Integration · AI Models · Nexus
Nexus + AI Models: Intelligent Operations Powered by LLMs
Nexus's AI Signal Layer uses large language models to analyze every customer conversation for intent, sentiment, complaint type, urgency, and churn risk — in Arabic and English, natively. On Growth and Scale plans, the AI runs on Nexus's own infrastructure. No OpenAI account needed. For teams who want to bring their own AI agents, Nexus also exposes a full MCP server and REST API so Claude, GPT-4, or any LLM can operate Nexus autonomously.
What This Integration Does
AI in Nexus operates on two levels. The first is built-in and always-on: the AI Signal Layer processes every inbound message in real time, classifying intent, detecting complaints, scoring sentiment, and updating the customer's churn risk profile. This happens automatically on Growth and Scale plans with zero configuration — no OpenAI key, no prompt engineering, no infrastructure to manage. The second level is extensible: for teams who want to build their own AI-powered workflows on top of Nexus data, the MCP server and REST API make the full operational surface available to any LLM. A GPT-4 agent can pull all orders in a "complaint pending" status, read the AI Signal scores, draft resolution messages, and send them via WhatsApp — all orchestrated by your agent, powered by your model, acting on live Nexus data. The two levels are independent; you can use either or both depending on your team's needs and technical capability.
AI Signal Layer — always-on conversation intelligence
Every message received in any Nexus inbox is analyzed in real time by the AI Signal Layer. The system extracts: customer intent (complaint, inquiry, return request, compliment, churn signal), sentiment score (positive / neutral / negative / urgent), complaint category (delivery, wrong item, product quality, payment issue), and urgency tier. These signals appear in the conversation sidebar instantly, feeding routing rules and escalation automation without any manual tagging by agents.
Churn risk scoring and RFM segmentation
Nexus continuously updates a churn risk score for every customer using a combination of LLM-analyzed conversation signals and order behavior data — order frequency decline, complaint rate, days since last purchase, cart abandonment patterns. These scores feed into RFM segments that are recalculated daily and available as automation triggers: automatically reach out to customers whose churn risk crossed a threshold before they place their next order elsewhere.
Bring your own AI agent via MCP or REST API
On Scale plan, full embeddings, transcription for voice messages, and AI-generated CS analytics dashboards are included. For custom AI agent workflows, the Nexus MCP server and REST API expose all operational data and actions. Connect Claude or GPT-4 using your own API account, point it at Nexus data, and build autonomous workflows that go beyond what the built-in AI handles — specialized triage bots, proactive outreach agents, or AI-powered inventory Q&A for your CS team.
Setup & Requirements
Built-in AI features require no setup beyond activating a Growth or Scale plan. The AI Signal Layer enables automatically when your plan is active and begins processing messages immediately. For custom AI agent workflows using external LLM providers, you connect through the MCP server or REST API — Nexus provides the data and actions; you manage your own model provider account separately.
- AI Signal Layer (intent, sentiment, complaint classification, churn risk) — included on Growth and Scale plans, no configuration required, activates automatically
- Conversation summarization and CS analytics dashboards — Scale plan only, runs on Nexus infrastructure, no external AI account needed
- Voice note transcription — Scale plan only, transcription powered by Nexus infrastructure, results stored in the conversation thread and customer profile
- Custom AI agents using Claude, GPT-4, or other LLMs — connect via the MCP server (see MCP Server integration page) or REST API using your own model provider account and API key; Nexus Growth or Scale plan required for write access
- No OpenAI, Anthropic, or other AI provider account is needed for any built-in Nexus AI feature — only required if you are building your own custom agent workflows on top of Nexus data
Frequently Asked Questions
Do I need my own OpenAI account to use AI features in Nexus?
No. All built-in AI features in Nexus — the AI Signal Layer, sentiment analysis, complaint classification, churn risk scoring, conversation summarization, voice note transcription, and CS analytics dashboards — run on AI infrastructure managed by AiForStartups. You do not need an OpenAI account, Anthropic account, or any AI provider subscription to use these features. They are fully included in your Nexus Growth or Scale plan at no extra cost. You only need your own AI provider account if you are building custom AI agent workflows that use an external LLM to call the Nexus API.
What AI features are built into Nexus?
Nexus includes a comprehensive suite of built-in AI capabilities: the AI Signal Layer for real-time message analysis (intent, sentiment, complaint type, urgency — on Growth and Scale); continuous churn risk scoring updated from conversation and order data (Growth and Scale); automated complaint classification and escalation routing based on AI signals (Growth and Scale); RFM-based customer segmentation recalculated daily from purchase and engagement data (Growth and Scale); conversation summarization that generates a one-paragraph summary of any long conversation thread (Scale); voice note transcription with Arabic and English support (Scale); and AI-generated CS analytics dashboards with insights on complaint trends, agent performance, and customer health (Scale).
Does the AI work in Arabic?
Yes. Arabic and English are both natively supported by the Nexus AI Signal Layer with no configuration or additional cost. The underlying models understand Egyptian Arabic dialect, Modern Standard Arabic, Levantine Arabic, and natural code-switching between Arabic and English — all common in Middle Eastern ecommerce customer conversations. Sentiment, intent detection, complaint classification, and churn risk signals work correctly in both languages. You do not need to configure the language or label training data — the system handles bilingual conversations automatically.
What is the AI Signal Layer?
The AI Signal Layer is Nexus's real-time message intelligence system. Every message that arrives in any Nexus inbox — WhatsApp, Instagram DM, Facebook Messenger — is analyzed immediately when received. The system extracts and stores: customer intent (inquiry, complaint, return request, compliment, urgent issue, churn signal), sentiment score with a directional label (positive, neutral, negative, or urgent), complaint category when applicable (delivery delay, wrong item, product quality, payment issue, courier problem), urgency tier (low, medium, high, critical), and a churn risk contribution score that updates the customer's overall churn profile. All signals appear in the conversation sidebar in real time and are available as conditions in Nexus automation rules.
Can I use Claude or GPT-4 as an agent against Nexus?
Yes. You can connect Claude, GPT-4, or any other large language model as an autonomous agent operating on Nexus data. For Claude and other MCP-compatible agents, use the Nexus MCP server — the agent discovers all available tools automatically from the manifest and can operate Nexus without any custom integration code. For GPT-4 or other REST-based agent frameworks, call the Nexus REST API directly using your Nexus API key for authentication. Your LLM provider account (Anthropic, OpenAI, or other) is entirely separate from Nexus — you pay your LLM provider for inference, and Nexus provides the operational data layer and action endpoints the agent calls.
Run AI-powered operations without managing your own AI infrastructure
Every conversation analyzed, every churn risk scored, every complaint classified — automatically, in Arabic and English, on your Growth or Scale plan.