Enhancing ServiceNow's AI Capability: RAG (Industry knowledge Base) + NowLLM

[Vivek Verma]

Hello,

I'd like to suggest something regarding the AI Capability of ServiceNow. It is widely acknowledged that LLMs represent both the future and, to an extent, the present. However, the current NowLLM is not equipped to meet industry-specific needs, such as those in telecommunications and healthcare. This limitation is understandable, given that the ServiceNow LLM model was trained on ServiceNow architecture, data, and publicly available data.

Therefore, I would like to suggest that ServiceNow should consider offering the NowLLM model with added RAG (Retrieval-Augmented Generation) capabilities. This would allow any company to purchase NowLLM as a base product and fine-tune their version of NowLLM with their company's knowledge base. I believe this would be a significant step forward for ServiceNow and would allow companies to have a more personalized and industry-specific version of NowLLM. 

Please inform me if ServiceNow is already implementing this.

Thanks,

Vivek Verma

LinkedIn || Github

[Mark Manders]

Sounds like a good idea, but only ServiceNow can tell you if they are already working on it, or thinking about it. Have you entered it as an idea? https://support.servicenow.com/ideas

That's where others can vote on it, making it a higher priority for ServiceNow.

[Laurent5]

I wrote an article on how it can currently be done with 3rd party LLMs which you may find interesting:

https://laurent-roussel.medium.com/leveraging-external-llms-and-rag-with-servicenow-virtual-agent-la...

[Vivek Verma]

Excellent Article 😍,

It would be great if ServiceNow provided a NowLLM capability similar to this.

[Shamus Mulhall]

Now Assist in AI Search does use a RAG architecture to answer questions using the existing vetted knowledge content from your knowledge base.

https://docs.servicenow.com/bundle/vancouver-intelligent-experiences/page/administer/now-assist-admi...

https://www.youtube.com/watch?v=sgVngXSzF3E&list=PLkGSnjw5y2U5_eA3Ta2NOwBcuqMtmP6e4&index=19&t=130s