AI Chatbot Development

Upwork currently lists chatbot developers at approximately $30–$61 per hour.

This represents marketplace hiring rates rather than guaranteed developer income. Actual earnings depend on experience, client acquisition, specialization, project complexity, marketplace fees, and the amount of billable work available.

Project prices vary significantly because the term "chatbot" can describe anything from a basic FAQ widget to a custom AI application integrated with several business systems.

Upwork currently provides examples ranging from small chatbot projects costing a few hundred dollars to substantially larger custom implementations.

A separate Upwork bot-development guide gives examples such as:

These are marketplace pricing examples rather than guaranteed or universal industry prices.

The final price usually depends on:

• Number of integrations

• Knowledge-base size

• Custom frontend requirements

• Authentication

• Human handoff

• Number of communication channels

• Analytics

• Security requirements

• Expected usage

• Maintenance requirements

The developer builds a chatbot for one client and charges a fixed project fee.

Example projects include:

• Website customer support chatbot

• Product FAQ assistant

• SaaS onboarding chatbot

• Lead qualification chatbot

• Internal company assistant

This is the most straightforward service model.

Many companies do not need a completely custom chatbot.

Instead, the developer configures an existing platform such as Botpress or another chatbot builder.

The service may include:

• Account setup

• Knowledge-base import

• Conversation design

• Prompt configuration

• Website installation

• Branding

• Testing

• Human handoff

• Analytics configuration

This can be easier to deliver than building the entire system from scratch.

Businesses often want the chatbot connected to existing software.

Examples include:

• CRM

• Help desk

• Email

• Calendar

• E-commerce store

• Customer database

• Internal API

• Ticketing system

Integration work can significantly increase the value and complexity of the project.

A developer can charge recurring fees for:

• Updating knowledge

• Monitoring conversations

• Fixing incorrect answers

• Improving prompts

• Updating integrations

• Reviewing analytics

• Managing AI costs

• Adding new features

Recurring maintenance can create more predictable revenue than relying only on new projects.

Instead of building a different system for every customer, developers can create a reusable chatbot package for one type of business.

Examples:

• Real-estate inquiry chatbot

• Dental clinic FAQ chatbot

• E-commerce support chatbot

• SaaS onboarding chatbot

• Local service lead qualification chatbot

The underlying system can be reused while the knowledge, branding, integrations, and prompts are customized for each client.

Botpress provides a visual chatbot builder, knowledge bases, integrations, API access, website chat deployment, and AI model support.

Current plans include:

The free pay-as-you-go plan currently includes 500 incoming messages and events per month, one bot, and $5 of monthly AI credit.

Paid plans provide features such as human handoff, watermark removal, larger limits, collaboration, analytics, and additional storage.

Tidio combines live chat, ticketing, automation, and its Lyro AI Agent.

Current published pricing includes:

A developer could use a platform like Tidio when the client primarily needs customer support rather than a completely custom application.

Intercom's Fin AI Agent is another example of usage-based customer-support AI.

Fin currently starts at $0.99 per successful outcome.

Intercom's annual seat pricing currently starts at:

Fin can also be used with some existing help desk systems without purchasing Intercom seats, although minimum commitments can apply.

These products are useful reference points when deciding whether a client needs a custom chatbot or would be better served by configuring an existing customer-service platform.

Do not begin with a generic offer such as "I build AI chatbots."

Choose a specific problem.

Examples:

• Answer repetitive customer questions

• Qualify website leads

• Recommend products

• Help customers find documentation

• Reduce basic support tickets

• Answer employee questions from internal documents

A chatbot is easier to sell when the business benefit is clear.

For a customer support chatbot, gather the information the chatbot is allowed to use.

This may include:

• FAQ pages

• Help center articles

• Product documentation

• Pricing information

• Return policies

• Shipping information

• Internal procedures

• Support documents

The chatbot should retrieve relevant information instead of relying entirely on the model's general knowledge.

Separate questions into categories.

The chatbot can answer:

• Product information

• Opening hours

• Shipping policies

• Basic troubleshooting

• Account setup instructions

The chatbot should escalate:

• Refund disputes

• Legal complaints

• Unusual billing problems

• Sensitive customer information

• Questions without reliable source information

This prevents the chatbot from pretending it knows information that should be handled by a person.

The basic flow may be:

User asks question → Chatbot identifies the request → Relevant company information is retrieved → AI generates an answer → User receives the response

For more advanced systems:

User asks question → Chatbot identifies the request → Retrieves company information → Calls a business system if necessary → Responds or performs an allowed action → Escalates to a human when necessary

A business chatbot should have a clear failure path.

Handoff can be triggered when:

• The chatbot does not know the answer

• The user repeatedly asks for a person

• The issue involves billing or refunds

• The customer is angry

• Required information is missing

• The chatbot's confidence is insufficient

The conversation history should ideally be passed to the human agent so the customer does not need to repeat the entire issue.

Do not test only questions copied directly from the FAQ.

Test:

• Normal questions

• Misspelled questions

• Short questions

• Ambiguous questions

• Questions containing incorrect assumptions

• Questions with no available answer

• Requests that should be escalated

• Attempts to make the chatbot ignore its instructions

Check whether the chatbot gives the correct answer, uses the correct information, and escalates when it should.

Common deployment channels include:

• Website chat widget

• Mobile application

• Customer support portal

• Slack

• WhatsApp

• Other messaging systems

Start with the channel that already receives customer questions.

After launch, review:

• Unanswered questions

• Incorrect responses

• Most common customer questions

• Human handoffs

• Cost per conversation

• Customer feedback

• Broken integrations

Real conversations reveal missing documentation and failure cases that may not appear during development.

Useful skills include:

• Conversation design

• Prompt writing

• Knowledge-base organization

• Chatbot platform configuration

• Basic API concepts

• Testing

• Understanding customer support workflows

These skills can be enough for simple website chatbots.

More advanced projects commonly require:

• Python or TypeScript

• REST APIs

• JSON

• Databases

• Authentication

• Retrieval systems

• Webhooks

• Frontend development

• Cloud deployment

• Logging and monitoring

The difficult part of a production chatbot is usually not generating a response.

It is making sure the chatbot uses accurate business information, handles failures correctly, protects customer data, integrates reliably with other systems, and knows when to hand the conversation to a human.

Client provides:

• Website

• FAQ

• Help center

• Product documentation

Developer provides:

• Chatbot setup

• Knowledge-base configuration

• Prompt and behavior rules

• Website widget installation

• Human handoff

• Basic testing

• Deployment

Optional upgrades:

• CRM integration

• Ticket creation

• Multilingual support

• Product recommendations

• Order lookup

• Analytics

• Monthly maintenance

This gives a new developer a clearly defined service that can later expand into larger projects.

AI models can generate incorrect information.

Important answers should be grounded in approved company information, and the chatbot should avoid answering when reliable information is unavailable.

A chatbot can continue returning outdated information if its knowledge base is not maintained.

Changes to:

• Pricing

• Products

• Policies

• Shipping

• Documentation

should be reflected in the chatbot's knowledge.

Users may intentionally try to make the chatbot reveal instructions, ignore restrictions, or perform unintended actions.

Systems that can access private information or external tools need stronger security controls than a simple FAQ bot.

Chatbots may process names, email addresses, account information, support history, or other customer data.

Developers need to understand what information is collected, where it is stored, which third-party services receive it, and how long it is retained.

A chatbot that costs very little during testing can become more expensive when thousands of users begin using it.

Monitor:

• AI token usage

• Platform limits

• External API usage

• Storage

• Hosting

• Messaging fees

before agreeing to a fixed monthly client price.

A chatbot is rarely completely finished after launch.

New questions, changing company policies, model updates, integration changes, and unusual customer behavior create ongoing maintenance work.

This can be a risk for the developer, but it can also support a recurring maintenance business model.

AI chatbot development is the process of creating conversational software that understands user questions and generates responses using AI models, business information, rules, and integrations.

A chatbot is primarily designed around conversation.

An AI agent is generally designed to complete tasks and may autonomously choose tools or actions as part of a larger workflow.

The two categories can overlap. A chatbot can include agent-like capabilities such as updating CRM records or performing account actions.

Not necessarily.

Platforms such as Botpress allow developers to create many chatbot workflows visually.

Programming becomes more important when a project requires custom interfaces, authentication, unusual APIs, complex business logic, or large-scale deployment.

A prototype can be built with free software and free platform tiers.

Botpress currently offers a $0 pay-as-you-go plan with a limited monthly allowance and $5 in included AI credit.

Production usage may introduce AI, platform, hosting, integration, and messaging costs.

Usually not.

Most independent chatbot developers use existing AI models through APIs and connect them to the client's knowledge and business systems.

No dedicated GPU is required when using hosted AI APIs.

The AI model runs on the provider's infrastructure.

Upwork currently lists chatbot developer rates at approximately $30–$61 per hour.

Fixed project prices vary substantially depending on whether the project is a simple FAQ bot or a custom system with AI, knowledge retrieval, authentication, multiple integrations, and human handoff.

There is no universal development time.

A basic chatbot built on an existing platform can be much faster to deliver than a custom chatbot with authentication, CRM integration, private customer data, analytics, and multiple communication channels.

Scope should be defined before quoting a delivery date.

A website knowledge or FAQ chatbot is one of the simplest starting points.

It provides a clear input source, limited scope, easy demonstration, and relatively few integrations.

Yes.

A developer can expand from individual chatbot builds into:

• Chatbot strategy

• Implementation

• Integration

• Knowledge-base preparation

• Customer-support automation

• Monthly maintenance

• Analytics and optimization

Reusable components can also make similar projects faster to deliver across multiple customers.