PesaBooks

1. One-liner

WhatsApp AI bookkeeper that turns M-Pesa transactions into eTIMS invoices and filed tax returns for Kenyan micro-businesses.

2. Trend signal — why now?

Three forces colliding simultaneously in Kenya’s SME landscape:

Regulatory hammer. Kenya’s eTIMS (Electronic Tax Invoice Management System) became mandatory for all businesses — not just VAT-registered ones — effective January 1, 2026. KRA now validates all income and expenses declared in tax returns directly against eTIMS data. Non-compliant invoices get disallowed as deductions, directly increasing the tax bill. Monthly Turnover Tax filing is due by the 20th, with KES 2,000/month penalties for individuals who miss it. The Kenya National Chamber of Commerce even pushed back on a proposed Sh2 million penalty for eTIMS non-compliance, signalling how aggressively enforcement is ramping up. Despite all this, only 18–20% of targeted businesses had registered for eTIMS by the original March 2024 deadline.

M-Pesa explosion. Safaricom celebrated 40 million M-PESA customers in Kenya in March 2026. The combined merchant base hit 2.4 million businesses accepting M-Pesa payments. Pochi la Biashara tills (informal traders) alone grew 81.7% year-on-year to over 1.15 million. Over 90% of SME transactions involve M-Pesa in some form — Paybill, Till, or direct transfer. Daraja 3.0, Safaricom’s most significant API upgrade since 2019, launched in November 2025, making programmatic access to transaction data meaningfully better.

Gap in tooling. KRA launched a Shuru WhatsApp chatbot for basic eTIMS invoice generation, and 60% of individual taxpayers now use it. But Shuru only generates invoices — it doesn’t reconcile M-Pesa transactions, track expenses, categorize income, or file returns. Meanwhile, web-based tools like Zoho Books, Cute Profit, and QuickBooks require business owners to learn accounting software — a non-starter for the kiosk owner or market trader who does everything on their phone. Pay Hero Kenya handles payment reconciliation but doesn’t close the loop to eTIMS compliance and tax filing.

Provenance:

• Signal 1: Kenya eTIMS mandatory for all businesses Jan 2026, KRA validates returns against eTIMS data, KES 2,000/mo penalties for non-filing — alphacap.co.ke — 2026-Q1
• Signal 2: M-Pesa merchant base hit 2.4M businesses (H1 FY26), Daraja 3.0 API launched Nov 2025 — safaricom.co.ke — 2025-11
• Signal 3: 80% of targeted businesses missed eTIMS deadline, existing tools are web-first and don’t handle M-Pesa reconciliation natively — the-star.co.ke — 2024-05 Category: Regulatory arbitrage

3. The opportunity

Kenya’s 2.4 million M-Pesa-accepting businesses just got dragged into a mandatory electronic invoicing regime, and the tools available to them are either too dumb (KRA’s Shuru — invoices only, no books) or too complex (Zoho, QuickBooks — require web dashboards and accounting knowledge). The gap is a chat-native AI bookkeeper that ingests M-Pesa transaction data, auto-categorizes it, generates eTIMS-compliant invoices, reconciles payments, and files Turnover Tax returns — all inside WhatsApp, zero accounting knowledge required.

The specific incumbent failure: Zoho Books has eTIMS integration but requires M-Pesa reconciliation workarounds through third-party integrators. Cute Profit has both but requires a web interface. KRA’s own Shuru bot can generate invoices but can’t match them to incoming payments, track expenses, or compute tax liability. None of these close the full loop for the informal micro-business owner who does everything on their phone.

4. Target market

• Primary customer: Kenyan micro-business owners — market traders, kiosk operators (duka), salon owners, mechanics, mama mbogas — who accept M-Pesa payments (Lipa na M-Pesa Till or Pochi la Biashara), earn KES 1M–25M/year, and are now required to comply with eTIMS and file Turnover Tax monthly.
• Why they buy: They face KES 2,000/month penalty for missing TOT filing (KES 24,000/year — painful for a business making KES 1–5M). They’re currently using notebooks or nothing. They’ve heard about eTIMS but find the web-based systems intimidating. They already live on WhatsApp.
• Rough TAM reasoning: 2.4 million M-Pesa merchants as of H1 FY26. Conservatively, 1.2 million fall in the KES 1M–25M turnover range subject to TOT. At KES 500/month, the theoretical TAM is KES 7.2 billion ($55M+). Realistically chasing the top 10% who feel the compliance pain most acutely — ~120,000 businesses, still a KES 720M ($5.5M) opportunity.
• Why now for them: Before January 2026, micro-businesses could fly under the radar. Now KRA cross-references eTIMS data against tax returns. No eTIMS invoice = expense disallowed = higher tax. The penalty clock is ticking monthly.

5. Product sketch (MVP)

• M-Pesa statement ingestion — connect via Daraja 3.0 API or forward M-Pesa SMS statements; AI auto-categorizes each transaction (sale, purchase, transfer, personal)
• eTIMS invoice generation — generate KRA-compliant invoices directly from WhatsApp, linked to M-Pesa payments received (closing the payment-to-invoice loop that Shuru can’t do)
• Expense tracking — snap a photo of a receipt via WhatsApp; AI extracts amount, vendor, category, and stores it
• Monthly P&L summary — WhatsApp message on the 15th of each month with plain-language income/expense breakdown in Kiswahili or English
• TOT auto-calculation and filing reminder — calculates 3% Turnover Tax liability, sends reminder before the 20th deadline, pre-fills iTax return data (manual filing for v1, auto-filing for v2)
• Audit-ready transaction log — all transactions timestamped, categorized, and exportable as PDF for KRA inspection
• USSD fallback — basic transaction logging and invoice generation via USSD for customers without smartphones/data

6. AI angle — what’s load-bearing

AI is doing the heavy lifting in three places:

1. Transaction categorization. M-Pesa statements are unstructured — a line says “JAMES KAMAU 0712…” and you have to figure out if that’s a customer payment, a supplier payment, or a personal transfer. The AI learns the business owner’s transaction patterns and auto-classifies with increasing accuracy over time. Without AI, this is manual bookkeeping — the exact thing micro-businesses refuse to do.

2. Receipt OCR and expense extraction. Photo of a receipt → structured data (amount, vendor, category, date). This eliminates the “shoebox of receipts” problem. Standard OCR + LLM extraction on Kenyan receipts with mixed English/Kiswahili text.

3. Natural language interface. Business owner texts “How much did I sell this week?” or “Nilipata pesa ngapi wiki hii?” and gets an answer. No dashboard, no charts, no learning curve. The conversational interface IS the product — remove the AI and you’re back to asking people to learn accounting software.

7. Localization angle

This is a Kenya-first play, purpose-built for the Kenyan business environment:

• M-Pesa native — not a bolt-on integration; M-Pesa transaction data is the primary ledger input
• eTIMS native — built around KRA’s specific compliance requirements, not generic tax software adapted for Kenya
• Kiswahili + Sheng support — conversational AI in the languages micro-businesses actually use, not just English
• KES pricing — KES 500/month (~$3.80) fits the budget of a kiosk owner earning KES 100K–400K/month
• WhatsApp-first — Kenya has 16M+ WhatsApp users; this is the primary business communication channel
• USSD fallback — for feature phone users who can’t use WhatsApp
• Expansion path — Tanzania (same M-Pesa ecosystem, similar tax reforms underway), Uganda, and eventually West Africa (with Mobile Money + local tax authority integrations)

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

• Pricing: KES 500/month ($3.80) for the base tier (up to 200 transactions/month). KES 1,500/month ($11.50) for the “Biashara” tier (unlimited transactions + auto-filing + multi-user). KES 3,000/month ($23) for “Fundi” tier (accountant dashboard + multi-business + advanced reports).
• ACV: KES 6,000–18,000/year ($46–$138). Blended ACV estimate: $70/year.
• Rough math to $1M ARR: ~14,300 paying businesses × $70/year = $1M ARR. That’s 0.6% of the 2.4M M-Pesa merchant base.
• Rough math to $5M ARR: ~71,400 businesses at $70 blended ACV, or 43,000 businesses with ACV expansion to ~$116 as usage grows and businesses upgrade tiers. That’s 1.8–3% of the merchant base.
• Expansion path: (a) Accountant/bookkeeper tier — charge accountants who manage multiple clients, (b) lending data — with consent, share anonymized financial data with mobile lenders for credit scoring (revenue share), (c) Tanzania/Uganda expansion with local tax authority integrations.

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

1. Safaricom M-Pesa merchant channels. Safaricom has a dedicated MSME division and regularly onboards merchants. Partner with Safaricom Business to be listed as a recommended eTIMS compliance tool during Pochi la Biashara and Lipa na M-Pesa onboarding. Even a small pilot with 500 merchants tests the channel.

2. KRA penalty fear — deadline marketing. Run WhatsApp broadcast campaigns and Facebook/Instagram ads targeting “eTIMS deadline” and “KRA penalty” keywords in Kiswahili and English. The message: “Usipatiwe fine ya KES 2,000 kila mwezi. Anza PesaBooks kwa WhatsApp — dakika 2.” (Don’t get fined KES 2,000 every month. Start PesaBooks on WhatsApp — 2 minutes.) Target the last week before the 20th of each month when the TOT deadline looms.

3. Market association partnerships. Kenya has organized market associations (Marikiti, Gikomba, etc.) with existing communication channels. Partner with 5 market associations to offer PesaBooks demos at monthly meetings. Each association has 200–2,000 members. Convert 5% = 50–500 customers per association.

4. Accountant referral network. There are thousands of small accounting practices in Kenya that service SMEs and charge KES 2,500–3,500/hour for bookkeeping. PesaBooks saves them time. Offer a “Fundi” tier with multi-client management and give accountants a 20% referral commission on clients they bring.

5. WhatsApp group seeding. Join 20–30 active Kenyan business WhatsApp groups (entrepreneur networks, market trader groups). Share eTIMS compliance tips as value-first content. Introduce PesaBooks as the solution. Each group has 100–256 members.

10. Build complexity — justification

Medium. Three non-trivial integrations required: (1) Safaricom Daraja 3.0 API for M-Pesa statement ingestion, (2) KRA eTIMS API (VSCU/OSCU) for invoice generation, and (3) WhatsApp Business API via a BSP like 360Dialog or Twilio. The AI layer (transaction categorization + receipt OCR) uses off-the-shelf LLM APIs and vision models. The USSD channel adds complexity but can be deferred to v1.1. A team of 2–3 engineers can ship a usable WhatsApp bot with M-Pesa + eTIMS integration in 10–14 weeks. The Kenya-specific regulatory knowledge (TOT calculations, eTIMS compliance rules) needs a domain advisor but isn’t technically complex.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Bookkeeping and tax preparation services are legal in Kenya. No special licensing required for software-based bookkeeping tools.
Ethical — no harm / dark patterns	✅	Helps businesses comply with tax law and avoid penalties. No dark patterns.
Market exists (evidence above)	✅	2.4M M-Pesa merchants, mandatory eTIMS compliance, KES 2K/mo penalties — demand is regulatory-driven.
1–5 person team can build this	✅	2–3 engineers + 1 Kenya-based ops person for partnerships. APIs are available.
Launchable with <$50K / ₹40L	✅	WhatsApp API costs are low (~$0.03/conversation), Daraja API is free, hosting on Hetzner/DO runs ~$50–100/mo. Total launch budget ~$15–25K.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Monthly KES 2,000 penalty + disallowed deductions. Businesses are getting fined now. Pain is acute, recurring, and has a dollar figure.
Demand evidence	15	13/15	eTIMS is law. 80% of businesses missed the deadline but are being forced to comply. 60% of taxpayers already use KRA’s WhatsApp bot — channel validated. 2.4M merchant base is growing 44% YoY.
Build feasibility	15	10/15	Three well-documented API integrations. AI layer is standard LLM + OCR. Medium complexity — a 2–3 person team ships in 10–14 weeks.
Distribution clarity	15	12/15	Named channels: Safaricom MSME partnerships, market associations, accountant referral, penalty-deadline Facebook/WhatsApp ads. Monthly penalty deadline creates natural urgency pulse.
Revenue mechanics	15	10/15	KES 500/mo is accessible but low ACV ($46/year). Need 14K+ customers for $1M ARR. Accountant tier and credit data upsell are realistic but unproven expansion levers.
Time to first revenue	10	8/10	MVP in 10–12 weeks, first paying customer within 2 weeks of launch. Monthly subscription = immediate recurring revenue.
Defensibility	10	7/10	Transaction data moat deepens with each month (AI learns per-business patterns). M-Pesa + eTIMS dual integration creates switching friction. WhatsApp-native positioning is a first-mover advantage in this specific market.
Total	100	77/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · sales-heavy

Technical for the API integrations (Daraja 3.0, eTIMS, WhatsApp Business) and AI categorization engine. Sales-heavy because the first 100 customers require on-the-ground partnerships with market associations, Safaricom MSME team, and accountant networks. A Kenya-based technical founder with business development experience is the ideal profile.

Key assumptions to validate (3–5)

1. Assumption: Micro-business owners (kiosk, market trader) will pay KES 500/month for an AI bookkeeper when they currently pay nothing. How to test: Run a 2-week WhatsApp ad campaign targeting “eTIMS penalty” keywords in Nairobi. Measure sign-up rate from 1,000 impressions. Need >5% click-through and >2% trial conversion to validate willingness to pay.

2. Assumption: M-Pesa transaction data via Daraja 3.0 is granular enough for AI to auto-categorize income vs. expenses accurately (>85% accuracy within 2 weeks of use). How to test: Get 10 M-Pesa merchant statement exports, manually label 500 transactions, train a categorization model, measure accuracy on held-out set.

3. Assumption: The eTIMS API (VSCU/OSCU) allows third-party software to generate invoices on behalf of a business without friction. How to test: Apply for eTIMS API access from KRA, build a prototype invoice generator, test with 5 real businesses.

4. Assumption: Market association partnerships can be secured within 4 weeks of outreach. How to test: Contact 10 market associations in Nairobi (Gikomba, Marikiti, Nyamakima). Secure at least 2 demo slots within 4 weeks.

Risk flags

1. [Platform dependency]: Heavy reliance on Safaricom’s Daraja API for M-Pesa data. If Safaricom changes API terms, pricing, or restricts third-party access, the product breaks. Mitigation: M-Pesa SMS statement forwarding as a backup ingestion method.

2. [Regulatory risk]: KRA could expand the Shuru bot’s capabilities to include reconciliation and tax filing, eating PesaBooks’ value proposition. Mitigation: Move faster, build deeper AI personalization and business intelligence features that a government bot won’t prioritize.

3. [Low ACV risk]: At $3.80/month, unit economics are thin. WhatsApp Business API costs (~$0.03/conversation, potentially 10–20 conversations/user/month) could eat into margins if usage is high. Mitigation: Careful conversation design to minimize API calls. Accountant tier at $23/month improves blended ACV.

4. [Market timing]: The compliance pain is real NOW but could subside if KRA enforcement becomes lax or if the government extends deadlines again. Mitigation: The product delivers genuine business value (cash flow visibility, expense tracking) beyond pure compliance — retention doesn’t depend solely on fear of penalties.

14. Structured verdict

Score:                  77/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Kenya-based technical founder with M-Pesa integration experience and local business networks
Time to revenue:        10–14 weeks to MVP, first paying customer within 2 weeks of launch
Capital to launch:      KES 2–3.5M ($15–25K)
Top 3 assumptions to validate first:
  1. Micro-businesses will pay KES 500/mo — test with WhatsApp ad campaign targeting eTIMS penalty keywords, need >2% trial-to-paid conversion
  2. M-Pesa Daraja 3.0 transaction data supports accurate AI categorization — test with 10 merchant statement exports, need >85% accuracy
  3. eTIMS API allows frictionless third-party invoice generation — prototype and test with 5 real businesses
Kill criteria:
  - Abandon if <3% of 500 WhatsApp ad respondents convert to paid after a 14-day free trial
  - Abandon if M-Pesa transaction categorization accuracy stays below 80% after 30 days of per-business learning
  - Abandon if KRA restricts or prices out third-party eTIMS API access for small software vendors

15. Next step — 1-week validation sprint

• Day 1–2: Apply for KRA eTIMS API (VSCU) access and Safaricom Daraja 3.0 sandbox. Simultaneously, collect 5 M-Pesa merchant transaction statements (ask friends/contacts running small businesses in Nairobi). Manually label 200 transactions into categories (sales income, supplier payments, personal transfers, fees).
• Day 3–4: Build a prototype transaction categorizer using GPT-4o-mini on the labeled dataset. Measure accuracy. Simultaneously, set up a WhatsApp Business API sandbox and build a basic conversational flow: “Forward your M-Pesa statement → see categorized transactions → generate an eTIMS invoice.” Run Facebook ads targeting “eTIMS penalty” and “KRA fine” in Kiswahili for KES 5,000 ($38), pointing to WhatsApp.
• Day 5: Count (a) how many people messaged the WhatsApp bot from the ad, (b) how many forwarded a real M-Pesa statement, (c) how many said they’d pay KES 500/month. Decide go/no-go based on: ≥50 WhatsApp messages from KES 5,000 ad spend AND ≥10 people who forwarded a real statement AND ≥5 people who said “yes” to KES 500/month. If those numbers hit, the demand is real and the channel works. If not, the idea stays as a VALIDATE and needs more research.