SelfBill — self-billing catcher for Malaysian SMEs

1. One-liner

Reads an SME’s spend and ad accounts, flags every payment owed a self-billed e-invoice, and drafts it for MyInvois.

2. Trend signal — why now?

Malaysia’s MyInvois e-invoice mandate hit Phase 4 (businesses RM1M–RM5M turnover) on 1 January 2026, with full penalty enforcement starting 1 January 2027. Every accounting vendor in the country is selling “submit your sales invoice to MyInvois” software — that market is a red ocean (Bukku gives it away free, SQL, AutoCount, QNE, Financio, JomeInvoice all compete on it).

But almost nobody is selling against the blind spot: the self-billed e-invoice. When an SME pays a foreign supplier (Google/Facebook Ads, AWS, overseas vendors), an agent/dealer/distributor commission, an individual non-business seller, a non-business landlord, or an e-commerce platform fee, the buyer — not the seller — is legally required to generate and validate a self-billed e-invoice. Miss it and (a) the expense deduction gets disallowed on audit, and (b) it’s a separate criminal offence under S.82C(6) of the Income Tax Act, RM200–RM20,000 per invoice plus up to 6 months jail.

SMEs focus on the invoices they send and forget the ones they’re supposed to issue to themselves. In February 2026 LHDN flagged 500,000+ non-compliant cases and RM14 billion in unreported income and is actively enforcing. Agent/dealer commissions and individual-payee transactions cannot be consolidated — every single one needs its own e-invoice, even ten commissions to the same agent in one month. That’s a recurring, high-frequency, criminal-liability blind spot with no focused tool aimed at it.

Provenance:
  - Signal 1 (Demand): Self-billing is a separate S.82C(6) criminal offence SMEs overlook; missing the UIN-cleared self-billed doc gets the deduction disallowed on audit — https://www.ktp.com.my/blog/cp58-common-mistakes/25feb2026 — 2026-02-25; 10 distinct mandatory scenarios, commissions can't be consolidated — https://jomeinvoice.my/self-billed-e-invoice-malaysia-scenarios/ — 2026
  - Signal 2 (Feasibility): MyInvois open API + cheap LLM classification of bank/expense/ad-account feeds makes auto-detection + draft generation buildable on off-the-shelf stack; imported-services self-billing (Google/Facebook Ads) is now a documented, repeatable pattern — https://jomeinvoice.my/self-billed-e-invoice-imported-services/ — 2026
  - Signal 3 (Economic): LHDN flagged 500,000+ non-compliant cases / RM14B unreported, active enforcement, RM200–20,000/invoice penalties; crowded well-funded e-invoice vendor market + MSME Digital Grant subsidies = money moving — https://jomeinvoice.my/article/lhdn-e-invoice-penalty-malaysia-what-sme-owners-must-know-in-2026/ — 2026
  Category: Regulatory arbitrage

3. The opportunity

The e-invoice incumbents (SQL, AutoCount, QNE, Financio, Bukku, JomeInvoice, Refrens) all treat self-billing as a feature you operate: a form you fill once you’ve already decided a transaction needs a self-billed e-invoice. They generate it; they don’t tell you it’s owed. The hard part isn’t generating the JSON and getting a UIN — it’s knowing which of your hundreds of monthly payments are self-billing triggers when the rules span 10 scenarios with edge cases (employee commission = Form E, not self-billed; foreign ads = yes; individual landlord = yes; registered-business landlord = no).

SelfBill flips the model: instead of “fill this form,” it’s “we read your money going out and tell you what you missed.” It connects to the SME’s bank feed / expense exports / Google & Meta Ads accounts, classifies each outflow against the self-billing ruleset, surfaces a worklist (“these 14 payments need a self-billed e-invoice this month — 9 are auto-draftable”), and pushes the validated documents to MyInvois. It’s a miss-detector with a generator attached, not a generator hoping you noticed.

4. Target market

• Primary customer: Owner/finance person at a Malaysian SME with RM1M–RM5M turnover (Phase 4), especially ones that run paid ads (foreign imported services), pay sales agents/dealers/distributors on commission, or rent from individual landlords. Also the outsourced bookkeeping / SST agent firms that service 30–200 such SMEs each.
• Why they buy: “I file my sales e-invoices fine. I have no idea if I’m catching all the self-billed ones, and I just read it’s a criminal offence with jail time.” The fear is asymmetric — the downside is disallowed deductions plus prosecution, the cost of insurance is small.
• Rough TAM reasoning: Malaysia has ~1M+ SMEs; the RM1M–RM5M Phase 4 band plus Phases 1–3 already live is a multi-hundred-thousand population. Realistically winnable beachhead: the subset that pays foreign ads + commissions (ad-spending SMEs and any business with a sales force) — tens of thousands — plus the ~10,000+ bookkeeping/tax-agent firms who each multiply reach.
• Why now for them: Phase 4 went live Jan 2026; enforcement bites Jan 2027. The 2026 calendar year is the exact window where every Phase 4 SME is being told to get compliant and is discovering self-billing is the part they don’t understand.

5. Product sketch (MVP)

• Connect bank transaction feed (CSV/statement import to start; Open Banking later) + expense exports from accounting software (SQL/AutoCount/Bukku CSV).
• One-click connect to Google Ads & Meta Ads accounts to catch imported-services self-billing automatically.
• AI classifier tags each outflow: “self-billed e-invoice required / not required / needs review,” with a plain-language reason and the matching LHDN scenario.
• Monthly worklist dashboard: “X self-billed e-invoices owed this period, Y auto-draftable, Z need a TIN/MyKad from the payee.”
• Auto-draft the self-billed e-invoice (correct fields, MSIC/classification codes, supplier details) and submit to MyInvois for validation → store the UIN-cleared document.
• Payee-detail collector: a shareable link/WhatsApp flow to request TIN/MyKad/passport + SST number from agents and individual sellers, so the SME isn’t chasing them manually.
• Audit binder: 7-year retrievable archive of every self-billed e-invoice + the evidence trail, exportable for an LHDN review.

6. AI angle — what’s load-bearing

The classification is the product. A payment of “RM4,200 to Meta Platforms Ireland” must be recognised as imported advertising services → self-billed required; “RM1,800 to Ahmad bin Ismail” must be triaged (individual seller? commission to a registered dealer? employee allowance under Form E? landlord?), and the model has to ask the one disambiguating question rather than 10. Off-the-shelf rules-only engines drown in edge cases; an LLM reading transaction descriptions, payee names, ad-platform metadata, and prior labels — improving from the SME’s corrections — is what turns a 2-hour monthly reconciliation into a 10-minute review. Remove the AI and you’re back to a form the user has to know to fill. The generation/submission half is commodity (MyInvois API); the detection half is the moat and is genuinely AI-shaped.

7. Localization angle

This is a Malaysia-first regulatory play by definition — MyInvois, LHDN scenarios, MSIC codes, SST numbers, MyKad/passport capture, RM pricing. Bahasa Malaysia + English UI and WhatsApp-based payee-detail collection (the dominant SME channel in Malaysia). The same shape (buyer-side self-billing detection) ports to other mandatory e-invoicing regimes with reverse-charge / self-billing rules — Philippines EIS, India GST reverse charge, EU ViDA — but each needs its own ruleset, so treat Malaysia as the wedge, not a global v1.

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

• Pricing: RM149/mo ($32) for a single SME (Starter); RM349/mo ($74) for ad-heavy/commission-heavy SMEs (more sources, higher volume). Agent/bookkeeping firm plans at RM699–RM1,999/mo (~$150–$425) for 25–150 client books — this is the real engine.
• ACV: ~RM2,400/yr direct SME; ~RM12,000/yr per bookkeeping firm.
• Rough math to $1M ARR (~RM4.7M): e.g. 250 bookkeeping firms × RM12,000 = RM3M + 700 direct SMEs × RM2,400 = RM1.68M ≈ RM4.7M. Or ~1,950 direct SMEs alone.
• Rough math to $5M ARR: Needs the agent-firm channel to compound (each firm a multiplier) plus expansion into Phases 1–3 mid-market and adjacent compliance modules (SST return prep, withholding-tax/CP58 cross-checks). ~900 firms + 3,000 SMEs gets there.
• Expansion path: Per-document overage on high-volume commission payers; add SST and withholding-tax (CP58) modules to the same outflow feed you already ingest; eventually buyer-side reconciliation of all incoming supplier e-invoices for input-cost matching.

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

• Bookkeeping / SST-agent firms first. There’s a finite, listable population of LHDN tax agents and accounting firms in Malaysia (directories + LinkedIn + the e-invoice vendor partner lists). Cold email/WhatsApp 500 of them with a one-page “here’s the self-billing liability your clients are missing, and the S.82C(6) jail clause” + a Loom showing the worklist. These firms feel the pain across dozens of clients and resell instantly. Target 5–8% to a paid pilot.
• Ride the fear content. Every Malaysian accounting blog (KTP, JomeInvoice, L&Co, BDO) is publishing self-billing explainers right now. Publish the definitive “self-billing scenarios checklist” + free self-audit tool (upload a month of bank statements → count of missed self-billed e-invoices), gated on email. The free tool is the demo.
• Google/Meta Ads angle as a sharp hook. Every SME running foreign ads owes monthly self-billed e-invoices and most don’t. Target “Google Ads agency Malaysia” / “digital marketing agency” partners who can offer it to their clients as a value-add.
• MSME Digital Grant co-marketing. Position SelfBill as a grant-eligible digital tool (50% subsidy up to RM5,000) — halves the effective price and gives a reason to buy this quarter.

10. Build complexity — justification

Medium. The MyInvois submission API, bank-statement parsing, and ad-account read APIs are all off-the-shelf; the classification layer is an LLM with a curated ruleset and a correction loop, not a custom model. Real work is in encoding the 10 self-billing scenarios + edge cases correctly (this needs a Malaysian tax advisor, not just an engineer) and in clean ingestion across messy bank/accounting exports. A 2-person team (one technical, one with MY tax/e-invoice domain depth) ships a credible v1 in ~3–4 months.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helps comply with a mandate; no regulated activity by us
Ethical — no harm / dark patterns	✅	Reduces under-reporting and penalties; pro-compliance
Market exists (evidence above)	✅	Mandate live, criminal penalties, 500K+ flagged cases
1–5 person team can build this	✅	2 people, ~3–4 months
Launchable with <$50K / ₹40L	✅	Off-the-shelf APIs + LLM inference; main cost is the tax advisor

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Criminal liability + disallowed deductions + recurring monthly + a genuine blind spot. Not quite 17+ because awareness is still rising — some SMEs don’t yet feel the pain.
Demand evidence	15	11/15	Strong regulatory + expert-source evidence and a crowded paid vendor market; but direct first-person SME complaint quotes are thin in indexed sources, so not 13+.
Build feasibility	15	11/15	Off-the-shelf APIs; the ruleset encoding + messy ingestion is the real work. Pair in 3–4 months.
Distribution clarity	15	11/15	Bookkeeping-firm channel is a named, listable population with a multiplier; conversion still unproven.
Revenue mechanics	15	12/15	Pricing benchmarked against MY e-invoice tools (~RM300/mo norm); agent-firm ACV credible. Direct-SME churn risk after enforcement settles.
Time to first revenue	10	8/10	Sellable now in the 2026 compliance window; pilot-to-paid likely 4–8 weeks via firms.
Defensibility	10	5/10	Execution + accumulating classification corrections + agent-firm lock-in. Incumbents could bolt on detection — head start and focus are the edge, not a hard moat.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

domain-expertise-required · technical-heavy — needs a Malaysian e-invoice/tax advisor co-founder or close advisor; the ruleset correctness is the product, and getting it wrong is a liability.

Key assumptions to validate (3–5)

1. Assumption: SMEs and their bookkeepers genuinely miss self-billed e-invoices today (not already handled by their accounting software). How to test: Run the free self-audit on 20 real SMEs’ bank statements; measure how many missed self-billed e-invoices surface per SME.
2. Assumption: Bookkeeping/SST-agent firms will pay a per-firm subscription to cover all clients. How to test: 30 cold outreach calls to firms; target ≥6 willing to start a paid pilot.
3. Assumption: The classifier can reach usable precision on real transaction feeds without endless manual review. How to test: Label 1,000 real outflows; measure auto-classification accuracy and how many need the disambiguating question.
4. Assumption: Penalty fear converts to purchase in 2026 (not deferred to late 2026 when enforcement bites). How to test: Track pilot close timing vs. enforcement-deadline messaging.

Risk flags

1. Regulatory risk: LHDN keeps revising scenarios, thresholds, and the consolidation/RM10,000 rules — the ruleset needs constant maintenance, and a wrong rule creates customer liability.
2. Platform dependency: Built on the MyInvois API, Google/Meta Ads APIs, and bank-statement formats — all can change.
3. Market timing / incumbent encroachment: The full-suite accounting vendors own the customer relationship and could add self-billing detection; the window to win the niche standalone is ~12–18 months.
4. Awareness risk: If SMEs stay unaware until enforcement starts, sales cycles drag into late 2026.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + Malaysian e-invoice/tax advisor co-founder
Time to revenue:        4–8 weeks via bookkeeping-firm pilots
Capital to launch:      RM40–80K / $9–17K (advisor + inference + ingestion build)
Top 3 assumptions to validate first:
  1. SMEs actually miss self-billed e-invoices today — free self-audit on 20 SMEs' statements
  2. Bookkeeping firms will pay per-firm — 30 cold calls, target 6 paid pilots
  3. Classifier hits usable precision — label 1,000 real outflows, measure accuracy
Kill criteria:
  - Abandon if the free self-audit surfaces <2 missed self-billed e-invoices/month per SME on average (no real blind spot)
  - Abandon if <4 of 30 bookkeeping firms agree to a paid pilot after a working demo
  - Abandon if a full-suite incumbent ships credible self-billing detection before your v1

15. Next step — 1-week validation sprint

• Day 1–2: Build the manual “self-audit” — take one month of real bank statements + a Google Ads export from 3 friendly SMEs, classify by hand against the 10 scenarios, count missed self-billed e-invoices. This is the falsifiable core: is there a blind spot or not?
• Day 3–4: Cold outreach to 30 Malaysian bookkeeping/SST-agent firms with the audit findings and a mock worklist screenshot. Ask for a paid pilot, not a “would you use it.”
• Day 5: Decide go/no-go. Go only if the self-audit averages ≥2 missed self-billed e-invoices per SME per month AND ≥6 firms verbally commit to a paid pilot. Anything less and the blind spot or the willingness-to-pay isn’t real.