Customer support chatbot

The AI customer support chatbot that answers, acts and hands off.

A customer support chatbot and AI support agent for your website, email, phone and WhatsApp. It answers from your help center and docs with citations, takes approved actions, and hands complex requests to your team with the conversation attached.

Free to start. No credit card. Live in minutes.

For online stores, the Shopify AI chatbot also connects product discovery, cart actions, and order lookup.

Resolution rate
72%
Answers with a cited source
Every one
Denser AIOnline

Hello, I am Denser assistant. How can I help?

Do you ship to Canada, and how long does it take?

Yes, we ship to Canada. Standard delivery is 5 to 8 business days, and express is 2 to 3 days.Source: /shipping-policy
  • Do you offer free returns?
  • Track my order

Can I talk to a person about order #48213?

Talk to a personConnected

Ava from Support joined the conversation

Hi, I have the full conversation. Let me check order #48213 for you.

Conversation context shared

Video transcript

Three customers helped at 2 a.m., nobody on shift. It's two in the morning. Three customers need help, and nobody's on shift. Your AI agent answers, takes care of it, and brings in your team only when it has to. Here's how, with Denser, in three parts: answer, act, and hand off.

Answer: from your help articles, with sources. First, answer. Upload the help articles you already have. Your agent answers with your own steps, and cites the article. Nothing made up.

Act: book calls and explain charges from Stripe. Second, act. Choose what your agent can do. It shows open times from your calendar, and books the one the customer picks. It checks who's asking first, then explains the charge from Stripe.

Hand off: your team takes over from the Inbox. Third, hand off. Some things need a person, like a refund. Your team sees the whole conversation in the Inbox, and takes over with one click. The reply lands in the same chat, so the customer never starts over.

Every channel in one inbox. Website chat, email, phone and WhatsApp all land in that same inbox. Build your AI support agent, free at denser.ai.

Watch on YouTube ↗

Customer proof

Trusted by businesses worldwide

  • 2k+active teams
  • 24/7support coverage
  • 50%fewer tickets
  • City of San José
  • Bridgelux
  • Onondaga County
  • EOS Australia
  • LenderDock
  • Portfoliobox
  • Upsiide
  • Valk Welding

How it works

What is an AI customer support chatbot?

An AI customer support chatbot understands a customer's request, finds the answer in your approved company knowledge, and decides whether to answer, take an action, or involve a person. Unlike a scripted, rule-based bot, it handles natural follow-up questions without a branch for every phrasing. Comparing vendors? See the best AI agents for customer service.

AnswerActEscalate

  1. 01

    Understand the request

    The agent identifies the customer's intent and keeps the conversation context across follow-up questions.

  2. 02

    Find an approved answer

    It retrieves the most relevant passage from your website, help center, or documents and cites the source.

  3. 03

    Resolve, act, or escalate

    It answers routine questions, runs an approved action, or sends the full conversation to a person.

Pick a channel

Deploy your customer support chatbot on every channel.

Choose where customers reach your support chatbot. Turn on the channels you need from the same dashboard, and each one uses the same knowledge base.

Website chat

A floating bubble on any page, or an inline frame that sits inside your layout.

How to set it up
Paste one script tag for the bubble, or an iframe for the inline version. On WordPress, install the Denser Chat plugin and write no code at all.

Every channel writes into the same chat logs, and any conversation the AI cannot finish hands off to your team. See how Helpdesk works

Denser AI chat widget answering a customer's return policy question with citations, linked to a Sources panel listing website, docs, and help center

Your knowledge, cited

Train your customer support chatbot and go live in 5 minutes

Point Denser at your website, PDFs, and docs. It answers with citations, scales to your whole knowledge base, and tells you exactly where it still needs help.

  • Live in 5 minutes with script, iframe, React component, or REST API.
  • Every answer links to the exact source passage it came from.
  • Scale from a small help center to a large knowledge base.
  • Flag unresolved questions and catch contradictions across sources.
  • Collect names, emails, and qualified intent from every conversation.

Use cases

Customer support chatbot use cases

Start with requests that have an approved answer or a bounded action. Keep exceptions, sensitive cases, and judgment calls with your team.

  • Product and policy questions

    Answer setup, shipping, returns, billing, and feature questions from approved documentation.

  • Technical troubleshooting

    Guide customers through documented fixes and escalate when the issue needs account access or engineering judgment.

  • Order and account workflows

    Check who the customer is, then look up the order in Shopify, the invoice in Stripe, or the record in your own API. Protected changes go to your team.

  • Lead and onboarding assistance

    Recommend the right resource, capture qualified details, and book a meeting on your calendar without leaving the conversation.

Actions

Support chatbot actions: bookings, orders, billing and Slack

Describe when each action should run, in plain words with a few example messages. The agent runs it for real, then answers from what came back: an order's status from your API, an invoice from Stripe, a meeting on your calendar.

Every action has its own test chat, so you can try it before customers see it. Anything that changes something, like cancelling an order, waits for the customer to say yes.

Start with Talk to a person, so every customer has a clear path from AI to your team.

See every action and template

ActionsChoose what the agent can doReady
  • Call your API

    Paste a URL, cURL command, or code snippet from your API docs, test it, and choose the fields the agent may read.

  • Book in chat

    Offer open times from your Cal.com calendar and book the one the customer taps.

  • Look up billing

    Answer invoice and subscription questions from Stripe after an identity check. Read only.

  • Notify your team in Slack

    Post a call-back or other request to the channel you choose, with a link to the chat.

  • Run your Shopify store

    Search products, check orders, and cancel or start a return once the shopper confirms.

  • Forms, links, and tickets

    Collect details, open the right page, or file a ticket in Zendesk, Freshdesk, Intercom, or HubSpot.

Cards

Answers customers can scan, not a wall of text.

When the answer is a list of products, an order, or a set of plans, the agent shows it as a card: a small layout your team designs once and your actions fill in.

  • Tables, lists, prices, buttons, and short forms, designed once in a visual editor
  • Filled in by the action that shows them, from your API's answer or the conversation
  • A one-line reply above the card, so the agent never reads the list out twice
The Denser test chat answering 'Do you sell snowboards?' with three product cards, each with a photo, name, and price, one marked Sold out. The trace under them reads 'Action: Shopify, Search products: ran'.
Search results as product cards, with the trace showing which action produced them.

Playbooks

Your procedures, followed step by step. And you can watch it.

A playbook tells your AI support agent how to handle one situation, step by step, the way you would brief a new teammate: what to ask, which action to run, and what to do when the conversation goes one way or the other. Build it from Say, Do and If blocks, or describe it in a sentence and let Ask AI draft it.

A playbook run in the Denser test chat. The reply's trace reads 'Playbook: Bug report troubleshooting, Now at: Try fixes from the docs one at a time', and the same Say / ask step is highlighted and marked Now in the flow editor beside it.
A real test run: the trace under the reply and the highlighted step in the flow point to the same block.
  1. 01

    Every reply names its step

    Under each reply in the test chat, a trace shows the playbook, the step the agent is on, the path it took at each decision, and the actions that ran.

  2. 02

    The flow lights up as it runs

    The same step is highlighted in the editor and marked Now, so you can see which block produced each answer, and change that block when an answer is wrong.

  3. 03

    Broken promises get caught

    If a reply says “I’ll connect you with a teammate” and no hand-off ran, the trace says so, before a customer ever sees it.

Start from a template

  • Troubleshoot, then hand off
  • Unhappy customer
  • Return request
  • Billing questions
  • Collect a lead and book a meeting

How playbooks work

Integrations

  • WordPress
  • Slack
  • WhatsApp
  • Telegram
  • Shopify

Connect the channels where customers already ask.

Embed Denser on your website, or deploy it to WordPress, Slack, WhatsApp, Telegram, and Shopify, so every cited answer reaches customers wherever they already are. No code required.

Human handoff

Human handoff to your team, Zendesk, Freshdesk, Intercom or HubSpot

The agent knows when to stop answering. A ticket opens in Denser's built-in helpdesk, or in the helpdesk you already use, the moment a conversation needs a person.

  • Asked for a human

    The visitor asks to speak with a person.

  • Stuck after multiple attempts

    The AI fails repeatedly and stops guessing.

  • Frustration detected

    The AI detects frustration and escalates early.

The transcript comes with it. Nobody re-types anything.

The helpdesk

A built-in helpdesk for every chat the AI agent hands off

Escalated conversations become tickets in one queue, whatever channel they started in. Assign an owner, track status, close it.

  • Chat logs become tickets

    Transcript and context arrive attached.

  • Live agent takeover

    Step into an active chat where the bot left off.

  • One queue, whole team

    Filter by All, Mine, or Unassigned.

  • Auto-replies for email and phone

    Customers hear back before a teammate picks it up.

Denser Helpdesk team queue showing tickets with subject, channel, assignee, and status, filterable by All, Mine, and Unassigned

See how Helpdesk works

What to automate

How the chatbot decides: answer, act or hand off

Automate approved answers. Hand off anything that needs account access, judgment, or an exception.

  • Answer

    Policy, product, and how-to questions

    Use approved content and show the source behind the answer.

  • Connect

    Order and account questions

    Use an authorized integration, or create a ticket without guessing.

  • Hand off

    Exceptions, complaints, and sensitive cases

    Send the transcript and context to an owner without making customers repeat themselves.

  • Improve

    Missing or contradictory information

    Stop, record the knowledge gap, and route the question for follow-up.

Compare approaches

AI support chatbot vs. rule-based chatbot vs. helpdesk add-on

The important difference is not whether a product can chat. It is how it grounds answers, handles uncertainty, and transfers work when automation should stop.

CriterionDenser support chatbotRule-based chatbotHelpdesk AI add-on
How answers are createdRetrieved from approved content with source citationsSelected from manually written branchesGenerated inside a separate helpdesk ecosystem
Unknown questionsRecords the knowledge gap and hands offFalls back when no branch matchesBehavior depends on helpdesk configuration
Taking actionCalls your API, books meetings, reads Stripe billing, posts to Slack, and runs Shopify orders, with confirmation before changesButtons and forms that hand the work to someone elseActions within the host helpdesk's own apps
Multi-step proceduresPlaybooks the agent follows in its own words, with a step-by-step trace of every test runDecision trees that break when customers go off-scriptVaries by vendor; often a separate workflow builder
Human handoffBuilt-in live takeover and ticket queue with transcriptUsually requires another tool or integrationNative to the host helpdesk
Best fitTeams that need cited answers plus an integrated helpdeskFixed forms and tightly controlled decision treesTeams committed to an existing helpdesk suite

Comparing vendors rather than approaches? See the best AI agents for customer support, with pricing, pros, and cons. Building the business case? Estimate your current cost per ticket and potential savings with our customer support cost calculator.

What Our Clients Say

A 72% resolution rate, and support tickets cut in half.

Denser Chat has been a game changer for us. It intelligently handles many of our customers' frequently asked questions, reducing customer support tickets by approximately 50%. This has freed up our marketing team to focus on more meaningful, high-value work.
Udara, Social & Digital Marketing Analyst, APL

Customer support chatbot FAQ

What is an AI customer support chatbot?

An AI customer support chatbot understands customer requests, retrieves answers from your approved help center and documentation, and decides whether to answer, take an action, or hand the conversation to a person. Denser cites the source behind every answer and opens a helpdesk ticket with the full context when the chatbot should not resolve the request on its own.

How is it different from a rule-based chatbot?

A rule-based chatbot follows a decision tree someone wrote by hand, so it only handles the paths that were anticipated. An AI support agent retrieves the relevant passage from your content and composes an answer, which means it handles phrasings nobody scripted and improves as you add documentation instead of as you add branches.

How do I set up a customer support chatbot?

Sign up for Denser and tell Ask AI your website, or upload your help center and docs yourself. Ask AI builds the agent, asks it the questions your customers ask, and helps you add it to your site, and every change waits for you to click Apply. No code is needed.

Which helpdesks does it work with?

Denser includes its own helpdesk on every plan: escalated conversations from website chat, the help page, email, the REST API, phone and WhatsApp become tickets your team can assign, track and close. If you already use Zendesk, Freshdesk, Intercom or HubSpot, the agent can file the ticket there instead.

Can the chatbot take actions?

Yes. Denser's actions let the agent call your own API, book a meeting on your Cal.com calendar, look up invoices and subscriptions in Stripe, post a request to a Slack channel, and search products, check orders, cancel orders, or request returns in Shopify. You describe when each action should run, test it in a real test chat, and switch it on. Actions that change something, such as cancelling an order, run only after the customer confirms.

How do I control what the agent does in multi-step cases?

Write a playbook: a step-by-step procedure for one situation, such as a return request or a billing question, built from Say, Do and If blocks. The agent follows the steps in its own words and runs the actions you chose. Billing lookups are read only, API keys are stored encrypted, and changes to an order wait for the customer's yes.

How do I test the agent before customers see it?

Every action and playbook has a test chat that works like the real widget. For playbooks, a trace under each reply names the step the agent is on, the path it took, and the actions that ran, and the same step lights up in the flow editor. You can test as a new visitor or as a known lead. Note that the test chat runs actions for real, so test bookings and Slack posts happen.

What is the best AI customer service chatbot?

It depends on where your support already runs. Denser is best for grounded answers with citations and flat monthly pricing. Intercom Fin suits teams already working in Intercom, Zendesk AI agents suit mature ticket operations, Ada and Sierra suit large enterprise programs, Gorgias suits Shopify stores, and Tidio Lyro suits small teams. Our comparison of the 10 leading tools covers pricing, pros, and cons for each.

How much does an AI customer support chatbot cost?

Denser has a free plan, and paid plans are flat monthly tiers from $39 (Starter) to $899 (Enterprise), with 20% off yearly. You are not charged per resolution. Actions, cards, and playbooks share one allowance per chatbot: 1 on Free, 5 on Starter, 10 on Standard, and 20 on Business and Enterprise. Collect leads and the Shopify actions don't count toward it. See the pricing page for current prices.

What happens if the chatbot cannot answer a question?

It opens a ticket in Denser Helpdesk with the full transcript attached, and tells the customer a person will follow up. Escalation happens when the visitor asks for a human, when the AI has failed repeatedly, or when it detects frustration - so a stuck conversation reaches your team before the customer gives up.

Can a live agent take over a conversation?

Yes. When a customer asks for a person, an agent can step into the live chat instantly, with full context from everything the AI already discussed.

Can the chatbot also answer phone calls?

Yes. Denser's voice agent answers calls on a phone number from the same knowledge base as your chat, and calls it can't resolve become tickets in the same queue.

Will customers know they're talking to an AI?

You decide how the agent introduces itself, and we recommend saying it's an AI assistant. Customers can ask for a person at any time, and the agent hands the conversation over with everything they've already said.

Will a customer support chatbot replace my support team?

No. The chatbot takes the repeated questions, such as policies, order status and how-tos, so your team spends its time on exceptions, complaints and sensitive cases. Those still reach a person, with the conversation and context attached.

What data sources can I train the chatbot on?

Websites, PDFs, Word and PowerPoint files, and plain text. Combine as many sources as you need into one knowledge base.

How does the chatbot avoid wrong answers?

Denser uses retrieval-augmented generation (RAG): it finds the relevant passage in your own content first, then answers from it and links the citation. If the answer is not in your sources, it says so and hands off to a person instead of guessing.

Can it handle a large knowledge base?

Yes. A single Denser AI support agent can search hundreds of thousands of pages and up to 50GB of documents, with three-layer retrieval that stays accurate at that scale. See the comparison page for the full breakdown.

What resolution rate can I expect?

Denser's customer support chatbot has a 72% resolution rate. Your own rate depends mostly on how much of what customers ask is covered by the content you add; Answer Quality shows the questions your sources missed, so you can close those gaps and raise it.

How long does setup take?

A first agent is live in about 5 minutes: paste your site URL or upload your docs, and copy one snippet onto your site. Most teams spend another 20 minutes testing answers before they switch it on for every visitor.

Answer what you can. Close the rest.

The helpdesk is there from the start.

No code. Free to start. Cancel anytime.