Quick summary
What AI chatbot development services build: a chatbot that answers your customers' questions in plain language, usually from your own help docs (a method called RAG), with a handoff to a person when it cannot help. For a SaaS product, the best ones also know the customer's account, plan and settings.
Build or buy first: if it only needs your help docs, an off-the-shelf tool such as Intercom's Fin ($0.99 per outcome) or Zendesk's AI agents is usually cheaper. Pay for custom chatbot development when it must live inside your product or answer from each customer's own data.
What it costs: vendor guides from 2026 put a custom customer support chatbot at about $15,000 to $40,000 (RaftLabs) and a RAG-powered chatbot at $50,000 to $150,000 (Space-O), plus $400 to $6,000 a month to run.
What to check: a chatbot they built that real customers use, how it hands off to a person, how one customer's data stays out of another's answers, and that the code and model accounts end up in your name.
AI Chatbot Development Services at a Glance
| Where to get it | What it costs | How fast | Best for |
|---|---|---|---|
| Off-the-shelf support bot (Intercom Fin, Zendesk AI agents) | Intercom Fin $0.99 per outcome, plus seats or a monthly minimum | Days to set up | Answering questions from your help center |
| A developer who works in your product | Their usual rate or a flat monthly fee | Weeks, in small pieces | A chatbot that needs your app's data and rules |
| Freelance chatbot developer (Upwork) | About $10 to $100 an hour for software developers | Days to start | One clearly scoped chatbot, if someone can review the work |
| Vetted network (Toptal) | No public rates | Average match under 24 hours | Screened candidates at a higher price |
| AI chatbot development company | About $15,000 to $40,000 for a customer support chatbot (RaftLabs) | 8 to 16 weeks | A large or multi-channel chatbot built by a team |
| Full-time AI engineer | A median $135,980 a year for US software developers, before benefits | Months to hire | Years of AI work ahead |
Before you hire anyone, answer these three questions:
- Question 1, "Does it only answer from our help docs?" If yes, try an off-the-shelf tool first. If it must know the customer's plan, usage or invoices, you are looking at custom work.
- Question 2, "What must it never say?" Refunds, discounts, legal answers and anything about another customer's account. Write the list down before anyone builds.
- Question 3, "How does a customer reach a person?" Every chatbot needs a clear way out. Decide who receives the handoff and how fast they reply.
Want an AI chatbot built into your SaaS?
A senior developer on subscription: $2,995 a month, as many requests as you like, each shipped within 2 business days, and your money back if the first 14 days don’t prove it.
In this article
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
This guide is for founders of SaaS products who want an AI chatbot that answers customer questions, inside the app or on the help center, and are deciding whether to buy a tool, hire a developer or pay an agency. It covers what a chatbot is in plain English, when an off-the-shelf tool will do, what custom chatbot development costs to build and run, and how to choose who builds it. If you want the AI to take actions, such as issuing refunds or changing settings, that is an agent, and our guide to AI agent development services covers it.
What an AI Chatbot Is (and What It Is Not)
An AI chatbot reads a question written in everyday language and writes an answer. Modern ones use a large language model (LLM), the same kind of AI behind ChatGPT and Claude, so they can understand questions phrased in ways nobody planned for.
A chatbot answers
It replies to questions, usually from your help docs. It does not change anything in your systems.
An account-aware chatbot answers about this customer
It can also read the customer's own data, such as their plan, usage or last invoice, so it can say "your plan includes 5 seats and you are using 4" instead of quoting the pricing page.
An agent acts
It decides which steps to take and uses tools to do them, such as cancelling a subscription or issuing a refund. More useful, and more risky, because a wrong decision changes real data.
Most SaaS founders want the first or second kind. Many vendors sell the third under the same name, so ask which one you are being quoted for.
Common AI chatbots in a SaaS product
- Help-center chatbot. Answers "how do I" questions from your docs and hands off to support when it cannot.
- In-app assistant. Sits inside the product and answers about the customer's own account, settings and data.
- Onboarding guide. Walks new users through setup and points them to the right page.
- Internal knowledge assistant. Helps your own team find answers in docs, past tickets and specs.
RAG development services: how a chatbot learns your product
Out of the box, an AI model knows nothing about your product. RAG, short for retrieval-augmented generation, fixes that. Before the AI writes an answer, your system searches your help docs (and, for an account-aware chatbot, the customer's data) for the most relevant passages and hands them to the model with the question. The answer is then based on your content, and a good chatbot shows which page it came from.
Agencies that sell "RAG development services" are selling this part: preparing the docs, storing them in a searchable form (often a vector database), choosing what to retrieve and checking that answers stay faithful to the source. Almost every useful support chatbot needs it, so treat it as part of the build, not an extra.
Buy First? Off-the-Shelf Chatbots vs Custom Chatbot Development
Search results for AI chatbot development services are almost all agencies, and most of them sell a custom build. For many SaaS products, the cheaper first step is a tool you switch on.
Intercom's Fin
Intercom prices its Fin AI agent at $0.99 per outcome and charges once per conversation, even if Fin takes several actions. An outcome includes the customer confirming their issue is solved or not asking for more help after Fin's answer, so a customer who simply leaves counts. Intercom says Fin also works with other help desks, including Salesforce, with no seat costs but a minimum monthly commitment.
Zendesk AI agents
Zendesk measures and bills its AI agents by automated resolutions, a conversation the AI handles without a person. Zendesk's own documentation says resolutions bought above your allotment cost more per resolution than ones you commit to in advance.
The chat tool you already use
Many help desks and chat widgets now include an AI answer feature trained on your help center. Check what your current tool offers before paying anyone to build.
An off-the-shelf chatbot suits you when
- It only needs to answer from your help center
- You already use Intercom, Zendesk or a similar help desk
- You want it live this month without a build
- Your support volume is modest
Custom chatbot development suits you when
- It must answer from each customer's own account data
- It has to live inside your product, not a chat widget
- Per-resolution fees would cost more than running your own
- It must follow product rules a help desk tool cannot see
A rough cost comparison, by our math
Take a help-center chatbot handling 2,000 conversations a month. If 1,200 of them count as Fin outcomes, Intercom's fee is about $1,188 a month (1,200 x $0.99), plus seats, with no build cost.
A custom chatbot pays the AI model directly instead. Anthropic's Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens (a token is roughly three quarters of a word). Assume a conversation of 4 messages, each sending about 4,000 tokens (instructions, retrieved help-doc passages and the chat so far) and getting back about 300. That is about 16,000 input tokens ($0.032) and 1,200 output tokens ($0.012), or roughly $0.044 a conversation and $88 a month for 2,000 conversations. Hosting, a vector database and monitoring come on top; Monocubed puts cloud hosting at $200 to $1,000 a month and vector databases at $50 to $500.
So a custom chatbot can be cheaper to run at volume, but it starts with a build bill of $15,000 or more and needs someone to maintain it. At modest volume, buying first usually wins. The case for custom is rarely the fee; it is what the chatbot needs to know.
What AI Chatbot Development Services Include
Whoever builds it, check that each stage is covered:
- 1
Step 1, Decide what it answers and what it never says
List the questions it should handle, the topics it must refuse (refunds, discounts, legal or medical advice) and when it hands off to a person.
- 2
Step 2, Prepare the knowledge
Clean up the help docs it will answer from, fix outdated pages, and decide which customer data, if any, it may read.
- 3
Step 3, Build the retrieval and the chatbot
Store the docs in searchable form, choose the model, write the instructions and connect it to your app or help desk.
- 4
Step 4, Connect the handoff
Route conversations it cannot handle to your support inbox or help desk, with the chat history attached so the customer does not repeat themselves. Our guide to API integration services covers connecting tools like these.
- 5
Step 5, Test, launch and keep it accurate
Test it against real past questions, review a sample of conversations every week after launch, and update the knowledge whenever the product changes.
What AI Chatbot Development Services Cost
Prices vary widely because "AI chatbot" covers everything from an FAQ widget to a voice assistant. These figures come from vendor guides, so use them as budgeting ranges, not quotes.
The guides disagree, which tells you how much depends on scope. RaftLabs puts a simple rule-based chatbot at $4,500 to $12,000 and an AI chatbot with LLM integration at $24,000 to $60,000 over 8 to 14 weeks. Monocubed, updated in May 2026, prices a generative AI (LLM) chatbot higher, at $75,000 to $300,000 over 12 to 20 weeks, and an enterprise agentic chatbot at $150,000 to $500,000 or more.
Running costs: the part many quotes leave out
The build is the first bill, not the last. Monocubed puts yearly maintenance at 15% to 20% of the build cost, and knowledge base updates and retraining at $5,000 to $20,000 a year. It adds that a high-volume RAG chatbot can reach $2,000 to $3,000 a month in model API fees alone. RaftLabs puts infrastructure at $100 to $500 a month for hosting and databases, and adds 5 to 10 hours a month of someone reviewing conversations.
Where to Get AI Chatbot Development Services
A developer who works in your product
An account-aware chatbot is mostly ordinary product work: reading the customer's data safely, respecting their permissions and showing answers in your own interface. A developer who already works in your codebase, or one who joins it through a development subscription, can build it in small, tested pieces: first a help-doc chatbot, then account answers, then the handoff. Our guide to AI integration services covers other AI features built the same way.
Best for: a live SaaS product that wants the chatbot inside the app.
Hire a chatbot developer as a freelancer
Freelance marketplaces list plenty of chatbot developers. Upwork's cost page puts software developers at about $10 to $100 an hour, and that spread reflects real differences in experience. On an open marketplace you do the vetting, so ask for a chatbot they built that real customers use and start with one small paid job. Our list of Upwork alternatives compares other marketplaces and their fees.
Best for: one clearly scoped chatbot, with someone who can review the work.
Vetted networks
Toptal has a dedicated page for chatbot developers. It says typically fewer than 3% of applicants are accepted, the average match takes under 24 hours, you can usually hire in about 48 hours, and there is a trial of up to two weeks. It publishes no rates; our Toptal review explains how its pricing works.
Best for: founders who want screened candidates and can pay above marketplace rates.
An AI chatbot development company
An AI chatbot development company brings a team and a process: discovery, design, build, testing and often a project manager. Agencies suit a large chatbot across several channels (web, email, WhatsApp, voice) or a company with no developer at all. Their case studies and client counts are self-reported, so ask for references you can call. Our guide on how to choose a software development company covers how to vet one.
Best for: a large or multi-channel chatbot with a fixed deadline.
A full-time AI engineer
If AI features are central to what you sell and there are years of work ahead, an employee builds deep knowledge of your product and data. The US Bureau of Labor Statistics puts median pay for software developers at $135,980 a year (May 2025), before benefits and recruiting.
Best for: a product with more AI work than one part-time person can handle.
The Two Mistakes That Cost the Most
1. The chatbot promises something you never offered
The amount was small; the lesson is not. A chatbot that answers from general knowledge, or improvises when your docs are silent, will eventually invent a policy. Keep it answering only from your own docs, tell it to say "I don't know" and hand off when the docs do not cover a question, and keep refunds, discounts and contract terms out of its reach.
Customers also want the exit. In a Gartner survey of 5,728 customers published in July 2024, 64% said they would prefer companies did not use AI for customer service, and the top worry, named by 60%, was that AI would make it harder to reach a person. A visible "talk to a person" button costs nothing and answers that worry.
2. One customer's data shows up in another customer's answer
An account-aware chatbot reads data from many customers who share one system. If the search step does not check whose data it is pulling, an answer for one customer can quote another's.
Put these in any brief:
- Filter by customer before the AI sees anything. The search step should only ever return documents and data that belong to the person asking, enforced in your server code, not in the instructions to the model.
- Keep shared and private knowledge apart. Public help docs can be shared; account data and private files stay partitioned by customer.
- Treat customer-supplied text as untrusted. OWASP's prompt injection entry warns that instructions hidden in "external sources, such as websites or files" can change the model's behaviour. A chatbot that reads uploaded files or tickets must not follow orders written inside them.
- Log and review conversations. Spot-check answers every week, and look for any that mention data the customer should not see.
Ask every candidate how they enforce the first point. A strong answer names where the check lives in the code and how it is tested. If your app was built quickly with an AI tool, our guide to vibe coding cleanup explains how to check its permissions before adding a chatbot on top.
How to Choose an AI Chatbot Development Company or Developer
Questions to ask any chatbot provider
- Can you show me a chatbot you built that real customers use today, and what share of questions it resolves?
- Would an off-the-shelf tool such as Intercom's Fin or our help desk's AI do this job? If not, why not?
- What will it answer from, and what happens when the answer is not in our docs?
- How does a customer reach a person, and what does our support team see when they do?
- How do you stop one customer's data from appearing in another customer's answer, and how is that tested?
- What will the model, hosting and upkeep cost each month at our volume?
- Whose accounts hold the code, the prompts, the data and the model API keys?
Red flags
- A custom build quoted before anyone asked about off-the-shelf tools. A good provider tells you when a tool will do.
- No answer for "what if it doesn't know?" A chatbot that always answers will eventually make something up.
- Data rules handled "in the prompt". Telling the model not to share other customers' data is not access control; the filter belongs in code.
- Accounts in their name. Your company should own the code repository, the vector database and the model accounts, with the developer added as a user.
Which Route Fits You?
Switch on Intercom's Fin, your help desk's AI agent or your chat tool's AI feature. Pay per resolution and skip the build.
It is built on your existing code, permissions and data from the start, and ships in small pieces you can test with real customers.
The fastest start. Ask for a live example and keep the accounts in your name.
A full team and process at a higher price. Get the running costs in writing.
Hire when there is more AI work than one part-time person can handle. Until then, the salary buys more than you can use.
If the chatbot needs new data or endpoints behind it, see our guide to backend development services.
How StackTie Fits
StackTie is a development subscription for founders who have a web app and more to build than people to build it. An AI chatbot inside your product is the kind of work you can send us, in pieces: answer questions from your help docs, add answers about the customer's own plan and usage, hand off to your support inbox with the chat attached, and log conversations so you can review them. You add requests in plain English to your own Trello board, and we work them one at a time in the order you set. Every request comes back within 2 business days as a pull request with a plain-English summary of what changed and why. One senior developer, Michael, leads every build. Everything we build belongs to your company and lives in your own accounts, including the model API keys. If an off-the-shelf tool will do the job better, we will tell you. There is no contract, you can pause when the work is quiet, and there is a money-back guarantee if it is not a fit. We do not build native iOS or Android apps, we are not a whole team working in parallel, and we do not provide 24/7 on-call support or compliance audits. The flat monthly price and what is included are on the homepage.


