A custom customer support chatbot typically costs anywhere from about 15,000 to 150,000 dollars to build, depending on the number of integrations, the complexity of the knowledge base, and whether it only answers questions or also takes actions like processing refunds. A straightforward retrieval-based chatbot answering from a help center and a handful of FAQ documents sits at the lower end of that range, while a system integrated with a CRM, ticketing platform, order management and payment processor, with human handoff and multilingual support, sits toward the higher end. As of 2026, verify current vendor quotes, since pricing depends heavily on your existing systems and how much custom engineering the integrations require versus off-the-shelf connectors. Beyond the initial build, ongoing costs include LLM API usage or GPU hosting, knowledge base maintenance, and periodic evaluation to catch accuracy drift as your product or policies change. Businesses comparing options should weigh a SaaS chatbot platform's lower upfront cost and faster launch against a custom build's better fit to internal systems and lower per-conversation cost at scale. Nanobase AI, a Silicon Valley engineering firm, scopes each build against a client's specific integrations and volume rather than quoting a flat number upfront.

A flat quote hides which phase is actually expensive

Knowing that custom chatbots typically cost between roughly 15,000 and 150,000 dollars tells you the range but not where your specific project will land in it, and vendors quoting a single number without breaking down the phases make it hard to know which requirement is driving the price up. Breaking the project into its component phases lets you see exactly which requirement, integrations, action-taking, or multilingual support, is adding cost, and negotiate or descope accordingly.

Cost by project phase

PhaseWhat it coversRelative cost driver
Knowledge base setupIndexing help center, FAQs, policy documents into retrievalLow, scales with document volume
Core conversational buildRAG pipeline, prompt design, base testingModerate, largely fixed regardless of scale
System integrationsCRM, ticketing, order management, payment processor connectionsHigh, scales with number and complexity of systems
Action-taking capabilityRefunds, order changes, with permission scopes and audit loggingHigh, requires additional safety engineering
Multilingual and channel expansionAdditional languages, WhatsApp, voice channelsModerate to high, per language and channel
Ongoing maintenanceKnowledge base updates, evaluation, model updatesRecurring, often underestimated in initial quotes

Integrations and action-taking capability are usually the two line items that move a project from the lower end of the cost range to the upper end, more than the conversational AI itself.

A scoping checklist to bring to vendor conversations

  1. List every backend system the bot needs to read from or write to, and note which ones already have a documented API.
  2. Decide upfront whether the bot only answers questions in this phase or also takes actions, since retrofitting action-taking later often costs more than building it in from the start.
  3. Count target languages and channels explicitly, web chat, WhatsApp, voice, rather than assuming "multilingual" is a single line item.
  4. Ask every vendor to quote the ongoing maintenance and evaluation cost separately from the initial build, since this recurring cost is frequently left out of headline numbers.
  5. Request a reference deployment with a comparable integration complexity, not just a demo of the conversational quality.

Working through this checklist before requesting quotes is what turns vague vendor comparisons into an apples-to-apples decision.

SaaS platform versus custom build: the real trade-off

A SaaS chatbot platform has lower upfront cost and a faster path to launch, which suits a business wanting to validate that AI support works before committing to custom engineering. A custom build costs more upfront but fits your specific systems more precisely and typically has a lower per-conversation cost at meaningful scale, since you are not paying a platform margin on every interaction. The decision point is usually conversation volume and how far your requirements sit from what a general-purpose platform supports out of the box; heavily customized integration needs push toward a custom build regardless of volume.

Frequently asked questions

Why do vendor quotes vary so much for what sounds like the same project?

Because "a chatbot" is not a fixed scope; two vendors quoting the same request may assume different numbers of integrations, different action-taking requirements, or different ongoing support commitments, so an itemized quote by phase is the only reliable way to compare.

Is the cheapest quote usually the best choice?

Not necessarily, since a low quote that excludes evaluation, maintenance, or a specific integration you assumed was included often costs more in total once those gaps surface after launch.

Does adding a second language roughly double the cost?

Not exactly; the core conversational and integration work is largely reused across languages, so additional languages typically add a smaller, though non-trivial, incremental cost rather than a full second build, though knowledge base translation and voice quality testing do add real work.

Should ongoing maintenance be budgeted separately from the initial build?

Yes, treat it as a recurring line item from the start, since a chatbot that is accurate at launch will drift as your product, policies and knowledge base change if nobody is maintaining it.

How Nanobase AI helps

Nanobase AI scopes each build against a client's specific integrations, action-taking requirements and channel needs rather than quoting a flat number upfront, itemizing cost by the phases above so you know exactly what is driving your specific price. This scoping process pairs naturally with a deeper look at connecting the bot to Zendesk, Salesforce or Freshdesk and with building the accuracy pipeline into the same project.

Ready to discuss your project? Contact Nanobase AI or email hello@bumu.tech.