AI chatbot development services for support and ordering
AI chatbot development is building a conversational assistant that answers customers, takes orders or resolves support requests in your own channels. Goldenroz builds customer support chatbots and ordering bots for WhatsApp, web and mobile apps, connected to your order, CRM and help content. Projects start at $25,000, fixed price per phase. A working prototype typically takes one to two weeks, and production roughly four to eight weeks depending on integrations.
From $25,000, fixed price per phase
Updated 2026-11-13. By Shai Goldenberg

Who this is for
- Support leaders whose team answers the same questions every day and who want those resolved without a person.
- Businesses that take orders or bookings by message and lose the ones that arrive after hours.
- Companies whose customers already write to them on WhatsApp.
- Product teams that tried an off-the-shelf bot and hit its limit on integrations or control.
Who it is not for
- Budgets under $25,000. A packaged chatbot tool is the better fit.
- Voice bots and phone call centers. We build text channels.
- Bots meant to hide the way to a person. Every bot we ship has a handoff.
- Medical, legal or financial advice given with no human review.
What is included
- Conversation designThe intents the bot handles, the ones it refuses, its tone and the exact point where it hands over to a person.
- ChannelsWhatsApp through the WhatsApp Business Platform, a web chat widget, or chat inside your mobile app.
- Knowledge base answersReplies drawn from your approved help content and policies. When the content does not cover a question, the bot says so and routes it.
- Actions in your systemsOrder lookup, order creation, booking, status updates and ticket creation through your APIs. See AI integration.
- FallbacksA model fallback chain, timeouts and retries, plus fixed replies that go out when no model responds.
- Handoff to humansTransfer to your team with the full transcript and a summary, triggered by rules, by customer request or by low confidence.
- Logs and cost trackingEvery conversation, tool call and model cost recorded, so you can review failures and see cost per conversation.
- Test setA set of real conversations the bot must pass before every release.
How the work runs
Scoping call
Before the projectWe review real transcripts, pick the intents worth automating and agree what the bot must never do. You get a fixed price.
Prototype
About 1 to 2 weeksA working bot on the top few intents, tested against your past conversations in a private channel.
Production build
Roughly 4 to 8 weeksSystem integrations, fallbacks, handoff, logging and channel setup. The range depends on integrations and on WhatsApp account approval.
Limited launch
At launchThe bot handles a share of conversations or selected hours first. Your team reviews transcripts daily.
Widen
After launchNew intents are added in order of volume, each tested against the conversation set.
WhatsApp chatbot development: what is different from web chat
A WhatsApp bot runs on the WhatsApp Business Platform from Meta. That brings rules a web widget does not have. Your business needs a verified account and number, messages you start outside the customer service window must use templates Meta has approved, and Meta bills messaging to your account.
The conversation is also different. Customers send voice notes, photos, locations and three short messages in a row. A WhatsApp AI chatbot has to wait, combine those messages and reply once. It has to cope with a customer who returns two days later to the same thread.
WhatsApp automation works best for ordering, booking, order status and reminders. The bot confirms the details back to the customer in plain text before it creates anything, and the order lands in the same system your staff already use.
Fallbacks: what the bot does when the model fails
Model APIs time out, return errors and sometimes return nonsense. A customer waiting on WhatsApp sees none of that. They see a business that did not answer.
Every bot we build has three layers. The first model gets a strict time limit. If it fails, a second model takes the request. If that fails, the customer gets a fixed, pre-written reply and the conversation is flagged for a person. Retries are safe to repeat, so a retried request never creates a second order.
Rules sit outside the model where they can. Opening hours, prices, delivery zones and refund limits come from your systems. The model phrases the answer and does not decide the policy.
AI customer service agent with handoff to humans
An AI customer service agent should resolve the routine requests and get out of the way on the rest. Handoff triggers are written down at the start: the customer asks for a person, the topic is on a restricted list, the bot has failed twice, or the request involves money above a set amount.
When a handoff happens, your agent receives the transcript, a short summary and the data the bot already collected. The customer does not repeat themselves. Outside staffed hours the bot says when a person will reply.
We do not promise a resolution rate before reading your transcripts. The share a bot can resolve depends on how many requests are routine and how much your systems expose through APIs. Bots that take wider actions across systems are covered under AI agent development.
What backs this up
- A production WhatsApp ordering botShai built an ordering bot on Claude that runs in production on WhatsApp with a model fallback chain, timeouts, retries and deterministic fallback replies.
- A support assistant with a knowledge baseHe also built a support assistant that answers from a maintained knowledge base inside a live marketplace platform.
- Bots tied to real paymentsThe platform behind those bots handles real money, with an escrow model and a double-entry ledger that uses idempotency keys.
- A named engineerShai, our CTO, designs and builds the bot. You own the code, prompts, transcripts and accounts.
What it costs
Each phase has a fixed price agreed before work starts. Model usage and WhatsApp messaging are billed by the providers to your own accounts.
- Support or ordering botFixed price per phaseFrom $25,000
- Multi-channel or multi-brand platformsScoped per project$100,000 and up
- Scoping callEnds with a written scope and a price$0
- Model and WhatsApp usageBilled to your own accountsPaid to the providers
Who does the work
Shai leads the technology side: business software, SaaS and mobile apps, payments and escrow systems, Postgres security, AI agents built with Claude Code and MCP, and IT and infrastructure projects with senior infrastructure engineers.
Questions buyers ask
How much does AI chatbot development cost?
At Goldenroz, chatbot projects start at $25,000 and are fixed price per phase after a scoping call. Larger multi-channel platforms run $100,000 and up. Model usage and WhatsApp messaging fees are separate and paid to the providers from your accounts.
How long does it take to build a chatbot?
A working prototype on the top few intents typically takes one to two weeks. Production takes roughly four to eight weeks, depending on how many systems the bot connects to. These are typical ranges, and each project is scoped on its own.
Can you build a WhatsApp chatbot?
Yes. We build on the WhatsApp Business Platform, with your verified business account and number. Shai has a WhatsApp ordering bot running in production on Claude.
What happens when the bot cannot answer?
It says so and hands the conversation to a person, with the transcript and a summary. Handoff also triggers when the customer asks for a human, when the topic is restricted or after repeated failures. Outside staffed hours the bot tells the customer when to expect a reply.
How do you stop the bot from making things up?
Answers come from your approved content and live data from your systems. Prices, hours and policies are read from the source, and the bot declines questions the content does not cover. A test set of real conversations runs before every release.
Which model powers the chatbot?
Claude is our default, with a second model as fallback. We also work with OpenAI models and others where they fit. The bot is built so the model can be changed without rewriting the integrations.
Is customer data safe?
The bot can reach only the records of the customer it is talking to, enforced in the database and the APIs. Conversations are logged in your own accounts. For an existing app, a security audit checks what is exposed today.
Should we buy a chatbot tool or build a custom one?
Buy when your needs are standard help-center answers on a web widget. Build when the bot must take orders, act in your own systems, run on WhatsApp with your rules, or when per-resolution pricing gets expensive at your volume. AI consulting covers that decision in detail.
AI chatbots: by industry and use case
- Appointment booking bots
- Chatbots for banking and insurance
- Customer service chatbot development
- Chatbots for ecommerce
- Enterprise chatbot development
- Chatbots for healthcare
- Instagram and Messenger chatbots
- Microsoft Teams and Slack bots
- RAG and knowledge base chatbots
- Chatbots for real estate
- Telegram bot development
- Website chatbot development
- WhatsApp Business API integration
- WhatsApp chatbot development
Guides on this topic
- Nine steps to build a WhatsApp chatbot on the official Cloud API: account, number, webhook, AI layer, handoff, templates and launch. See the full checklist.
- How the WhatsApp Business API works: Cloud API, templates, the 24-hour customer service window, per-message pricing and the setup steps. See what to plan for.
- An AI agent takes actions across your systems. A chatbot answers questions. Compare the two, see examples and use cases, and use the guide to pick one.
Related services
- Custom AI agents that take actions in your systems, with permissions, approvals and logs.
- Connect ERP, CRM, SQL and Microsoft 365 to Claude or ChatGPT through MCP, read-only and audited.
- An assessment that ranks your AI use cases and ends with a build-vs-buy call and a pilot plan.
- iOS, Android and Web apps from one React Native and Expo codebase, submitted to both stores.
Talk to the person who will do the work
One hour, no sales team. You talk to Shai Goldenberg, the CTO, about your systems or your product, and you leave with a clear next step and a price range.