Generative AI refers to an artificial intelligence system that can produce various types of content, such as text, images, audio, and other data formats, based on the inputs provided to it.
ServiceNow offers a generative AI controller that allows this functionality to be integrated into existing workflows on the platform.
How does the ServiceNow Generative AI Controller work?
The ServiceNow Generative AI Controller is a low-code integration that connects generative AI application programming interfaces (APIs) to the ServiceNow platform. The controller leverages the ServiceNow platform's Integration Hub functionality, which allows third-party APIs to be executed as part of a flow when a specific event occurs in ServiceNow. Customers can connect to generative AI providers like Azure OpenAI and OpenAI using Integration Hub spokes without having to write scripts.
After integrating capabilities into desired workflows, you can begin leveraging generative AI. In a workflow (process flow or virtual agent topic) or custom script, ServiceNow sends a request to the generative AI provider with a message (i.e., "What are the best places to travel"?). When the provider receives the request, it processes it using its models and sends a response to ServiceNow. ServiceNow can use the response to create new content, answer questions, summarize text, and more.
What problems are solved with generative AI?
Creation of knowledge articles.
Accurate answers to user questions.
Summary of complex information.
Increased accuracy and scalability of personalized content.
Improve user experience and satisfaction by reducing errors or misunderstandings between agents.
Data requirements and considerations
Machine learning requires data to create, train, and continually improve models used for automation and prediction. The accuracy of the models depends on good quality data. But what type of data do you need and what does good quality data look like?
What should your data look like?
To help generative AI vendors generate more accurate content, the data analyzed by the generative AI vendor must be diverse, high quality, and relevant to the task. This may include:
The data must be of high quality and free of errors and inconsistencies. This can help ensure that generative AI models return more reliable generated content.
The data must be relevant to the requested task. For example, if the goal is to answer user questions with knowledge articles, the knowledge base should include a variety of articles that cover common user questions, concerns, or problems.
Generative AI requests should be clear and concise.
Supported Customizations
ServiceNow provides a list of supported customizations that do not impact the integrity of your AI solution.
Supports different ChatGPT models.
Works with Flow Designer, Virtual Agent and in custom scripts.
Integration support
ServiceNow can easily integrate with other business applications using out-of-the-box integrations from the ServiceNow Store and Integration Hub, or with advanced scripts, protocols, and features via APIs. These capabilities extend to most of the platform's capabilities.
Does the Generative AI Controller support integrations?
Yes, the ServiceNow Generative AI Controller integrates with OpenAI and Azure OpenAI using spokes. The Generative AI Controller comes with the following APIs out of the box:
Completions
The Completions API uses legacy ChatGPT models to generate content, such as text-davinci-003.
Chat Completions
The Chat Completions API uses newer ChatGPT models to generate content, such as gpt-4 and gpt-3.5 turbo.
Moderations
The Moderations API checks whether the content complies with the provider's usage policies. Therefore, developers can identify prohibited content and remediate it.
Which ServiceNow products can use the Generative AI Controller?
The Generative AI Controller is a platform capability that all applications can take advantage of. Future versions of the ServiceNow Store may have extended application-specific generative AI capabilities.
