Linkealls - AI Sales

An AI sales and support team for every small business.

Comment

-What it does

Linkealls gives small businesses an AI-powered sales and customer service agent that works through chat and voice calls.

Customers can ask questions in Angolan Portuguese, discover available products or services, receive relevant product cards with images and prices, and speak directly with the AI agent. The agent qualifies leads by asking relevant questions about the customer’s needs, budget, timeline, and preferred contact method.

For business owners, Linkealls centralizes the business profile, catalogue, customer conversations, qualified leads, marketing campaigns, payments, and follow-up.

The platform helps small businesses respond faster, present their products more professionally, and turn more conversations into real sales opportunities.

The same conversational experience can also initiate purchases and support Multicaixa Express payments, allowing customers to move from a question to a completed transaction without leaving the chat.

-How we built it

We built Linkealls as a full-stack web application using React, TypeScript, Node.js, Express, and PostgreSQL.

Gemini is the intelligence layer behind the customer-facing agents. We use the Gemini API for:

  • real-time voice conversations through Gemini Live;
  • speech input and output transcription;
  • natural-language customer support;
  • lead qualification;
  • product recommendations;
  • campaign copy generation;
  • image analysis for advertising creatives;
  • structured tool calls.

The AI agent can call application tools to show catalogue products, send written information during a voice call, initiate a checkout, and check payment status.

The frontend communicates with the backend through REST APIs and WebSockets. WebSockets power the live call experience, audio streaming, transcripts, product cards, agent messages, and real-time checkout updates.

We also integrated local payment infrastructure for Multicaixa Express through the Ekwanza payment flow. Payment validation, order creation, and business ownership checks are handled securely on the server.

-Challenges we ran into

One of our biggest challenges was making a real-time voice agent feel natural while also allowing it to take actions inside the application. The agent needs to speak, listen, call tools, display information, and continue the conversation without making the experience feel robotic.

We also learned that an AI model does not always call a tool when we expect it to. For example, Gemini may answer a question about a product by voice without explicitly calling the catalogue tool. To make the experience reliable, we added server-side intent detection as a fallback so relevant product cards are still shown.

Another challenge was keeping AI responses short enough for chat and voice. Long responses create friction, especially on mobile devices. We added stricter instructions, response limits, and message splitting so the agent communicates in short, natural Portuguese messages.

The payment flow required additional care because orders, payment states, and business accounts must remain correctly scoped. We added validation for phone numbers and quantities, checked that orders belong to the current business, and made checkout status visible in real time.

We also had to adapt the experience to the Angolan market, including Portuguese language, local communication patterns, mobile-first interaction, and Multicaixa Express payments.

-Accomplishments that we're proud of

We are proud that Linkealls is more than a chatbot. It is an AI-operated sales workflow for small businesses.

The agent can:

  • answer customers in Angolan Portuguese;
  • hold real-time voice conversations;
  • qualify leads naturally;
  • show relevant products and services inside the conversation;
  • send written information while continuing a voice call;
  • start a purchase or payment flow;
  • track payment status;
  • notify the business owner about new leads;
  • help generate advertising campaigns;
  • analyze campaign images;
  • create advertising copy based on the uploaded image.

We are especially proud of connecting the conversational experience to real business actions. A customer can move from a question, to a qualified conversation, to a product recommendation, and finally to a payment flow without leaving the experience.

We are also proud of building the product around the needs of businesses in Angola instead of simply adapting a generic global chatbot.

-What we learned

We learned that useful AI products need deterministic systems around the model. The model is responsible for understanding language and making recommendations, but the application must validate actions, enforce permissions, and provide reliable fallbacks.

We learned that voice agents need short turns, clear questions, and natural pacing. A technically correct answer can still feel frustrating if it is too long or asks too many questions at once.

We also learned that showing information visually at the right moment is extremely powerful. A customer may hear about a product during a call, but seeing the image, price, and key details immediately makes the conversation easier to understand.

Local context matters. Language, payment methods, trust, mobile connectivity, and customer habits all affect how an AI product should be designed.

Most importantly, we learned that AI adoption for small businesses depends on reducing complexity. Business owners should not need to understand AI tools, prompts, or automation. They should configure their business once and let the agent handle the repetitive work.

-What's next for Linkealls

Our next step is to make Linkealls a complete AI operating system for small businesses.

We plan to add:

  • smarter product and service search by category, price, and customer needs;
  • better matching between customer requests and catalogue items;
  • automatic appointment and consultation scheduling;
  • lead scoring and prioritization;
  • automated follow-up campaigns;
  • reminders for customers and business owners;
  • dashboards showing conversion from conversation to sale;
  • stronger integrations with catalogues, calendars, and CRM systems;
  • mobile notifications when a high-intent lead arrives;
  • multilingual support for Portuguese and other local communication needs;
  • more payment and business automation integrations.

Our long-term goal is to help every small business in Angola operate with an AI sales and customer service team that is available 24/7, understands the local market, and helps turn more conversations into completed transactions.

Built With

Share this project:

Updates