HaloCRM

HaloCRM Guides

Guides > ChatGPT/Open AI Integration

ChatGPT/Open AI Integration

In this guide we will cover:

- What is the ChatGPT Integration?

- Connecting to ChatGPT

- Common Uses for OpenAI



What is the OpenAI Integration?

The native integration with ChatGPT can be used to write responses on tickets or improve on response agent's have written themselves. It can also be integrated into your live chat to respond to a customer, taking the customer's given context into account.


There is various other functionality that can be achieved once chat GPT/Open AI is connected to your Halo instance. For information on the functionality available checkout Leveraging AI in Halo. 


Note that free versions of OpenAI are limited in the connection they provide, although can be used for testing. If you are looking to use the OpenAI connection long term a paid account will be required. 


A list of available models and their pricing per million tokens can be found here.


Connecting to OpenAI

First, make sure that the OpenAI integration is enabled. To do this, head to Configuration > Workspace > Integrations > OpenAI and enable the integration using the '+' icon.

Fig 1. Enable Integration module.


Once enabled, click into it.


You will be asked to provide your OpenAI API Key and the model you are using. 

Fig 2. Connection details.


Obtain API Key

You can obtain this by going to: https://platform.openai.com/


Log  into your account (or create one!).


Once logged in, click on your icon in the top right-hand corner, and go to 'view API keys'.


Fig 3. View API keys in OpenAI.


Here, you can generate a new secret key. Once generated, make sure you copy the key into your clipboard as this will be your only opportunity to do so.


Fig 4. New API Key.


Paste that key into the API key field on the OpenAI integration page in your Halo instance. 


Obtain OpenAI model

The model used will depend on which model you would like to use to complete requests. You will need to have purchased tokens for this model within the connected account. 


Add details into Halo

Now you have your connection details you can plug these info Halo from the integration setup page (shown in figure 2). 


You can also plug them in by heading to Configuration > AI > select 'Own OpenAI Connection' as the 'Default AI Connection' > Configure OpenAI connection. 

Fig 5. Enter OpenAI connection details.


Note: If you have added the connection details via the integration module the details entered will also be visible from when using 'Configure OpenAI Connection' under Configuration > AI. 


Common Uses for OpenAI

Once you have OpenAI connected to your instance you can achieve various functionality in Halo. Some common uses for this include improving agent responses using AI and plugging AI into the chatbot, allowing AI to respond to users rather than agents. 


Improve Agent Responses

Create a new, or add to an existing action, under Configuration > Forms (Tickets) > Actions > New.


At the bottom of the 'Details' tab, you will see the fields shown in figure 6.

Fig 6. AI options for new action.


Set the 'Operation' to 'Improve Agent Note'. 


Set which AI model should be used for this. 


In the prompt box, you can train OpenAI to respond in the way you'd like it too. So in this example, I want OpenAI to improve a pre-written response to make sure it is both friendly and polite.


Fig 7. Action configuration to improve agent note.


Note: This will only improve input to the 'Note' field, not any free text field so ensure you are using the 'Note' field on the action. 

 

Now when an agent uses this action:


Before: 

Fig 8. Agent note before AI improvement.



After:

Fig 9. Agent Note after AI improvement. 



Chatbot

Head to your chat profile that you use and go to the 'bot flow' tab. Press 'edit'.


Where you deem necessary to add in OpenAI's response, you can add this to the bot flow. Typically, it is used in replacement for agent chat - but it is important to add that you must feed the relevant information to OpenAI in order for it to work most seamlessly. I.e. Point it to your website or other.


Edit the step you'd like to add ChatGPT and select 'generate response with ChatGPT' as your action type.


Fig 10. Chat step in which an AI gives a response. 


Underneath, you can train OpenAI with its tonal behaviour and where to pull knowledge from. 


The failure message (if OpenAI is slow/down etc) can be used as a fallback. 


For example; 'hang on while we connect you to an online agent' and then have a step in the bot flow following which does actually connect the customer to an agent. 


You need to ensure that the customer has already input their issue/question/enquiry in order for OpenAI to work efficiently on the live chat.

Fig 11. Chat flow example. 


The above example would be a good way to build your bot flow.


Finally, you can build custom integrations using OpenAI. To see how this is done, please refer to the following guide.


Popular Guides

Pin It on Pinterest