System search results typically respond to user input with a file or links and minimal context to the underlying resource. In contrast, Workgrid’s knowledge apps understand the intent of the user’s query and answers it directly in natural language, in addition to providing a link to the sources of the response, and in some cases suggested prompts to take direct action based on the inferred intent.

Workgrid combines our proprietary query analyzer with the search API from your systems and knowledge bases to surface responses to employee questions. If an answer comes from more than one system, the Assistant uses ranking algorithms and RAG to evaluate the potential answers for the most relevant response. In some cases, this may be a summarized, text-based response. In other cases, the Assistant may surface an app to connect with system-based information such as IT Help Desk. 

The AI Assistant recognizes an employee by user authentication, which ensures that the responses include only information the user is permitted access to. Workgrid relies on existing configurations in your enterprise systems ensuring responses are refined to what each individual employee is authorized to see and reduces the need to add additional management responsibility. 

Workgrid uses a multi-model architecture leveraging a range of classification, re-ranking, and large language models to aggregate and assess information.

In addition to RAG technology used to isolate answers to the content connected to the knowledge app, we also build guardrails into all our prompts to prevent them from using outside information or hallucinating. Workgrid runs a series of prompt injection tests to ensure that users cannot override the prompts in our apps or ask for information outside of the source system. The knowledge app data sets are run against a sophisticated Model Testing & Evaluation framework designed to test reliability, accuracy, and performance against industry benchmarks for continuous improvement. While we do all we can to ensure the accuracy and reliability of our responses – as all generative AI platforms do – we recommend appending knowledge responses with encouragement to confirm important information using the source link provided in the answer. 

No, Workgrid does not store data or train our model based on your organization’s data and will not build new content to respond to queries. The responses are searched and summarized from the existing systems your business already leverages. However, our generative AI app templates empower users to connect their own LLM to Workgrid. Paired with a combination of data sources that exist today, users can build an experience to generate data within their organization’s model.