Building Ekaette: An AI Voice Agent for African Businesses Using Amazon Nova Sonic
Ekaette is a configurable multimodal AI voice and messaging assistant built on Amazon Nova and AWS Bedrock. It handles live phone calls, WhatsApp, and SMS for customer-facing businesses.
The Problem: African Businessess Are Missing Calls and Losing Sales
In Africa, WhatsApp is the primary way customers interact with businesses. Over 50 million Africans use WhatsApp daily, to ask about products, check prices, place orders, and arrange deliveries. For small and medium businesses like electronics retailers, fashion stores, and service providers, every missed WhatsApp call is a missed sale, every late response can also lead to lost sales.
But most African SMBs can't afford dedicated call center staff. The owner answers calls between serving walk-in customers. After hours, calls go unanswered. During peak periods, customers wait or give up. And when they do get through, the person answering may not know current prices, stock availability, or delivery options.
The businesses that can afford call centers, still rely on static recordings, long hold queues, and rigid call-centre routing. Customers spend significant time and airtime waiting to resolve simple requests, and urgent needs are delayed by generic queue systems that do not understand intent or priority. It is a frustrating experience that almost everyone recognizes: too much waiting, too little help, and no fast path for straightforward tasks.
This is the gap Ekaette fills the gap
What Ekaette Does
Ekaette is an AI-powered customer service agent that answers voice calls and whatsapp calls, and replies messages for businesses across whatsapp,instagram. and facebook. When a customer calls, Ekaette:
- Greets them in warm, natural Nigerian English — the way a friendly Lagos professional would talk
- Answers questions about products and prices — with accurate, up-to-date catalog data in naira
- Explains store policies — hours, pickup options, warranty terms, payment methods
- Handles payments — generates virtual account numbers via Paystack so customers can pay immediately
- Checks delivery fees — uses Topship to quote delivery costs and estimated arrival times
The agent sounds natural, keeps responses short, and knows when to stop talking. It doesn't sound like a robot reading a script — it sounds like a knowledgeable, helpful colleague.
Why Amazon Nova Sonic
Building a voice AI agent for Nigerian customers requires a model that can handle real-time, bidirectional conversation — not the traditional "transcribe, think, synthesize" pipeline that introduces awkward delays. Amazon Nova 2 Sonic on Bedrock provides exactly this: a single model that listens and speaks simultaneously, with built-in turn detection that understands conversational rhythm.
This matters for Nigerian English specifically. Nigerian English has distinct prosody, pacing, and conversational patterns. A pipeline approach with separate ASR and TTS models trained primarily on American English introduces errors at every stage. Nova Sonic handles the full audio loop natively, reducing compounding errors and producing more natural interaction.
The technical architecture uses Nova 2 Sonic for voice streaming and Nova 2 Lite for text-based WhatsApp message handling — both through Amazon Bedrock.
Architecture: From WhatsApp Call to AI Response
The system works like this:
- A customer places a voice call via phone or whatsapp to the business number
- The call routes through Africa's Talking SIP infrastructure to our EC2-based SIP bridge
- The bridge handles Opus codec conversion, SRTP encryption, and media packetization
- Audio streams bidirectionally to Nova 2 Sonic via Bedrock's streaming API
- Nova Sonic receives a grounded system prompt containing the business's identity, product catalog, pricing, and policies
- The model responds conversationally, and audio flows back through the same pipeline
- An AWS Polly-synthesized greeting plays immediately on call connect, so the customer hears a voice within the first second
For text messages on WhatsApp,instagram and messenger the same knowledge base powers Nova 2 Lite responses through the Bedrock Converse API, running on ECS Fargate.
The entire infrastructure is AWS-native: EC2 for the SIP bridge, ECS Fargate for the API service, S3 for media storage, DynamoDB for session state, Secrets Manager for credentials, and Terraform for infrastructure-as-code.
How Ekaette Can Positively Affect African Communities
Leveling the Playing Field for Small Businesses
Africa has over 40 million micro, small, and medium enterprises. Most compete against larger businesses with dedicated customer service teams, call centers, and CRM systems. Ekaette gives a solo electronics shop owner in Ikeja the same 24/7, knowledgeable, professional phone presence as a major retailer for a fraction of the cost.
A customer calling at 10 PM to ask "do you have iPhone 15 Pro Max in stock, and how much?" gets an immediate, accurate answer with the price in naira, instead of waiting until morning. That's a sale that would have gone to a competitor.
Reducing Language and Accessibility Barriers
Many Nigerian customers are more comfortable speaking than typing. Voice-first AI meets them where they are on Phone orc WhatsApp, speaking naturally, in Nigerian English. There's no app to download, no website to navigate, no chatbot syntax to learn. You call a number and talk to someone helpful.
Building Trust Through Accuracy
One of the biggest pain points in Nigerian e-commerce is price inconsistency. Customers call, get quoted one price, then see a different price online or in-store. Ekaette pulls from a single source of truth — the business's actual product catalog — so every customer gets the same accurate, current pricing. This builds trust in a market where trust is the primary barrier to commerce.
Enabling After-Hours Commerce
Nigerian business hours don't match customer availability. Many customers work during store hours and can only shop in the evening. Ekaette handles calls 24/7, turning after-hours inquiries into confirmed orders with payment links — extending the effective business day without extending staff hours.
Real-World Application: Ogabassey Gadgets Store
Our live demo showcases Ekaette serving "Ogabassey Gadgets Store" a trade-in focused gadgets store in Lagos. The agent handles:
- Product inquiries across 11 items (phones, laptops, accessories) with naira pricing
- Store policy questions (9 AM - 7 PM WAT hours, same-day pickup before 2 PM, next-day delivery)
- Payment processing (bank transfer, debit cards, USSD, OPay, PalmPay, international cards)
- Trade-in warranty details (30-day guarantee, 14-day re-evaluation window)
- Delivery fee quotes via Topship integration
This isn't a prototype, it's a working system handling real WhatsApp voice calls today.
Plans for Encouraging Adoption
Multi-Tenant Architecture
Ekaette is built as a multi-tenant platform from day one. The grounded prompt system loads company identity, knowledge base, and product catalog from vendored JSON configuration. Onboarding a new business means:
- Creating a company profile (name, overview, location, hours)
- Adding knowledge entries (policies, FAQs, service details)
- Uploading a product catalog (names, prices, descriptions)
- Assigning a WhatsApp number
No code changes required. The same infrastructure serves multiple businesses, each with their own persona, catalog, and policies.
Nigerian Market Expansion
The immediate roadmap targets three high-impact verticals in the Nigerian market:
- Electronics retailers — price inquiries, stock checks, trade-in quotes
- Fashion and lifestyle — size guidance, style consultation, order tracking
- Hospitality — room availability, booking confirmation, local recommendations
Each vertical gets an industry-specific knowledge template with pre-built policy structures and common FAQ patterns.
Vernacular Language Support
As Nova Sonic's language capabilities expand, Ekaette can extend to Pidgin English — the most widely spoken language in Nigeria with over 75 million speakers. A customer in Aba market should be able to call and ask "how much be this phone?" and get a natural response. This is the ultimate adoption driver: meeting every Nigerian customer in their own language.
Conclusion
Ekaette demonstrates that Amazon Nova Sonic can power production-grade voice AI for emerging markets where voice-first, WhatsApp-native customer service isn't a nice-to-have — it's the primary channel. The combination of real-time bidirectional voice, grounded knowledge, and AWS-native infrastructure creates a platform that can scale across thousands of Nigerian businesses, giving each one an always-on, knowledgeable, and genuinely helpful voice agent.
The technology is ready. The market need is urgent. And with Amazon Nova, the cost of serving each additional business drops to near zero.
Built with Amazon Nova 2 Sonic, Amazon Nova 2 Lite, Amazon Bedrock, AWS Polly, and the Africa's Talking Voice API.
#AmazonNova
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article