AI Chatbot and Business Automation for Saudi Companies: WhatsApp, Website and Back Office
WhatsApp and website assistants, lead capture and workflows between your store, CRM and ERP, with a person always one step away. We quote by what you need.
Devvista builds chatbots and business automation for companies in Saudi Arabia, in Arabic and English: a WhatsApp or website assistant that answers the common questions and captures the lead, and the back-office automation that moves orders, leads and documents between your tools without retyping. You see a working version each week, and the finished workflows, accounts and documentation are registered to your company. We price either as a fixed project from $500 to $100,000 or hourly from $10 to $100, depending on complexity and developer level.
This page explains what a chatbot and automation project really contains, what a machine should and should not be left to decide, how WhatsApp rules affect the design, and what moves the price. For a quote, contact us or use the Get an Estimate button.
What can be automated, and what each piece needs
Start with the job, not the technology. These are the jobs we are most often asked about, and what each one involves.
| The job | What gets built | What still needs a person |
|---|---|---|
| Answer common customer questions on WhatsApp or the website | An assistant that answers from your own price lists, policies and product data and hands over to staff when unsure | Complaints, negotiation, anything not in your approved material |
| Capture and qualify leads | A conversation or form that collects name, need and budget range and writes it into your CRM | Deciding which leads are worth a call |
| Move data between systems | Workflows that sync orders, customers and stock between your store, ERP and CRM | Fixing records that fail validation |
| Read documents | Extraction of fields from invoices, forms or emails into a structured record, checked against your rules | Documents the system flags as uncertain |
| Send reminders and updates | Approved message templates for order status, appointments and payment reminders | Replies that need judgment |
If a simple rule does the job, we build the rule. A chatbot is worth adding only where customers ask the same questions in many different words.
WhatsApp chatbots: what the platform requires
WhatsApp is the channel we are most often asked to automate first. Three rules from Meta's own documentation shape the design (WhatsApp Business Platform, Cloud API overview).
Incoming messages arrive through webhooks. The documentation says the platform relies heavily on webhooks, and delivery status updates for outgoing messages come the same way. That is how a bot, a CRM or your ERP learns that a customer wrote.
Templates are the only message type allowed outside a customer service window. A company that wants to start a conversation, for example to remind a customer about an invoice, must use a template, and templates generally need approval before they can be sent.
You need opt-in. Meta states that you must obtain user opt-in before sending message templates, and that the opt-in must make clear your business name and intent. We build the opt-in into your forms and checkout.
Cost is a separate line from our fee. Meta's pricing page says it charges per delivered message, with rates that depend on the message category and the recipient's country calling code, and that it does not charge for messages from users to a business (WhatsApp Business Platform pricing). The same page gives each business phone number a monthly allowance of service messages at no charge (1,000 per phone number, per the page, with a change noted from 1 October 2026), so replies are not unlimited at no cost. Read the rate card for your own volumes before you plan the budget.
If you run Odoo, its documentation says the WhatsApp integration is Enterprise-only, works only with a WhatsApp Business Platform account (not a personal account or the WhatsApp Business app), and requires a Meta business account and developer app plus a webhook set up in Odoo (Odoo documentation, WhatsApp). Our Odoo implementation team sets this up together with the quotations, invoices and tickets you want to send.
Arabic, dialects and honest limits
Customers write in formal Arabic, in their own dialect, in Arabizi (Arabic in Latin letters), in mixed Arabic and English, and often in voice notes. In our experience, tested per project, a language model reads much of this well but not all of it, and it can sound sure of itself when it is wrong. We do not promise a figure for accuracy before we have seen your real conversations.
What we do instead is practical. We collect a sample of your real customer messages (with your permission, with names removed), test the assistant against them before launch, and write down where it fails. Answers come only from material you approved, such as price lists, delivery terms and opening hours. When a question falls outside that material, or the customer is upset, or the message mentions money or a complaint, the assistant hands the chat to a person with the conversation attached. We build the hand-over first and the clever part second.
Voice notes, images and documents need their own handling and their own cost, so we decide in scoping whether they are in the first release. Tell us your case and we will say honestly if a bot is the right tool.
Back-office automation: n8n, Make and direct API links
The unglamorous half of the work saves the most time: an order placed in the store creates the customer and the invoice in the ERP, a new lead lands in the CRM with the right owner, a supplier PDF becomes a draft bill for a person to approve.
For flows of low to medium complexity we use workflow tools. n8n describes itself as a workflow automation tool that combines AI features with business process automation, it is fair-code licensed, and it can be hosted on n8n Cloud or on your own server with Docker (n8n documentation, n8n hosting). Self-hosting is the option to consider when you want the workflows and their data to stay on infrastructure you control. We also build on Make, and where a tool does not fit we write the integration in code against the system's API.
Many Saudi stores run on platforms that publish developer interfaces. Salla's documentation describes RESTful merchant APIs, OAuth 2.0 authorization and webhooks for store events such as orders, products, shipments, customers and carts (Salla developer documentation). Zid's documentation describes public and private apps and a webhook log for checking data exchanges, and notes that a private app needs the store to be on a Professional or Enterprise plan or to have API Access activated (Zid developer documentation). That is what lets an order in your store trigger a message, an invoice or a stock update elsewhere. See our online store development page for when to stay on those platforms.
Every workflow we build has error handling, logging and an alert, so a failed step is reported and does not silently stop the process.
What should stay with a human
Automation is strongest on repeatable, rule-based work and weakest where judgment, relationships or legal duties are involved. This is how we split the work.
| Leave to the system | Check with a human first | Keep entirely with people |
|---|---|---|
| Opening hours, order status, delivery terms, price-list answers | Quotes, discounts, refunds, anything with a legal or tax effect (for example invoices) | Complaints and disputes |
| Copying data between your store, CRM and ERP | Extracted fields from scanned documents | Hiring, firing and contract decisions |
| Reminders using approved templates | Messages to customers who did not opt in | Negotiating large deals |
Invoices deserve a special mention. If an automation creates or sends tax invoices, it may fall under ZATCA e-invoicing, depending on your VAT registration and your wave, and the connection to the Fatoora platform may then be part of the scope. See our ZATCA Wave 25 guide and Odoo e-invoicing page.
Customer data and privacy
A chatbot reads and stores personal data: names, phone numbers, addresses and the content of conversations. Before building we ask four questions and write the answers into the scope: where the data is stored, which outside services touch it (the messaging platform, the AI provider, the automation tool), how long conversations are kept, and who can read them.
Our global AI and automation page (English) describes the option of locally hosted open-source models for cases where data must stay on your own infrastructure. Which option fits depends on the data and on your own legal duties. We have not quoted Saudi data protection legislation on this page because we could not open the official text while writing it, so please have your legal adviser confirm what applies to you; we will build to the answer.
Our own examples, and what they do not show
We have no Saudi clients to show on this page and we will not invent any. These are published case studies from other markets, so you can judge how we build and document automation:
- AI lead intake and blueprint automation with Make.com: a 112-field intake synced into GoHighLevel and Airtable, with six platforms integrated and an AI-generated document that is held for internal approval before delivery. The case study lists Typeform, GoHighLevel, Airtable, Claude AI, Google Docs and Slack. The client is confidential.
- AI credit repair and automation platform for Harold: a multi-agent AI analysis system with human checkpoints and CRM automation, built in a regulated US context.
Both are English-language, US-style projects. They show the method, human review included, not Arabic chatbots. Our CRM development page (English) covers the system a chatbot usually feeds.
How a chatbot or automation project runs
Every project follows the same steps.
| Step | What happens | What you get |
|---|---|---|
| 1. Discovery call | Which process, how often it happens, how it is done today, which tools are involved | A short written summary of the goal |
| 2. Process map | We map the steps and mark what can be automated, what needs approval and what stays manual | A one-page map and a list of first-release features |
| 3. Build | Working workflows or a working bot every one to two weeks, shown on a video call, connected to a test copy of your tools | A version you can try yourself |
| 4. Test with real data | Real (anonymised) messages and records, edge cases and failures | A written list of what passed and what did not |
| 5. Launch and watch | Go live on one channel or one process first, with alerts on errors | Monitoring and a hand-over path to staff |
| 6. Handover | Workflows, accounts and documentation in your company name; optional monthly support | You own everything |
What decides the price, with one worked example
Devvista prices this work either as a fixed price from $500 to $100,000 depending on complexity, or hourly from $10 to $100 per hour depending on the type and level of developer. Five things move the number most: how many channels (WhatsApp, website, email), how many systems must be connected, how much of the answer material must be organised first, how much Arabic testing is needed and whether the data must stay on your own server.
Illustrative estimate, not a quote. Take a company that wants new website and WhatsApp enquiries captured, written into the CRM with an owner assigned, and a confirmation sent. Assume 40 to 80 hours of development, setup and testing (an assumption for this example). At an illustrative $40 per hour that is $1,600 to $3,200 of our time. Your own scope will differ, and we write a plan before you commit.
Running costs are paid to the providers, not to us: the messaging platform, any AI model usage and any workflow-tool subscription. We list them in the plan so there are no surprises.
Where Devvista fits
A chatbot project at Devvista starts with a list of the questions your customers really ask and ends with a hand-over path to your staff. In between we build, test against your own messages and launch on one channel first. Tell us your case and we will say honestly whether a bot is the right tool.
Related pages: Odoo ERP implementation, software development company, website development, online store development, hiring developers and the Saudi home page. To discuss your process, contact us.
Contact Devvista
Our address, phone and opening hours.
Amna Bint Wahb St, road, Jeddah 23622, Saudi Arabia
- SundayOpen 24 hours
- MondayOpen 24 hours
- TuesdayOpen 24 hours
- WednesdayOpen 24 hours
- ThursdayOpen 24 hours
- FridayClosed
- SaturdayClosed
Tell us what you need and we will send an estimate. Get an estimate.
Straight Answers.
Services
Custom Software DevelopmentAI & AutomationSaaS DevelopmentCRM DevelopmentOdoo Implementation & DevelopmenteCommerce DevelopmentDedicated Development TeamsWordPress DevelopmentUI/UX DesignSEO ServicesIT Staff AugmentationWhite Label DevelopmentGuides
ZATCA e-invoicing for OdooSoftware development for Saudi ArabiaZATCA Wave 25Wave 25 CheckerContact and bookingOdoo ERPERP Cost GuideAccounting and e-invoicingLogistics softwareWebsite DevelopmentMobile App DevelopmentHire developers for Saudi ArabiaHire .NET developers for Saudi ArabiaAI Chatbots and AutomationOnline Store DevelopmentContracting softwareManufacturing ERPWholesale and distributionHire Odoo developers for Saudi Arabia Case Studies Devvista BlogTell us about your process
Tell us which process eats your team's time or which questions customers keep asking. We will quote it in writing after a short scoping call.
- 30-minute consultation
- No commitment required
- Response within 24 hours
- Book a call30 minutes with the team.
- Get a clear planSteps, timeline and a cost range for your setup.
- DecideGo ahead with us, or take the plan with you.