Quickstart
In this quickstart guide, you will create a simple support bot that interacts with a customer and attempts to solve their issue, based on the knowledge you have given it.
In the data tab, connect one or more data source. Currently supported sources are:
- Websites
- Articles from a Salesforce Knowledge base
- Tickets from a Salesforce Case database
- Articles from a Zendesk Help Center
- Tickets from a Zendesk Tickets database
- Pages from a Notion workspace
- Files from a GitHub repository
- Discussions from GitHub Discussions
Head over to the Markprompt playground and give the assistant custom instructions, such as branding and tone, or how to behave in specific situations.
Try to ask it some questions!
Now that your assistant is ready, you can deploy it as:
- A chatbot on your website
- A case creation form on your website
- A Slack bot
- A Salesforce draft assistant
- A Zendesk draft assistant
- A Zendesk chat bot
- A Raycast extension
Follow the links for step-by-step guides on how to do it – it's easy!
Once the assistant is live, data will start pouring in. In the threads tab, you can inspect all conversations coming in in real time, and the data that was used to generate a response. You can enable automatic triaging, sentiment analysis, urgency and CSAT scores.
The insights dashboard gives you precise understanding of the performance of your assistant, such as the rate of answered or escalated questions, or the most cited sources.
Now the work begins to improve your assistant! With these insights tools at your disposal, you will be able to spot shortcomings of the system, such as a missing or out of date article, or common issues your customers are dealing with. You will have strong data points that allow you to take action and continually improve the performance of the assistant, so that you can serve your customers even better.