Build your own chatbot or use AI Chatbot: a comparison
Building your own chatbot means you run the AI model, the prompts, the knowledge pipeline, every channel connection and the staff inbox yourself; AI Chatbot skips those parts and gives you a working assistant built from the content your store already has. Building fits teams with engineers who need custom behaviour or full control, and most small stores do not need to go that route.
What does it take to build your own store chatbot?
A chatbot that answers from your store's data is much more than one call to an AI model. You have to build and keep running each of these:
- The model: pick or self-host a language model, manage keys, cap spending, and handle errors and overload.
- Prompts and behaviour: write, test and tune prompts, stop the bot from inventing prices or policies, and decide when it must admit it does not know.
- Knowledge: crawl your site, read PDF and DOCX files, split the text, index it and search it (often called RAG), then refresh it when prices and policies change.
- Channels: a website widget, Messenger, Zalo, TikTok Shop, Telegram, each connected in its own way.
- Inbox and handoff: the place where staff read chats, take over, reply and give control back to the bot.
- Operations: logging, quality monitoring, alerts and updates whenever any of the above shifts underneath you.
A team with engineers can do all of it, but each piece is a long-term commitment, not a one-off project.
Building your own vs AI Chatbot, side by side
| Criterion | Build it yourself | AI Chatbot |
|---|---|---|
| Time to a first answer | Time to design, build and test before a customer sees anything | Add your existing content, test in the private preview, often done in a single sitting if your documents are ready |
| Engineering ownership | Engineers build it and keep it running | The usual setup needs no code and happens in the dashboard |
| Knowledge ingestion | Write your own crawler, file readers, OCR and store sync | Import a whole website (recognises Shopify, WooCommerce, Magento and the Vietnamese platforms Haravan, Sapo and Pancake), PDF, DOCX, TXT and CSV files (scanned PDFs read with OCR in Vietnamese and English), written answers; products sync from WooCommerce |
| Channel integrations | Register apps, request permissions, handle webhooks and tokens for each platform | Website, Messenger, Zalo Official Account (Vietnam) and Telegram are ready to connect; TikTok Shop chat is coming soon |
| Handoff and inbox | Build the UI, assignment, notes and live updates | A shared inbox is included: assign, tags, internal notes, snooze, saved replies; the assistant takes a chat back after 15, 30, 60 or 240 minutes, or manually |
| Keeping data fresh | Write and monitor your own refresh jobs | Websites are re-checked weekly or daily depending on the plan; changes to money or time terms (refunds, fees, delivery days) wait for your approval; every version is kept and can be restored |
| Ongoing maintenance | Yours: model API changes, platform changes, bugs, monitoring | AI Chatbot maintains the platform and integrations; you keep your knowledge accurate and current |
| Data control and customisation | Full control over model choice, where data lives and custom logic | Customise the name, tone, language, handoff rule, things never to say and the knowledge; the model and processing flow are managed for you (see security) |
| Cost model | Engineering time, hosting, model usage fees and upkeep | A monthly plan based on AI credits, conversations and knowledge sources, with a free plan (see pricing) |
Why are channel integrations the hard part?
Each platform has its own rules on access, expiring tokens, webhooks and when you may message a customer. Those rules change over time, so someone has to keep watching them.
- Website: the lightest, a single snippet added to your pages.
- Messenger: the Page inbox only, not post comments. A custom app has to pass Meta's permission review.
- Zalo Official Account (Vietnam): an Official Account only, never a personal Zalo. AI Chatbot requires the OA to be verified and on Zalo's Tăng trưởng+ plan.
- TikTok Shop: you register an app with the platform and each seller has to authorise it.
With AI Chatbot the website, Messenger, Zalo Official Account and Telegram connections already exist, and TikTok Shop is coming soon. The integrations page lists each channel.
When is building your own the right call?
Building fits you if you:
- have engineers and need custom behaviour, such as a direct link into an internal order system, ERP or CRM;
- need full control over where data lives or which model runs, for compliance reasons;
- treat the chatbot as part of your core product rather than a support tool;
- need a channel AI Chatbot does not offer, such as Instagram, WhatsApp or Lazada.
AI Chatbot fits you if you:
- have no engineers, or yours are busy elsewhere;
- want customers answered soon and to refine as you go;
- sell through common channels such as your website, Messenger, Zalo Official Account or Telegram;
- mainly need repeat questions answered from existing content, with hard cases passed to your staff.
It depends if you want your own front end but do not want to run models and knowledge yourself. In that case keep your interface and call the public chat API from your server with a workspace API key, while AI Chatbot handles knowledge and handoff. How to retry those calls safely is covered in the client message ID entry.
Where should you start?
Do not decide on paper. Import your website, ask the questions customers raise most, and judge whether the answers are good enough to sit on the front line. Try the live demo, then look at the features. If you want to talk through your own case, contact us. If you are weighing a chatbot against staffing a chat box instead, read AI Chatbot vs live chat widgets.
Frequently asked questions
What does it take to build your own sales chatbot?
You choose and run an AI model, write and tune prompts, build a pipeline to ingest and search your content, connect each chat channel, build an inbox for your staff, and then maintain all of it as platforms change.
Do I need to code to use AI Chatbot?
Not for the normal setup. You add your website or documents, test the answers, connect channels and choose how handoff works, all from the dashboard.
When should I build my own instead of using AI Chatbot?
When you have engineers and need deep custom behaviour, full control over data and models, or a channel AI Chatbot does not support yet. For most small stores a ready-made assistant is faster and less work.
Which channels does AI Chatbot cover so I do not integrate them myself?
Your website, Facebook Messenger (Page inbox), Zalo Official Account (Vietnam) and your own Telegram bot, all feeding one shared inbox. TikTok Shop chat is coming soon. Instagram, WhatsApp and Lazada are not available today, and Shopee chat is not answered because Shopee does not allow chatbot replies.
Can I keep my own front end and still use AI Chatbot?
Yes, through the public chat API with a workspace API key. Your server sends the customer's message and gets back the answer and its sources, while AI Chatbot handles the knowledge and the handoff.
How do the costs of building and using AI Chatbot compare?
Building means engineering time, hosting, model usage fees and ongoing maintenance. AI Chatbot is a monthly plan based on AI credits, conversations and knowledge sources, with a free plan. Check the pricing page against your own volume.