AI Agents

AI that sells and serves —
and knows when to hand off

Agents that answer in Arabic and English from your own knowledge base, and turn conversations into structured CRM data — inside guardrails you define.

app.nexus.com/inbox · AI Agent active
Hi, do you deliver to Al Wakrah? How long does it take? 21:14 — outside business hours
Hello! Yes — we deliver to Al Wakrah within 1–2 business days, free for orders above QAR 200. Would you like to place an order? 21:14 ✓✓
AI · Sara — answered from source: "shipping-policy.pdf"
I want to order 60 units for my company — who do I talk to? 21:16
Excellent! Bulk orders are handled by our sales team directly. I've logged your request — Omar will contact you tomorrow morning. 21:16 ✓✓
AI · Sara — new deal "Bulk order ×60" + task for Omar + human handoff
How it works

From setup to first reply in minutes

1

Connect your knowledge

Upload PDFs, add URLs and text — Nexus builds a knowledge base your agent answers from, citing the source of every answer.

2

Set the guardrails

What may the agent do autonomously, and what does it only suggest for human approval? When does it hand off? You decide — per agent.

3

Activate & monitor

The agent works on the channels you choose, and every action lands in the AI activity log — full transparency for your team and customers.

AI ↔ CRM

Reads customer context. Writes structured outcomes.

This is what sets Nexus apart: the agent isn't an isolated chatbot — it's a team member that knows the customer's history and updates their records.

  • Reads: contact data, open deals, past conversations
  • Writes: lead qualification, deal creation, stage updates, follow-up tasks
  • Suggests: sensitive changes go to a human for approval first
AI core connected to CRM data
●Autonomous: shipping, pricing and hours questions
●Suggest-only: discounts and large deal edits
●Hands off: complaints, bulk orders, frustrated customers
●Never: legal promises, payments, other customers' data
Guardrails

Clear boundaries. Compounding trust.

The AI never impersonates a human: every automated reply is labeled "AI · Sara". And every action — reply, update or handoff — is recorded in an auditable log.

  • Authority levels per action type: autonomous, suggest, never
  • Handoff with full context — customers never repeat themselves
  • LLM-agnostic: plug in the language model that suits you
FAQ

Questions every team asks

Does the agent speak fluent Arabic?

Yes. The agent answers in Modern Standard Arabic by default and adapts to the customer's language — Arabic, English or a mix — within the same conversation.

What happens when the agent doesn't know the answer?

It doesn't invent one. The agent only answers from your knowledge sources; when it can't find a grounded answer, it hands off to a human with a full summary.

Can the agent change CRM data without permission?

Only within the limits you set. Every action type has an authority level — autonomous, suggest-for-approval, or never — and every action is logged.

How do customers know they're talking to AI?

Every automated reply is clearly labeled. Transparency is a hard rule in Nexus: AI never speaks as a human agent.

What are the usage limits?

Each plan includes a monthly AI-reply quota. At the limit, the agent quietly steps back and conversations continue with your team — no disruption for customers.

Hire your team's first AI member

Create your agent, connect your knowledge and activate it on WhatsApp — in under an hour.