It is easier to judge a product by watching it work than by reading a feature list. So here is how ConvoRail works, told as the story of two visitors on a website: one who asks a question the bot knows, and one who asks something it does not. The example business is fictional and the conversations are illustrative, not customer results.
The Setting
Imagine a small business with a website, a handful of Q&A pairs about delivery, returns and opening hours, and the ConvoRail widget installed with one script tag. Its owner has put three common questions on the widget's first screen as shortcuts and set a friendly opening message for the pricing page.
Visitor One: The Question the Bot Knows
When the question is in your knowledge, the visitor gets a clear answer in seconds and nobody on your team is interrupted.
Step 1: The Visitor Opens the Chat
A visitor on the pricing page sees the chat bubble and clicks it. The first screen shows the greeting for that page and three tappable questions. This matters because most visitors do not know what to type. A shortcut removes the blank-box problem.
Step 2: They Ask in Their Own Words
The visitor types: "how long does delivery usually take?" They did not use the exact wording of the Q&A pair, and they do not need to.
Step 3: The Bot Checks Its Knowledge
ConvoRail looks for the closest match in the knowledge the business has given it, including its Q&A pairs and its website content. Only if there is a close enough match does the bot reply, and the reply is built from that knowledge.
Step 4: The Visitor Gets a Clear Answer
The answer comes back in seconds. It can include bold text, a short list or a link, so the visitor gets a readable answer rather than a wall of text. The visitor rates the conversation when it ends, and the business can see the rating later. No one on the team was interrupted.
Visitor Two: The Question the Bot Does Not Know
When the question is not in your knowledge, the bot says so, offers a short contact form and sends your team an urgent flag instead of guessing.
Step 1: A Different Kind of Question
Another visitor asks: "Can you deliver this to a small island, and can I pay half now and half later?" The business has never written an answer to that, and may not even have decided its policy.
Step 2: The Bot Refuses to Guess
Nothing in the knowledge matches closely enough, so the bot says plainly that it does not have confirmed information. It does not invent a delivery policy or a payment plan. A guess here could become a promise the business has to keep, which is the problem described in why bots make up answers.
Step 3: The Visitor Leaves Their Details
Instead of leaving the visitor at a dead end, the bot offers a short form asking for their name, email and phone number. The visitor fills it in. The details travel with the conversation.
Step 4: The Team Gets an Urgent Flag
The conversation arrives in the shared inbox, flagged as urgent, with the whole thread and the visitor's details. A teammate opens it, assigns it to themselves and replies. The visitor, still on the site, sees the reply appear in the chat with the teammate's name. If the visitor has gone, the contact details mean the team can follow up by email or phone.
Step 5: The Bot Stays Out of the Way
While the teammate is in the conversation, the bot does not interject. The visitor hears one voice.
What Happens Afterwards
The business owner opens the dashboard at the end of the week and sees that two visitors asked about island delivery and half payments. They decide on a policy, add two Q&A pairs, and the next visitor who asks gets an instant, consistent answer. That is the loop that improves the bot over time, and it is covered in using unanswered questions to improve your chatbot.
The Whole Flow at a Glance
The flow has two paths: known questions are answered automatically, and unknown ones become urgent conversations for your team.
| Moment | What ConvoRail does | What your team does |
|---|---|---|
| Visitor opens chat | Shows greeting and first-screen shortcuts | Nothing |
| Question is in the knowledge | Answers in seconds from your own knowledge | Nothing |
| Question is not in the knowledge | Says it does not have confirmed information and offers a contact form | Gets an urgent flag in the inbox |
| Visitor leaves details | Attaches them to the conversation | Replies live or follows up later |
| End of week | Lists unanswered questions on the dashboard | Adds the missing answers |
Why This Design
Many chatbots are built to always answer. ConvoRail is built to answer only when it can stand behind the answer, and to turn the rest into a conversation your team can finish. The design accepts that sometimes the right response is "I don't know, but here is how we will find out." For a business, that is the safer default, and it keeps trust intact.
What the Visitor Sees and What Your Team Sees
The visitor sees a short, readable conversation, while your team sees the same conversation with ownership, priority and notes attached. The two views are deliberately different.
- The visitor sees: a greeting, tappable questions, a clear answer or an honest "I don't have confirmed information", a short form if needed, and a live reply with a teammate's name.
- Your team sees: the full thread, the visitor's details, an urgent flag where the bot could not answer, and fields for status, priority, assignee, tags and internal notes that the visitor never sees.
Try the Flow Yourself in Ten Minutes
You can walk through this exact flow yourself in ten minutes on the free plan: add a few answers, open your site in a private window, ask one known and one unknown question, then reply from the inbox. Seeing both paths firsthand is more convincing than any description.
- 1Add three Q&A pairs, such as hours, delivery and returns.
- 2Open your site in a private window and ask one of them in your own words.
- 3Ask something you did not cover and fill in the contact form.
- 4Open the inbox, reply, and watch the reply appear in the chat.
Conclusion
A visitor either gets a quick answer built from your own knowledge, or an honest "I don't know" followed by a contact form and a live reply from your team. Either way the conversation ends with something useful, and the questions the bot missed become the answers you write next.
Key takeaways
- The first screen offers greetings and one-tap shortcuts, so visitors do not face a blank box.
- The bot replies only when your knowledge has a close match.
- When it does not know, it says so and offers a contact form instead of guessing.
- The conversation reaches the inbox flagged as urgent, and your team replies live.
- Missed questions become the Q&A pairs you add next.
Quick Summary
How ConvoRail works: a visitor opens the chat, taps a shortcut or types a question, and the bot answers only if your knowledge has a close match. If not, it says it does not have confirmed information, offers a contact form and sends the conversation to your shared inbox flagged as urgent, where a teammate replies live. Questions the bot missed become the Q&A pairs you add next, so coverage improves over time.
Want to see grounded answers on your own site?
Start free, add your Q&A pairs and paste one script tag. No credit card required.

Founder & CEO
Writes about grounded AI chatbots for websites: how they work, how to measure them and what to check before you buy.
Topics
Get new posts by email
Guides on AI chatbots for websites, straight to your inbox.
