AI Chatbot Developer: Chatbots for Websites, WhatsApp and Support

What AI chatbots have I built?

Three projects show my AI chatbot and copilot work, and each links to its case study. The MagizAI details also come from magiz.ai and its WHMCS Marketplace listing as I read them on 19 October 2026.

Disclosure: MagizAI is my own product. I am its founder and CEO, so I gain if you buy it. ElySpace Ask AI belongs to my employer, where I am CTO. I list it as proof, not as a pitch: it is a chatbot I designed and built myself, and anyone can check it at magiz.ai. A custom build and MagizAI solve different problems, and if the ready-made product covers your need, I will say so on the scoping call.

Which AI models and tools do I use?

For custom builds I use the OpenAI and Anthropic APIs, and I pick the model by testing it on your own questions, not by brand. The rest of the list is what my shipped projects use; I do not list tools I have not shipped with.

Can you build a chatbot trained on your own documents?

Yes, though "trained" is the sales word. What I build is retrieval: the bot looks up your documents at question time, and the model is never retrained on them. The technical name is retrieval-augmented generation, or RAG.

1. Collect and clean. I gather your pages, help articles, PDFs and policies, and we remove outdated pages and duplicate policies so each fact has one source.
2. Chunk and index. I split the text at headings and turn each chunk into an embedding, a list of numbers that captures its meaning, kept in a vector store with metadata such as product and language.
3. Retrieve. The chunks closest to each question are sent to the model with it. Below a similarity threshold the bot does not answer; it says it does not know and offers a person.
4. Answer with the source. The reply links its source page, and the log keeps the chunks used and the cost.
5. Refresh. When a page changes, its chunks are indexed again. An outdated page gives an outdated answer, so the documents matter more than the model.

Rules that matter, such as never quoting a price or discount that is not on your published pages, are enforced in code as well as in the system prompt, because a prompt can be argued with. Visitor text that tries to change the rules, known as prompt injection, is part of the test set.

Can you build a WhatsApp AI chatbot?

Yes. MagizAI, my own product, answers WhatsApp through Meta's official WhatsApp Cloud API: the bot uses the same content and hand-off rules as the website chat, and WhatsApp conversations land in the same team inbox, where a person can take over.

WhatsApp differs from a website widget in ways I plan for from the start. I connect the number through your own Meta business account, so the number and its message templates stay yours. Meta's rules decide when a business may write first and when it must use a pre-approved template, and campaigns need the customer's opt-in. Customers also write shorter messages and send photos and voice notes, so the hand-off triggers are tuned on real WhatsApp chats. If MagizAI already covers what you need on WhatsApp, I will say so before quoting a custom build.

Can an AI chatbot work with WHMCS and hosting support?

Yes, and two other pages cover it in depth, so this is the short version. A WHMCS chatbot answers a signed-in client from their own services, invoices and tickets, with the identity taken from the client-area session and never from what is typed, and with access limited to the actions you allow.

WHMCS AI chatbot compares a ready-made module, an n8n workflow and a custom build. MagizAI for hosting providers describes my own module, which needs no WHMCS admin API key and keeps every account action switched off until you turn it on and the client confirms it. For running chat in a hosting company around the bot, read live chat for web hosting companies.

Disclosure: besides MagizAI, I founded WHMCSPilot.com, which sells WHMCS modules and themes, and I am CTO at ElySpace, a hosting company.

Can you add an AI chatbot to my website or Laravel app?

Yes. On most websites the visible part is one line of code that loads a chat widget; the real work is on the server, where an endpoint fetches your documents, calls the model, enforces the limits and writes the logs.

• Any website. The widget script goes before the closing body tag and should run inside a Shadow DOM, an isolated part of the page, so your site's styles and the chat cannot break each other. MagizAI's widget is built this way.
• A Laravel application. The chatbot can live inside your app as a route, a queued job that indexes documents and a few tables, and use your existing login, so it knows which customer is asking. It needs no separate service unless traffic calls for one.
• WordPress. A small plugin adds the widget without editing the theme. MagizAI already has a WordPress plugin.
• Your own SaaS product. If the chatbot is a feature you sell to your customers, each customer's documents must stay separate from the others'; see SaaS development.

AI chatbot or AI agent: which do you need?

Choose by what the visitor needs done: find an answer, ask in their own words, or get something changed. A chatbot answers from your documents and its only write is the hand-off ticket; an agent is given tools that act in your systems.

If a good FAQ page would do, I will say so. AI chatbot vs AI agent compares the trade-offs in depth, and AI agent development covers bots that act.

How much does it cost to hire an AI chatbot developer?

With me, building a custom AI chatbot that answers from your documents is about 35 to 70 hours by my planning estimate, which is $1,925 to $4,550 at $55-$65 an hour, or a fixed quote. Running it is a separate monthly bill that the model provider charges to your own account.

The hours are planning estimates, not quotes. What an AI chatbot or agent costs breaks them down task by task and shows the token arithmetic. Document cleanup varies most: a tidy help center sits at the low end, while PDFs, conflicting policies or a second language push the hours up. After a free scoping call of about 30 minutes I send a fixed quote within one business day, as on how I work.

Three limits keep the running cost predictable: a cap on messages and reply length per conversation, a per-visitor rate limit, and a daily spend cap with an alert before it is reached. From the first weeks of logs I report the real cost per conversation.

How long does AI chatbot development take?

A chatbot that answers from your documents is 35 to 70 hours of my work. The calendar time depends mostly on how quickly the questions and documents arrive, and the written quote gives you a date.

For scale: a whole WhatsApp marketing platform I built earlier, in which the AI chatbot was one part, took five weeks. A WhatsApp, WHMCS or helpdesk connection adds hours to a chatbot; tools that change data turn it into the agent row in the cost table.

What happens to my data and my customers' data?

The visitor's message, the recent conversation and the document chunks needed for the answer go to the model provider's API, and nothing else should.

• Redaction. I mask email addresses, phone numbers and card-like numbers before text is sent or logged, unless the answer needs them. Pattern matching misses some formats, so retention and access limits matter too. Card data, passwords and API keys are kept out of prompts, and text a visitor types is masked on a best-effort basis.
• Retention and access. Logs are kept for a period you choose, then deleted, and only named staff accounts can read them.
• Provider terms. OpenAI and Anthropic publish data-use terms for their business APIs. Read the current ones for your plan.

The widget tells visitors they are talking to an AI. I do not claim a chatbot makes you compliant with any law; ask your own adviser what applies.

What should I ask any AI chatbot developer before hiring?

Put these seven questions to me or to anyone else.

• Can I see a chatbot you built that is running now, and what was your part in it?
• Which model is used, and can it be switched without a rebuild?
• What does the bot say when it does not know?
• Can I see test results on my own questions before launch?
• What does one conversation cost, and what stops a runaway bill?
• Who reads the logs, and for how long are they kept?
• Who owns the code, prompts and accounts at handover?

What chatbots will I not build?

I will not build a chatbot that misleads people or answers where a wrong answer costs too much.

• One that pretends to be a human. It says it is an AI, and a visitor can always ask for a person.
• Deceptive or spam uses: fake reviews, impersonation or mass unsolicited messages.
• Bots that give medical or legal advice to the public. The OnDemand Psychiatry work was a copilot for clinicians, not a bot that advises patients.
• Any bot sold on a promised resolution or deflection rate. Those depend on your documents and customers, so I measure them and promise nothing.
• A chatbot that refunds, cancels or changes accounts without a person confirming. Actions belong to agent work, with narrow permissions and approval.