Artificial Intelligence (AI) is revolutionising how businesses operate, and presents ServiceNow customers with a brilliant potential opportunity, given the value proven. One key decision that ServiceNow customers face is whether to rely on ServiceNow’s dedicated NowLLM or utilise a general-purpose large language model (LLM) like OpenAI’s GPT-4 or Google’s Gemini. This decision will impact user experience, accuracy, integration, security, and the cost to deploy and maintain.
In this article, we’ll explore the key differences between these two approaches, highlight use cases where each excels, and provide guidance on how to choose the best option for your ServiceNow strategy.
Understanding NowLLM and general-purpose LLMs
What is NowLLM?
NowLLM is a specialised AI model designed to enhance ServiceNow’s platform. Unlike general-purpose LLMs, NowLLM focuses on ServiceNow’s workflows including IT Service Management (ITSM), HR services, and customer support. It’s trained on data relevant to these contexts, making it a highly focused subject matter expert. This specialisation allows NowLLM to provide highly accurate and context-aware responses tailored to ServiceNow’s ecosystem.
What is a general-purpose LLM?
General-purpose LLMs, like OpenAI’s GPT-4 or Google’s Gemini, are trained on massive datasets spanning multiple industries and topics. They’re designed to handle a wide variety of requests but are not specialised in any one domain. While these models offer versatility and creativity, they may struggle to provide precise, context-specific responses for tasks that require deep knowledge of a specific platform like ServiceNow.
Comparing NowLLM and general-purpose LLMs
| General-Purpose LLM | NowLLM | Advantage |
Accuracy and Contextual Relevance | General-purpose models are trained on a broad range of topics, which can sometimes dilute their focus when dealing with specialised contexts like ServiceNow’s workflows. This broader training can result in irrelevant responses or a lack of precision. | NowLLM’s training is tailored to ServiceNow’s data structures and workflows. For instance, when summarising a customer service incident, NowLLM can identify key data points and prioritise essential details, providing a clear and concise summary. General-purpose LLMs might not have the same insight into ServiceNow’s specific data structures, leading to less accurate or incomplete summaries. | NowLLM |
Integration and Customisation | While it is possible to integrate general-purpose LLMs into ServiceNow, in practice this often requires custom development, API connections, and ongoing maintenance. This approach is more complex, time-consuming, and costly compared to the pre-integrated approach of NowLLM. | NowLLM’s tight integration with ServiceNow’s workflows, ITSM, and HR services means it works out-of-the-box with minimal configuration. Additionally, customers can use ServiceNow’s Now Assist Skill Kit to build custom AI-driven workflows. | NowLLM |
Security and Data Privacy | Public LLMs require data to be sent to third-party servers, raising concerns about data privacy and confidentiality. Organisations using public general-purpose LLMs must evaluate whether sensitive data is at risk of exposure or worth the significantly increased cost of running a private general-purpose LLM. | ServiceNow prioritises security and transparency. Customers have control over whether their data is used for model training, and in Stanford University's Foundation Model Transparency Index ranks first in transparency, scoring 100% across most transparency dimensions and outperforming major providers like Google, OpenAI, and Meta. | NowLLM |
Flexibility and Innovation | General-purpose models are more versatile and can handle multiple use cases beyond the ServiceNow platform, such as content creation, customer sentiment analysis, or support for non-ServiceNow workflows. | Since NowLLM is optimised for ServiceNow’s ecosystem, it’s less effective for use cases outside this context. For instance, it’s not suitable for tasks like analysing social media sentiment. However, the Now Assist Skill Kit does empower rapid customer innovation within these constraints. | General-purpose LLM |
Cost and Maintenance | Integrating general-purpose LLMs with ServiceNow will always require a significant level of custom development, even though ServiceNow simplifies this with the ability to call an external LLM via its generative controller. This process can be more complex, time-consuming, less reliable and costly. | Updates and maintenance of NowLLM are included in the ServiceNow software license agreement. ServiceNow continuously improves NowLLM’s capabilities via updates from platform upgrades and frequent releases to the ServiceNow Store. | NowLLM |
Example uses: When to choose NowLLM vs. general-purpose LLMs
- ITSM Incident Summarisation. Winner: NowLLM. NowLLM’s familiarity with ServiceNow’s data structures allows it to prioritise key information in an incident report, ensuring concise and clear summaries. A general-purpose LLM might struggle to identify which details are most relevant.
- Multi-Domain Knowledge Base Creation. Winner: General-Purpose LLM. If a business needs to create a knowledge base that spans multiple platforms, like internal CRMs or ERP systems, a general-purpose LLM’s broader scope could be advantageous.
- Onboarding. Winner: NowLLM. For HR workflows managed through ServiceNow, NowLLM’s understanding of the HR module allows it to automate onboarding processes, create FAQs, and manage employee support requests with precision.
How to choose the right AI strategy for your ServiceNow implementation
- Define the scope of your use case: If your use case is tightly bound to ServiceNow’s ITSM, HR, or CSM workflows, NowLLM is the logical choice. If the goal extends beyond ServiceNow’s ecosystem, consider a general-purpose LLM or a hybrid approach.
- Take a value-led approach: Identify the benefit AI provides for each use case and ensure measurable success metrics are in place.
- Evaluate integration needs: If seamless integration with ServiceNow’s workflows is critical, NowLLM’s built-in compatibility offers significant advantages.
- Assess cost and maintenance: If budget constraints exist, leveraging NowLLM’s built-in support and updates under the ServiceNow license agreement may be the most cost-effective option.
- Consider a hybrid approach: Sometimes, the best approach is a hybrid one. Use NowLLM for ServiceNow workflows and general-purpose LLMs for broader, cross-platform applications.
- Run a Proof of Concept (POC): Before committing, run a POC to see how each model performs in real-world tasks.
Conclusion
Selecting the right AI model for your ServiceNow strategy will depend on your specific use cases, budget, and business needs.
To formulate your ServiceNow AI strategy, an AI Readiness Assessment is an effective way to start. This should baseline your current AI maturity and provide a target maturity level with recommendations and suggested implementation priorities tailored for your organisation.
2 Comments
