One AI Assistant Overview

One AI Assistant is a core part of the One Model platform that gives users a faster, more intuitive way to explore, understand, and use workforce data. It supports creating charts, uncovering insights, navigating Storyboards, performing research, and retrieving trusted answers through a single conversational interface. Each user's experience is shaped by configuration — organizations decide which capabilities to enable, what data the Assistant can access, and how much functionality to provide for each user group. All interactions respect role-based data permissions.

Video Overview

One AI Assistant capabilities

From guided answers to open-ended analysis, One AI Assistant offers six capabilities. Each supports a different way to access information, explore trends, or make decisions depending on the user's question, skill level, and goals.

Answers

Answers is the most structured and governed way to deliver information through One AI Assistant. Admins configure specific questions, choose an existing visual to return, and organize content by topic. Users can browse all available questions, see questions by topic, or ask their own question and be matched to the closest configured entry.

Each Answer includes four parts designed to help users quickly understand and trust what they're seeing:

• A short written answer generated from the question and the data shown in the linked chart or table (can be permissioned separately or disabled)
• A visualization (chart, table, or key value)
• A link to a relevant Storyboard for additional context and deeper exploration
• A list of related questions users can ask next

This setup suits users who don't have the time or context to verify accuracy on their own. It guarantees consistency — everyone sees the same answer, based on the same chart, with no need to fact-check or interpret.

See the One AI Assistant Answers user guide for full details.

Visualizations

Visualizations lets users create custom charts, tables, or key values by describing what they want to see in the prompt bar. It's a flexible way to explore data without building visuals manually or understanding the underlying model.

To support accuracy and consistency, the Assistant first checks for any configured Answers. If nothing matches, it interprets the prompt and builds a visual using the appropriate metrics, dimensions, filters, and time period. Users can refine the result by typing a short iterative prompt like add terminations or split by department, editing the original prompt, or using the Define panel to adjust the underlying query. Even when the first result isn't perfect, it provides a clear starting point for further refinement or deeper analysis. 

See the One AI Assistant Visualizations user guide for full details.

Analyze

Analyze enables users to ask natural language questions about any chart or table in the Assistant and get clear, contextual responses that simplify data interpretation. Analyze is available for both Visualizations and Answers and is powered by a secure large language model (LLM). The LLM receives the user's request along with the data displayed in the chart or table, so responses stay grounded in the data.

Analyze is built directly into the Assistant for in-flow analysis. Because conversations are threaded, users can ask follow-up questions that retain the full context of the discussion. Responses may include rich formatting such as lists, icons, and tables. Chart data is sent to the OpenAI LLM when using Analyze, so organizations should confirm alignment with their corporate AI policies before enabling this feature.

Optional web search capability is also available for Analyze. Web search adds up-to-date public information with sources cited as hyperlinks.

See the One AI Assistant Analyze user guide for full details.

Chat

Chat is a conversational AI feature built directly into One AI Assistant. It enables natural, back-and-forth dialogue, whether you're looking up HR concepts or brainstorming ideas. For example, you might ask about SHRM-recognized predictors of voluntary separation, get a concise answer, then follow up with clarifying questions — all while staying in the same thread.

Chat requires no configuration. You switch to Chat mode and start asking questions. Responses can include lists, tables, and rich text formatting.

As with Analyze, web search is available for Chat as a permissioned setting. Web search adds up-to-date public information with sources cited as hyperlinks. Enabling web search also makes the One Model Help Center available to Chat as a tool for answering questions about the One Model platform. Help Center access can be permissioned separately from full chat web access.

See the One AI Assistant Chat user guide for full details.

Insights

Insights help users focus on what's most important in a visualization without exporting data or analyzing it manually. Available as part of the Visualizations capability, insights are statistical observations that surface trends, comparisons, outliers, and other notable patterns automatically identified by One Model's Insights Engine.

Each insight includes a Gen AI-generated recommendation to suggest what to analyze or consider next. Recommendations are permission-controlled and can be turned off. One AI Assistant also shows its work, letting users see how the insight was calculated and verify it if needed.

See the One AI Assistant Insights user guide for full details.

Storyboard Navigation

Storyboard Navigation lets users ask the Assistant to find relevant Storyboards by topic. The Assistant searches Storyboard titles, descriptions, and page names to identify the best matches, then returns a list of direct links to the most relevant content. This helps users find existing content without relying on search or manual navigation.

See the One AI Assistant Storyboard Navigation user guide for full details.

How One AI Assistant works

One AI Assistant combines context identification, configuration, vector search, and generative AI to deliver fast, relevant results based on your organization's own data and content.

Configured Answers, metrics, dimensions, and Storyboards are transformed into embeddings — numeric representations of meaning. The system stores those embeddings in a secure vector database, allowing the Assistant to match user prompts to the right concepts even when different terms or language are used.

When a user submits a prompt, One AI Assistant routes it to the appropriate capability. Intelligent routing can be overridden by manually selecting a mode. If multiple modes could answer the prompt, the Assistant asks which one you would like to use. Each capability then uses generative AI to produce the result — using the embeddings to identify relevant matches, interpreting the question, generating an answer, or providing a recommendation.

Mode selection

One AI Assistant is designed to provide a seamless experience in AI Mode, where it automatically decides whether to return an Answer, a Visualization, an Analysis, a Chat response, or a Storyboard based on your prompt. This lets you ask questions naturally without worrying about which feature to use.

Because results depend on your organization's configuration and your own data permissions, AI Mode may not always route exactly where you expect. In those cases, Mode Selection lets you take control. Selecting a mode explicitly ensures the Assistant responds the way you want.

This balance of automation and control lets you work fluidly — rely on AI Mode for a unified experience, or switch modes when you want more precision.

Configuration and permissioning

Admins can tailor One AI Assistant to their organization using a range of configuration options. These include setting general behaviour, defining example prompts, configuring Answers, and determining which metrics, dimensions, and Storyboards the Assistant can access. The Assistant returns no results until this configuration is complete. All results respect each user's role-based data access permissions.

Each of the Assistant's core capabilities can be individually permissioned through One Model's Application Access Roles. Specific features powered by generative AI — such as Insight recommendations and the text-based portion of Answers — can be permissioned separately.

See the One AI Assistant Configuration Guide for setup details.

Use of generative AI

Your data is never used to train OpenAI or One Model models. What is sent to OpenAI depends on the capability being used:

• Prompts are sent when users submit a request
• Embeddings are sent as part of configured content
• When users generate a written interpretation of a chart or table (for example, in Analyze or the text portion of Answers), the prompt and the data shown in that chart or table are sent to OpenAI

These inputs may be stored for up to 30 days to monitor for abuse, in line with OpenAI's API policy.

Generative AI supports several parts of the Assistant's workflow. Each of these features can be permissioned separately by role, so only authorized users can access them:

• Context identification
• Embedding creation
• Recognition of metrics, dimensions, and time selections in prompts
• Generation of written Answers
• Performing analyses
• Ranking of Insights
• Generation of Insight recommendations

See the One AI Assistant Generative AI Usage article for full details on what is sent to OpenAI and when.