Introduction
What ikazebot is, the five ideas everything else builds on, and how a reply is made.
On this page
ikazebot lets your business build AI agents that answer customers from your own content. They answer on your website, on a hosted help page, and on WhatsApp, Messenger and Instagram. When a person is needed, your team takes over from one shared inbox.
This guide follows the order you would set things up in. Each page explains what a feature does, then how to use it.
The five ideas everything builds on
| Idea | What it is |
|---|---|
| Workspace | Your company's space in ikazebot. Agents, knowledge, conversations, contacts and your team all belong to one workspace. One person can belong to several. |
| Agent | An AI assistant with a name, instructions and a model. You can run several, for example one for sales and one for support. |
| Knowledge | What the agent learns from: files, websites, text and question-and-answer pairs. It answers from these, not from guesses. |
| Channel | Where customers talk to the agent: the website widget, the help page, WhatsApp, Messenger, Instagram or your own app through the API. |
| Credits | What replies cost. Each plan gives a monthly allowance, and a more capable model uses more credits per reply. |
How a reply is made
When a customer writes, ikazebot finds the parts of your knowledge that match the question. It then asks the agent's model to answer from those parts only.
- 1A customer writeswidget, help page, WhatsApp…
- 2Guardrails check itspam and, if you choose, topics
- 3Your knowledge is searchedthe closest passages
- 4The model answerswith your instructions
- 5The reply is savedwith its sources and its cost
If nothing in your knowledge matches well enough, the agent says it does not know instead of inventing an answer. That is why good knowledge matters more than any other setting.
Your path through this guide
- 1Create a workspace
- 2Create an agent
- 3Add knowledge
- 4Test in the playground
- 5Deploy to a channel
- Create your account and workspace.
- Create your first agent and give it instructions.
- Add knowledge so it knows your business.
- Test it in the playground before customers see it.
- Put it on your website or your messaging apps.
- Run it day to day from the Inbox and Analytics.
When you are ready for more, add actions so the agent can do things, and procedures so it follows your playbook step by step.
Related
- Your account and workspaceSign up, verify your email, create a workspace and switch between workspaces.
- Create your first agentName it, pick a model, write its instructions and set its temperature.
- Frequently asked questionsShort answers about agents, knowledge, channels, credits, your data and your team.
Something not working as described here? See Troubleshooting.