AI agent for customer service in Arabic, and an AI sales agentBuilt as projects in Jordan & Saudi Arabia
Customers write at night, call at lunch and ask the same questions in Arabic and English. An AI agent for customer service in Arabic and English answers from your own knowledge bank, works as an AI sales agent that qualifies enquiries, and hands the serious or sensitive ones to your team with the context attached. It is part of our AI agents service.
The same questions, at every hour, in two languages
Most customer conversations fall into a few groups: opening hours and locations, prices and availability, bookings, order or case status, and a first enquiry from someone who might buy. Staff answer the same questions many times a day, and the ones that matter, a buyer ready to visit or a guest with a problem, wait in the same queue.
An AI agent for customer service Arabic and English customers can message removes the queue for the routine questions. It answers from your own documents, in the language the customer wrote in. An AI sales agent does the first half of sales work: it asks the qualifying questions, records the answers and books the visit or call. Your team keeps the conversation as soon as a person adds value.
Nplay builds these as agents connected to your data, not as a fixed menu. The backbone is the same conversational approach as our AI Receptionist and the knowledge-bank avatars inside AI hologram displays. For Jordan and Saudi Arabia, Arabic is a first-class language, not an afterthought. See our guide to AI agent use cases by department for the wider picture.
Where AI agents help with service and sales
Each sector needs its own knowledge, hand-off rules and hours. We scope one sector and one channel first.
Retail and malls
Answer store-hours, location, stock and promotion questions, guide shoppers to a department and capture a call-back request. An in-venue AI receptionist can do the same on a screen.
Hospitality
Answer questions about rooms, facilities, check-in, dining and events, take enquiries and pass booking changes, complaints and special requests to the front desk.
Virtual tours for hotels →HospitalityReal estate sales
Qualify buyers and renters, share unit details from an approved list, book viewings and brief the adviser before the visit.
Virtual tours for property →Real estate salesClinics
Answer questions on services, opening times and preparation instructions and book appointments. The agent does not give medical advice and sends clinical questions to staff.
Education
Answer admissions, programme and fee-process questions for prospective students and parents, capture enquiries and book visits and open-day slots.
Service counters
For government-style counters, offices and utilities, explain the steps for a service, list the documents needed, give queue guidance and route complex cases to a person.
A lead-handling agent, step by step
This is an illustrative scenario of an AI sales agent, with no figures, to show how the pieces fit together. It does not describe a named client.
A property developer runs a show flat and an online listing. A visitor writes on WhatsApp late in the evening, in Arabic, asking whether two-bedroom units are available and what the payment plan looks like.
- The agent greets the visitor in Arabic and answers the availability question from the approved price and unit list in its knowledge bank.
- It asks a few qualifying questions: the budget range, the intended use, the preferred handover period and whether the visitor is buying or investing.
- It explains the payment plan in the wording approved by the sales team, without adding promises, and says that final terms come from a sales adviser.
- It offers two time slots for a show-flat visit, checks the adviser's calendar and books one.
- It writes a summary into the CRM: who the lead is, what they asked, what they answered, the language used and the booked slot.
- The next morning the adviser opens a lead that is already qualified, with the conversation attached, and calls the visitor before the visit.
If the visitor had asked for a discount, disputed a reservation or sounded upset, the agent would have stopped, told the visitor a person would pick it up and flagged the conversation for an adviser in working hours.
What the agent handles and what goes to a person
| The agent handles | A person handles |
|---|---|
| Opening hours, locations, directions and contact details | Complaints and anything from an upset customer |
| Product, unit, service and price-list questions from approved material | Discounts, negotiation and custom terms |
| Qualifying questions and lead capture | Closing a sale and signing contracts |
| Booking, rescheduling and cancelling within set rules | Refunds, payment disputes and exceptions |
| Order, case or application status read from a connected system | Cases where the record is missing, wrong or disputed |
| Summaries and notes written into the CRM or helpdesk | Medical, legal, financial or safety questions |
| Reminders and follow-up messages that staff have approved | Any case the agent marks as uncertain |
Where the agent meets your customers
- Website chat: answers on product and service pages, with the page the visitor is on as context
- WhatsApp: the channel many customers in Jordan and Saudi Arabia already use, with message templates where a business message needs one
- Phone: a voice AI agent Arabic callers can speak to, for after-hours and busy periods, with transfer to a person
- Email: replies drafted or sent for routine questions, with sensitive threads left for staff
- In-venue: an AI receptionist or hologram avatar in a lobby, showroom or mall that uses the same knowledge bank
- One brain, many channels: the same knowledge and rules, so customers get the same answer wherever they ask
From first enquiry list to a live agent
Choose the enquiry group
We review your real enquiries and pick the group with the most repeated volume, such as availability questions or booking requests.
Build the knowledge bank
We collect your approved price lists, policies and answers in Arabic and English and set what the agent may say and may not say.
Connect the systems
We link the CRM, calendar or booking system, read-only first, then allow the actions you approve, like creating a lead or a slot.
Test on real conversations
We run past enquiries and recorded calls through the agent and review the answers with your staff.
Launch with hand-off
We go live on one channel, with a named person for hand-overs and review of every conversation in the first period.
Measure and expand
We review the measures below and add a channel, a language or another enquiry group.
How to judge whether the agent is working
- Share of conversations resolved without a person, read alongside customer feedback rather than alone
- Quality of hand-overs: did the person receive the context and avoid asking again
- Response time outside working hours compared with before
- Qualified leads and booked visits that came through the agent
- Questions the agent could not answer, used to grow the knowledge bank
- Errors found in review, such as a wrong answer or a missed hand-off, and how fast they were fixed
- Staff view: is the team spending more time on cases that need them
Further reading: NIST AI Risk Management Framework (external source).
AI agent for customer service in Arabic: frequently asked questions
Can you build an AI agent for customer service Arabic customers will actually use?
Yes. We build for Modern Standard Arabic and for Gulf and Levantine dialect, and we test on your own real messages. Customers can switch between Arabic and English in the same chat, and the agent replies in the language they are using.
What does an AI sales agent do and what does it not do?
It answers product and price questions from approved material, asks qualifying questions, books visits or calls and writes a summary for your salesperson. It does not negotiate discounts, agree custom terms or close a deal. A person does that.
Can the agent work on WhatsApp and the phone, not just a website?
Yes. We build agents for website chat, WhatsApp, phone and email, all using one knowledge bank. Each channel has its own rules and testing, so we usually start with one and add others.
What happens when the agent does not know the answer?
It says it does not know, offers to pass the question to a person and records the question with the conversation. Those gaps are reviewed and added to the knowledge bank if they are common.
Do you publish results such as response times or conversion lifts?
No. Results depend on your enquiries, your team and your channels, so we agree measures with you and review them together rather than promise a figure in advance.
Keep exploring
Let customers get an answer at any hour
Tell us your sector, the channel your customers use most and the five questions they ask every day. We will propose a first agent, the hand-off rules and the measures we will review with you.