AI tools have been evolving by leaps and bounds over the past several years. LLMs revolutionized content creation and finding information. 

Public LLMs such as ChatGPT have indexed a lot of information, but sometimes it is out of date, or does not include specific, detailed information that you or your organization may require it to know. Additionally, these public LLMs do not have, nor should they have, access to your private internal data.

It is now possible to privately host and customize LLM models that can be prompted and generate results using data or information that you choose and is important to your organization, and connect it to your website. 

There are open source models available that can be hosted, and then enhanced via Retrieval Augmented Generation (RAG).

Any number of these RAGs, or vector datasets, can be created each containing different data sets and configured in a variety of ways, for use in different AI based functionality.

In this way when your website's AI integration prompts an LLM, it can have access to information that is not in the public model, while keeping your data private and secure. This can include any combination of information from external websites, internal documents, site content,  and CRM data.

Hosting your own LLM model isn't necessarily required, but there are API costs for using public LLMs, and potential privacy issues with sharing your prompts, and the augmented information used to customize them. Website AI functionality can mix and match both private and public models with custom vector databases as well.

In this way it can be possible to place AI functionality in your site that references and utilizes information relevant for your organization, such as:

• AI Chat bots and assistants
• AI Search
• AI Generated content
• AI translations
• AI moderation
• AI Image generation

The AI space is rapidly evolving, with more and more open source software being created each month. For example a large suite of AI modules available for Drupal now. This suite of modules contains the functionality to generate RAG vector datasets of all Drupal content.

By installing and configuring CiviCRM Entity, CiviCRM data can be indexed into multiple RAG vector datasets and made accessible to privately hosted LLMs via the RAG methodology. In this way you can take advantage of all the Drupal AI tools with your CRM data in a private, secure infrastructure.

For example, we can create vector datasets of your CiviCRM Events so your users can search for and ask questions about them in an AI chatbot. These vector datasets can combine other site content as well, including code of conduct or other documentation, putting all the necessary information in the hands of your user via prompt.

Your administrators could use an AI based search to find contacts. 
 

Image

At Skvare, we continue to monitor and learn about ongoing technological trends to best service our clients and stay competitive in this dynamic marketplace. Contact us to explore private, custom, AI options and infrastructure for your CMS/CRM.