Skip to main content

Agents

AI agents normally involve a fair amount of programming and tweaking. Because we have already done the hard part of embedding the Agent directly into the system's core network, the clever stuff is already set up and ready to roll out as soon as you set some basic information for the Agent.

Agent initial setup

When you first enter the Agent screen, you will be required to set up basic customer information.

 

Click on 'Run the initial setup' to complete this process.

 

image.png

 

image.png


Customer Defaults
These settings will become the default for all agents built on this customer's account. Agents can override these defaults by selecting their own settings.

image.png

image.png

This is the name of the customer's account and is only used for reference.
Company Name is the default name the agent will use when greeting a caller. It is worth remembering to write the name as you want it said. So in this case, Simply Cool Ltd should be written as Simply Cool Limited to stop the agent spelling out L T D as individual letters. This, of course, is easily changeable, so you can adjust it as you go along to get the best results.

image.png

image.png

The Default voice will be used by all agents if no voice is selected on a per-agent basis. There are many to choose from.

 

When you have selected a voice, a play image.png icon will appear, allowing you to listen to the voice before selecting it. Again, nothing is set in stone, so you are welcome to come back and make a change at any time.

image.png

image.png

Timezone works with your default business hours and helps agents understand the time. The default for the system is Europe/London GMT, but you can select other time zones if need be.
This enables or disables all agents on the customer's system and stops agents from taking calls. The default is 'Active'

image.png

image.png

The call recording system will capture all calls made and received by the agents. It is on by default.

The company description is the set of details agents will use as background to help them understand what they should say when answering a call. This is particularly useful if the agent is working as a Receptionist.

 

The Expand button just makes the screen bigger so it is easier to read all the information provided in the company description

image.png

image.png

System prompt override. This will replace all the clever, predetermined network control over how the agent performs. These settings can be removed if any issues occur; however, we recommend trying the default system prompt first before overriding for all agents. This setting is recommended for expert setup only.

 

The Expand button just makes the screen bigger so it is easier to read all the information provided in the system prompt override.

Additional instructions. These are appended to the default prompt. This is a good way to provide a bit more clarity on how you want the agent to respond or act.

 

 

The Expand button just makes the screen bigger so it is easier to read all the information provided in the additional instructions

image.png

image.png

Default business hours
The default business hours are the times you want agents to respond to calls during working hours. Outside these working hours, the agent may just take messages or simply say that you are closed.

image.png

image.png

Toggling the switch to off will close your office for the day.
You can also add special days, such as bank holidays.

image.png

Keep in mind that all these defaults can be overridden by any and all agents. This is to set up the most frequently used settings to make each agent's setup easier.

Final steps.
Finally, click Enable and Save Defaults to complete the initial agent setup.

image.png

image.png

You can now return to the initial setup screen or return to the AI Agents list. 
Customer-wide knowledge sources
You can also add some company-wide knowledge sources. The agents will use this to build their LLM (Large language model), which is basically an understanding of their role. Agents can 'absorb' a lot of information to use whilst a call is in progress. The agent doesn't really absorb and hold the information. The LLM provides the agent with instant recall of information when a question is asked. it is all about the information at hand for the agent to recall and project the information provided.

image.png

image.png

Wow, that was a lot of words. The AI agent that will read this in the next 12 hours should blush when it knows I am talking about them/it.

image.png

To start a new knowledge source, click on the new button.
You will now see the box with a few options.

image.png

image.png

The source type

 

1. This can be simple text entered in the prompt box below.

2. Scrape a website will download information from a chosen website.

3. Upload a file like a PDF.

 

Text:

This will be added to the LLM to help the agent locate the offices.

image.png

image.png

Once saved, you will see confirmation showing the text document name and the last time it was ingested into the LLM

Scrape a website:

 

You can collect information from a website to build the knowledge base. Enter the website address, then select whether you would like the URL to be revisited every 12 hours to check for any changes. This is great, for example, if you have user guides online that are constantly being updated. You can choose how deepdeeply you want it to collect information on each page, andas alsowell as the maximum number of pages. The defaults are probably what we would recommend leaving it set to.

 

 

 

image.png

 

image.png

The system will prepare itself to start collecting the information

It will keep collecting page information until it either reaches its maximum request limit or has read every page of the requested website.

image.png

image.png

Once completed, it will provide how many pages were collected, when it was last ingested and when it will run again.

Upload a file:

 

PDF's, Word documents and text files can all be uploaded

image.png

image.png

Clicking the Choose File button opens your PC's file system to find the document you want to upload.

Once selected, press the saveSave knowlodgeKnowledge button.

 

image.png

 

image.png

Click on continue editing the customer defaults to see the knowledge base data again


You will then see your document and the date and time it was ingested into the LLM.

image.png