ACE Technologies Limited · Nigerian C-Corp · Delaware flip option post-Series A
ACE is an invisible operating system for informal commerce. We turn WhatsApp chaos into autonomous businesses through radical AI automation. Merchants keep texting customers on WhatsApp — our AI handles inventory, payments, negotiations, logistics, and customer retention completely in the background.
No new apps for customers. No manual work for merchants. Just autonomous growth.
In Nigeria and across emerging markets, $2.3 trillion in commerce happens entirely on WhatsApp. ~480,000 merchants in Lagos alone manage 30–150 customer conversations daily through unstructured text, voice notes in Pidgin, and screenshots of bank transfers. These merchants are growing 40% YoY but hit a hard ceiling at $50K annual revenue — they can't scale past their personal bandwidth.
| Current "Solution" | Why It Fails |
|---|---|
| WhatsApp Business App | Zero state management, no payment integration, no inventory tracking |
| Respond.io / Zoko / Hilos | Routes chats to human agents — still requires merchants to check bank apps, call riders, update inventory manually |
| Shopify / WooCommerce | Requires customers to leave WhatsApp → 78% conversion drop in emerging markets |
| Traditional ERP (Zoho, Odoo) | Desktop-oriented, requires structured data input — alien to merchants who think in conversations |
The core insight: The merchant's current system isn't broken — it's optimised for n=1. WhatsApp works perfectly for one customer. It catastrophically fails at 50 concurrent conversations. The merchant doesn't need a better inbox. They need an autonomous operating system that executes business logic while they sleep.
| ❌ What We're NOT Building | ✅ What We ARE Building |
|---|---|
| An AI chatbot that drafts responses for merchants to review | An autonomous OS that executes |
| A fancy CRM with "AI-powered insights" | An event-driven engine with no manual steps |
| Another unified inbox with message routing | Invisible back-office infrastructure |
| A tool requiring customers to change behaviour | Invisible layer on existing WhatsApp behaviour |
The AI doesn't ask permission. It executes complex multi-step business logic and reports outcomes. The merchant's job shifts from operational execution to strategic exception management.
┌──────────────────────────────────────────────────────────────────┐
│ LAYER 3: ENTERPRISE INTELLIGENCE PLATFORM (B2B) │
│ AI Training Data · FMCG Market Pulse · ACE TrustScore API │
│ The "Scale AI of Informal Commerce" — $16.5M ARR by Year 2 │
└───────────────────────────────┬──────────────────────────────────┘
│ (Refined data → enterprise buyers)
┌───────────────────────────────▼──────────────────────────────────┐
│ LAYER 2: AUTONOMOUS STATE ENGINE (Core IP) │
│ 11 microservices: Rust core + Vercel AI SDK agent layer │
│ Ingestion → Identity → AI Negotiator → State Machine → │
│ Payment → Logistics → Supplier → Visual Context → Comms Router │
│ Every interaction: data collected, refined, monetised │
└───────────────────────────────┬──────────────────────────────────┘
│ (Raw commerce events)
┌───────────────────────────────▼──────────────────────────────────┐
│ LAYER 1: INTERFACE LAYER │
│ Merchant: React Native app (exception dashboard) │
│ Customer: WhatsApp → PWA Trojan Horse → Global Buyer ID │
│ Vendor Communiqué: SMS reply-code for merchant decisions │
└──────────────────────────────────────────────────────────────────┘
ACE automates the entire commercial lifecycle — from first customer message to post-sale retention. The merchant interacts only with exceptions.
| Workflow | What ACE Does | Merchant Input |
|---|---|---|
| Order Fulfillment | Customer texts → AI parses intent (Whisper → BERT → GPT-4o) → AI Negotiator closes deal → virtual account issued → payment verified → rider dispatched → merchant notified | Zero |
| Demand-Driven Restocking | Inventory Oracle detects stockout (16hr lead) → pings supplier WhatsApp → negotiates price → margin analysis → drafts PO | 1 tap (8 seconds) |
| Customer Retention Engine | Nightly analyser detects at-risk VIP → generates culturally-nuanced message → sends after 4hr hold if no merchant action | Optional review |
| Multimodal Visual Resolution | "That blue dress in your last reel" → CLIP embeddings → SKU resolved → price negotiated autonomously | Zero |
ACE doesn't apply discounts. It negotiates — like a skilled market trader who knows the customer's full history, operates in their dialect, and closes deals autonomously within merchant-defined boundaries.
The negotiation arc: Anchor → Acknowledge → Counter → Close / Pivot / Escalate
6 autonomous tactics: Relationship Anchor · Bundle Pivot · Inventory-Verified Scarcity · Future Credit · Urgency Window · Sentiment-Aware Soft Close
The Rust circuit breaker: The AI is physically prevented from closing below the merchant's floor price. On below-floor requests: Bundle Pivot → Future Credit → Vendor Communiqué SMS to merchant.
Every negotiation is a data asset: NegotiationTrace logs price elasticity per SKU, per geography, per customer tier → sold to FMCG brands as market intelligence.
ACE keeps merchants in control without requiring them to be at a dashboard.
Exception detected → Channel selected by urgency:
> ₦50K order dispute → AI voice call
Pricing exception → SMS: "Amaka wants dress at ₦12K (floor: ₦14,250). Reply 1-approve, 2-hold, 3-bundle"
Restock approval → SMS: "Red Ankara running out. Alhaji: ₦42K for 50yds (44% margin). Reply 1 to approve."
Routine orders → WhatsApp morning digest
Stats → App push (weekly)
Merchants respond with a single digit from any phone. No app needed. Works on feature phones.
By operating the primary product, ACE becomes the most valuable dataset in emerging markets — data that is ungoogleable, unscrapeable, and doesn't exist in structured form anywhere else.
| Enterprise Product | Buyers | Year 2 ARR |
|---|---|---|
| AI Training Data Marketplace | OpenAI, Anthropic, Google DeepMind, Meta AI, Mistral | $7.2M |
| FMCG Market Pulse | Unilever, Nestlé, P&G, PZ Cussons, Dangote | $5.76M |
| ACE TrustScore API | Kuda, FairMoney, GTBank, Access Bank, MFIs | $3.57M |
| Total B2B (Year 2) | $16.53M |
Five data types collected at zero marginal cost: Conversational transcripts (AI labs) · Commerce signals (FMCG) · Negotiation traces / price elasticity (FMCG) · Merchant behavioural signals (TrustScore) · Goods-level product intel (FMCG + TrustScore)
We are Scale AI meets Respond.io, built for the $2.3T informal economy that traditional SaaS completely ignores.
A merchant considering leaving ACE must calculate:
| What They Lose | Monthly Cost of Leaving |
|---|---|
| Aggregate logistics pricing (₦600 → ₦280/delivery) | ₦20,000/month |
| Escrow trust signal (34% higher conversion, 22% higher AOV) | ₦35,000/month revenue impact |
| Exclusive supplier network (20% COGS discount) | ₦15,000/month margin loss |
| Global Buyer ID 1-tap checkout network | ₦40,000/month revenue impact |
| Total cost of leaving | ₦110,000/month |
| Cost of staying (subscription) | ₦12,000/month |
| ROI of staying | 817% |
| Layer | Stack |
|---|---|
| Agent / AI orchestration | TypeScript + Vercel AI SDK (generateObject for typed intents, tool-calling into Rust services) |
| Core microservices | Rust (Actix-web, Tokio, SQLx, Tonic, rdkafka) |
| AI/ML inference | Python + FastAPI (Whisper, regional dialect BERT, GPT-4o) |
| Merchant app | React Native + Expo (OTA updates) |
| Customer interface | Progressive Web App (WhatsApp in-app browser, < 3s on 3G) |
| Primary DB | PostgreSQL (ACID, partitioned by merchant_id) |
| Vector DB | Qdrant (conversation embeddings, visual product embeddings) |
| State cache | Redis (conversation state, service window tracking, distributed locks) |
| Event bus | Apache Kafka (immutable event log — all domain events) |
| Data warehouse | ClickHouse (FMCG dashboards, enterprise analytics) |
| Size | |
|---|---|
| TAM — Sub-Saharan Africa, SE Asia, Latin America | $2.3T annually |
| SAM — WhatsApp-dominant markets (Nigeria, Kenya, Indonesia, Brazil, Mexico) | $840B |
| SOM Year 1 — Nigerian fashion, food distribution, personal care | $12B |
| Revenue Stream | Year 1 | Year 2 | Year 5 |
|---|---|---|---|
| Merchant subscriptions (B2C) | $1.2M | $8.27M | $45M |
| AI Training Data | — | $7.2M | $18M |
| FMCG Intelligence | — | $5.76M | $22M |
| TrustScore API | — | $3.57M | $16M |
| Proprietary ASR API | — | — | $15M |
| Total ARR | $1.2M | $24.8M | $116M |
| Milestone | Date |
|---|---|
| Project start | May 18, 2026 |
| 50 beta merchants onboarded (Lagos) | July 31, 2026 |
| All 4 autonomous workflows operational | August 31, 2026 |
| Payment verification live (banking API + escrow) | September 15, 2026 |
| Logistics aggregator live (Kwik + Gokada) | September 30, 2026 |
| Demo-ready: full lead-to-close pipeline autonomous | October 31, 2026 |
- Phase 1 Architecture — 11 microservices, full system diagram
- Autonomous Workflows — 4 core workflows with step-by-step flows
- Lead-to-Close Pipeline — 10-stage full automation pipeline
- Vendor Communiqué System — SMS decision protocol
- AI Negotiator — Negotiation arc, 6 tactics, circuit breaker
- Financial Model — Subscription tiers, unit economics, projections
- Go-to-Market Strategy — Lagos pilot, merchant acquisition
- Enterprise Products — B2B data product strategy
- Competitive Moats — Four structural lock-in mechanisms
- Risk Mitigation — Critical flaws and hardened solutions
- Engineering Guidelines — 8 principles, critical vulnerability mitigations
- AI Training Strategy — Model hierarchy, RLHF pipeline
- Data Collection Architecture — 5 data types, Kafka topology
- Phase 2 Architecture — Multi-channel enterprise platform
- Vercel AI SDK Config — Tools, schemas, training middleware
- Database Architecture — All 5 databases with design rationale