client document portal” cannot read a nota fiscal or reason about Simples Nacional. Pricing in reais (R$ tiers) lands where a $/seat global tool can’t. Adjacent expansion later: same playbook for Portugal, then Spanish-speaking LATAM with CFDI (Mexico) — but Brazil alone clears the bar.

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

  • Pricing: tiered by client count, benchmarked to incumbent R$99–R$499/mo band. Starter R$199/mo (up to 30 clients), Pro R$499/mo (up to 120), Scale R$999/mo (up to 300+). Sits above dumb-portal pricing because it replaces review labor, not just reminders.
  • ACV: ~R$4,800–6,000/yr blended (≈ $900–1,100).
  • Math to R$6M ARR ($1.1M): ~1,000 firms × R$500/mo × 12. Out of 103,176 firms, that’s <1% penetration.
  • Math to R$28M ARR ($5M): ~4,500 firms at blended R$500/mo, OR ~2,500 firms after ACV grows via per-client overage + stack integrations. Still low-single-digit % of the market.
  • Expansion path: overage as the firm onboards more clients (usage-aligned), upsell to a per-client “deliver to ContaAzul/Nibo” automation tier, and a per-client white-label client app the firm resells to its own clients.

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

  • CRC + association directories: CRCSP and regional councils publish firm rosters; accounting associations (e.g. SESCON-SP, Fenacon) have member lists. Build a list of 2,000 SP firms with 30+ clients, send a 60-second Loom (in PT) showing the agent catching an incomplete bank statement; expect 3–5% reply.
  • Contador influencer/creator channels: Brazil has a large accounting-creator scene on YouTube/Instagram (e.g. Roberto Dias Duarte’s audience, Jornal Contábil). Sponsor/partner 3–5 of them for a “stop being a cobrador” demo; their audience is exactly the buyer.
  • Land via one painful client: offer a free 30-day pilot scoped to a firm’s 10 worst late-delivering clients — the ones that blow up fechamento. Saving those 10 sells the whole roster. Falsifiable conversion test built into the wedge.
  • Pasta Contábil / Ottimizza switchers: target firms already paying for a dumb portal (they’ve admitted the budget and the pain) with “yours reminds; ours reads and checks.”

10. Build complexity — justification

Medium. Off-the-shelf: WhatsApp Business API, an LLM for classification/extraction/conversation, standard web dashboard, file storage. Custom work is the judgment layer — per-regime checklist logic, nota fiscal XML / OFX parsing and completeness checks, and a chase state-machine that suppresses on arrival. No novel models, no hardware, no marketplace. A 2–3 person team (one strong backend + one with Brazilian accounting domain knowledge) ships a credible v1 in ~3–4 months.

11. Gating checklist

GatePass?Note
Legal in target market✅Document collection on behalf of a firm; WhatsApp Business API ToS-compliant; LGPD applies — handle as data processor with consent, standard.
Ethical — no harm / dark patterns✅Reduces nagging, doesn’t manufacture it; opt-out respected.
Market exists (evidence above)✅Funded incumbents (Pasta Contábil, Ottimizza, Nibo) already charge for the adjacent product.
1–5 person team can build this✅2–3 people, ~3–4 months.
Launchable with <$50K / ₹40L✅API + inference + design; well under cap.

12. Feasibility score

AxisWeightScoreNotes
Problem intensity2016/20Monthly, every client, both time-cost and relationship-cost. Documented, not hypothesized. Just shy of hair-on-fire because firms have limped along with manual FUP.
Demand evidence1512/15Multiple independent BR sources naming the exact pain + paid incumbents in the adjacent slot. Strong, but no proof yet that firms pay a premium for the read-and-verify upgrade.
Build feasibility1511/15Standard stack; the nota fiscal / OFX completeness logic and checklist-per-regime are the real work. ~3–4 months.
Distribution clarity1511/15Named directories + creator channels + “worst 10 clients” pilot. Conversion math plausible but unproven; PT cold outreach is the variable.
Revenue mechanics1512/15Pricing benchmarked to a real incumbent band; <1% penetration to $1M. Premium positioning is the assumption to prove.
Time to first revenue107/10Pilot-to-paid in 4–8 weeks; needs a 30-day pilot to prove the read-and-verify value.
Defensibility105/10Execution + accounting-domain checklist data + workflow lock-in once clients are trained to use the thread. Incumbents could bolt on verification, so head-start and depth matter.
Total10072/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required — needs solid LLM/integration engineering and someone who genuinely understands Brazilian fechamento and tax regimes. A Brazilian accountant co-founder is close to mandatory.

Key assumptions to validate (3–5)

  1. Assumption: Firms will pay a premium over dumb-portal tools (R$499 vs R$99–199) for read-and-verify. How to test: Price-test the Pro tier against 30 firms already paying for Pasta Contábil/Ottimizza; measure willingness at the higher price.
  2. Assumption: The agent can read real-world client uploads (phone photos, partial statements, mixed XML) accurately enough to be trusted. How to test: Run 500 historical client uploads from 3 pilot firms through the verifier; measure precision/recall on “complete vs incomplete.”
  3. Assumption: Clients will engage with an AI agent in WhatsApp rather than ignore it. How to test: Pilot on the “worst 10 clients” at 3 firms; measure on-time, complete delivery rate vs the prior 3 months.
  4. Assumption: Cold outreach to CRC-listed firms converts at 3–5%. How to test: 200-firm Loom campaign, measure reply and pilot-start rate.

Risk flags

  1. Incumbent fast-follow: Pasta Contábil/Ottimizza/Nibo could add LLM verification. Mitigation: go deep on regime-specific checklist intelligence and client-facing conversation quality faster than a feature team bolts it on.
  2. Platform dependency: WhatsApp Business API ToS and pricing changes could squeeze margins or limit automated messaging. Mitigation: email/portal fallback channel; stay within template/session rules.
  3. Accuracy liability: A false “you’re complete” that causes a missed fechamento deadline burns trust fast. Mitigation: conservative flagging, human-in-the-loop confirmation for the firm before close, clear “agent-assisted, firm-approved” framing.
  4. LGPD: Handling clients’ financial documents at scale. Mitigation: data-processor agreements, consent flows, retention controls — table stakes, not a blocker.

14. Structured verdict

Score:                  72/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + Brazilian accountant co-founder (domain)
Time to revenue:        6–10 weeks (pilot-to-paid)
Capital to launch:      R$60–120k ($11–22K)
Top 3 assumptions to validate first:
  1. Premium WTP over dumb portals — price-test 30 firms already paying for collection tools
  2. Read-and-verify accuracy on messy real uploads — 500 historical docs through the verifier, measure precision/recall
  3. Clients engage the agent — "worst 10 clients" pilot at 3 firms, measure on-time complete-delivery lift
Kill criteria:
  - Abandon if <15% of 50 piloted firms convert to paid after a 30-day pilot
  - Abandon if verifier precision on "complete" can't clear ~90% on real client uploads (false-complete is the trust-killer)
  - Abandon if a major incumbent ships equivalent LLM verification before your v1 and bundles it free

15. Next step — 1-week validation sprint

  • Day 1–2: Pull CRCSP / SESCON-SP member lists; assemble 50 SP firms with 30+ clients. Recruit one friendly pilot firm and collect ~100 real historical client uploads (anonymized).
  • Day 3–4: Build a thin prototype of just the verifier — feed the 100 uploads, measure how often it correctly calls “complete vs incomplete vs wrong CNPJ.” Record a 60-second PT Loom of it catching a half-month bank statement.
  • Day 5: Send the Loom to the 50 firms; book pilot calls. Go/no-go: ≥6 firms (12%) request a pilot AND verifier precision on “complete” ≥85% on the 100-doc set. Below either → rework or pass.

The result is falsifiable: a real precision number on real documents, and a real reply rate from real firms — not “people seemed interested.”

Interested in a detailed proposal?

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

Contact us

info@startupbasket.ai