An AI chatbot for SaaS is an AI agent trained on your help centre, product docs and pricing page that answers customer and trial user questions in plain language, at any hour, and hands account or billing problems to your team. It takes the repeat "how do I" questions off the queue, so people handle what only people can.

This guide is for a SaaS founder or support lead who sees the same setup and pricing questions every day and wants them answered without hiring ahead of growth. It covers which questions to hand to the agent, how it compares with the support channels you already run, how to keep it in step with a product that ships every week, and which plan fits. There is a six-point readiness check near the end.

What is an AI chatbot for SaaS customer support?

It is an AI agent that answers from your own documentation instead of from a script. A trial user asks "can I invite a teammate on the free plan?" and the agent answers from your pricing and docs pages, in a sentence, on whichever page they are looking at.

With Dante AI the agent learns from four kinds of source: website URLs it crawls, file uploads such as PDF and Word, typed text, and question and answer pairs. You can mix them up to your plan's memory limit (knowledge sources).

Which SaaS support questions can it take, and which should go to a person?

The handover route matters as much as the answers. With Human Handover on (Advanced and Pro plans), a Talk to a human pill sits under every reply and the conversation lands in your team Inbox, where someone can take it over and reply live. Without it, a visitor who asks for a person is offered a short name and email form by default, on every plan. See how to choose an AI chatbot with human handover for the details that matter.

How does it compare with a help centre, a ticket form or live chat?

ChannelAvailableWho finds the answerCost grows withBest for
Help centre articlesAlwaysThe customer, by searchingWriting timePublishing the answers you control
Ticket form or support emailAlways open, replies in your hoursYour teamTicket volumeAccount-specific problems
Live chat with your teamOnly while someone is onlineYour teamChat volume and hours coveredComplex sales and upset customers
AI agent trained on your docsAny hourThe agent, from your docsPlan and credits usedRepeat how-to and plan questions

The agent sits in front of the other three, not instead of them. Your help centre becomes its source, and your ticket queue and inbox get the conversations that need a person. For the wider build or buy question, read AI agent for customer service.

How do you train it on your product docs and keep it current?

  1. Add your marketing site and your docs site as separate sources. The crawler follows internal links but stays on one domain, so a docs or help subdomain needs its own source. This is the step SaaS teams miss most often.
  2. Upload what is not public. Release notes, onboarding guides or internal FAQs can go in as files or question and answer pairs.
  3. Connect a Google Drive folder if your docs live there. Share it with the address shown on the Knowledge page. On a paid plan it re-syncs on its own: every week on Starter, every 3 days on Advanced and every day on Pro, picking up new and changed files and dropping removed ones.
  4. Place it where questions start. Use the floating launcher across your marketing site, or the Script embed inside a help centre or contact page (embed guide). Add each hostname under Allowed domains.
  5. Read the transcripts every week. In chat history, a Knowledge search step shows which sources the agent matched, so a wrong answer points straight at the page that needs fixing.

What does a first rollout look like? A worked example

Picture a project management tool with a marketing site, a docs subdomain and a pricing page. Week one, the team adds both domains as sources, uploads its changelog as a file, and writes ten question and answer pairs for the plan limits trial users ask about most. Before going live they ask the agent their twenty most common ticket subjects and check each answer against the docs. Two answers are wrong because an old docs page is still public, so they fix the page rather than the agent. They turn on Human Handover for billing and access questions and put the launcher on the pricing page first, where pre-sale questions start. Week two, they read every transcript where the agent could not find an answer and write those docs pages. Run the check below on your own setup.

Which plan fits a SaaS team?

Dante AI prices are per workspace, not per seat: Free $0, Starter $40 a month, Advanced $120 a month and Pro $400 a month, with yearly billing giving two months free. Credits are 100, 2,000, 8,000 and 30,000 a month. Free includes the website embed and 150,000 characters of knowledge, enough to test on a small docs site. Starter adds 25 million characters and weekly knowledge updates. Advanced adds Human Handover, lead capture and 3 AI agents, so a team can run one agent for the marketing site and one for support. Pro has 10 agents and API access. Compare them on the pricing page.

Want to see it answer from your own docs? Get started free at dante-ai.com.

Sources