One AI Assistant Answers Best Practices Guide

Answers shapes how users experience their data. The questions you write, the visuals you choose, and the order you place them in turn a static Storyboard into a clear and cohesive story. Instead of navigating multiple pages or guessing where to look, users start with a curated question that takes them directly to a validated chart, an optional natural language explanation, a link to the full Storyboard page, and related questions to continue exploring. This guide covers how to design strong Answers, select the right visuals, write effective questions, organise content into topics, and maintain Answers over time.

What Answers is

Answers turns your Storyboard tiles into guided questions for users. You write the question, choose the visual it displays, and organise it inside topics you define. When users click a question, or ask something similar in natural language, they get a written explanation if permissioned, the correct chart, a link to the Storyboard for more detail, and related questions to explore next.

Because every Answer uses a visual you have already reviewed and approved, users get an accurate and reliable explanation based on content you trust.

Why Answers is valuable

Well-built Answers highlight what matters and give users a structured, reliable way to move through information. They are especially valuable for anyone new to analysis or unsure where to begin.

Answers gives users a trustworthy way to understand their data without needing to locate the right chart or interpret it alone. Each Answer uses a visual you have already validated, so users know the data, definitions, and logic are correct. The experience is structured and approachable, and it supports natural storytelling because you control the sequence and flow of the questions.

Key benefits include:

• Accurate information that comes from visuals you have already reviewed and approved
• An entry point for casual or new users who prefer guided questions instead of navigating large Storyboards
• A natural learning path, because you control the order and flow of questions
• Storytelling through curated questions that present information in a deliberate sequence
• Context through links back to the Storyboard, with metric definitions and drillthrough available when permitted
• Faster access to common questions without searching through multi-page Storyboards
• A transition into Analyze for follow-up questions using the same chart
• Alignment with Role Based Security, so users only see what they have permission to view
• Discovery through example prompts, topic browsing, and asking what questions the Assistant can answer

Plan topics and structure

Topics form the main navigation path for users, so the structure matters. Create topics that reflect how users naturally think about the business rather than how you model your data. Good topic categories include Workforce, Turnover, Compensation and Pay Equity, Spans and Layers, Diversity, Recruiting, and Engagement.

Avoid topics that turn into miscellaneous containers. If a topic grows too large, split it into logical sections. Within each topic, arrange questions in a meaningful order. Start with broad or high-level questions, then follow with trends, comparisons, deeper analyses, and diagnostic questions. This creates a guided progression rather than a random list.

Use example prompts to make Answers easy to discover. These appear in the suggested prompts shown when a user opens the Assistant. Add What questions can you answer? along with a few high-value examples to help users understand how Answers works.

Write strong questions

The question you write determines the quality of the AI-generated explanation the Assistant produces. Strong questions are short, natural, answerable by the tile, and written in the voice of your real users.

Keep questions short and natural

Questions must be 100 characters or fewer. Keep them concise, and write the way HR business partners, recruiters, analysts, and leaders speak. Avoid jargon unless all users share it.

Prefer interpretive questions

Interpretive questions help the Assistant describe meaning, not just surface-level facts. Questions such as What does this trend suggest? or Where do we see risks? guide the Assistant toward more thoughtful responses than questions like Describe this chart.

Only ask what the tile can answer

The Assistant only uses the data in the selected tile. It cannot supplement missing breakdowns or additional time periods, generate forecasts, or introduce group comparisons that do not exist.

Avoid vague or undefined comparisons

If a chart does not contain a reference line or benchmark, the Assistant cannot safely answer questions that rely on one. Questions such as How do we compare to optimal ranges? are too vague. Instead, name the benchmark directly or add it to the visual — for example, How do we compare to our internal benchmark range of 6 to 8?

Use richer framing when it adds clarity

Instead of questions that produce simple numeric answers, frame questions to encourage interpretation:

• How is gender representation changing over time?
• Based on the current trend, are we moving toward gender parity?

These encourage more meaningful narrative responses.

Split overly complex questions

A two-part question works if both parts rely on the same logic. If the question mixes unrelated analyses, split it into two separate questions.

Use questions rather than commands

Commands such as Show me headcount for the last 12 months return basic summaries. Questions such as How has headcount changed over the last 12 months? produce clearer, more narrative explanations.

Quick reference — question patterns

Choose the right visuals

The visual behind an Answer plays a major role in the quality of the explanation. Choose visuals that show meaningful variation, clear structure, or patterns the Assistant can describe.

Use visuals with patterns for richer answers

The best visuals include time series, distributions, grouped comparisons, hotspots, scatterplots, geomaps, sankeys, and structured tables. These provide enough structure for the Assistant to describe trends, differences, and relationships. Single-value tiles work for simple KPI questions but produce short answers.

Use tables effectively

Tables work well if they contain a focused set of columns. Avoid broad prompts like Tell me about this table. Instead, write several targeted questions that guide the Assistant to specific insights from the table.

Note: If the tile shows Query too large to display, refine or filter the table until that message disappears. The Assistant cannot use a tile it cannot render.

Permissions and access

Answers respects all existing Role Based Security. A user only sees a question and Answer if they have permission to view the underlying Storyboard tile. The natural language explanation at the top uses its own access role, so if you want a group to see the chart but not the narrative, you can remove that permission.

Performance

The written Answer should not time out in most cases. Very large tables or unusually complex questions might take longer. If performance slows, reduce the amount of data the visual returns, or simplify the tile.

Maintain Answers over time

Treat Answers as a living content library. Review them whenever Storyboards change. Retire or update questions that no longer apply, and add new questions based on usage trends or changes in business priorities.

Use usage statistics to understand which topics and questions users rely on. This helps you uncover gaps and identify where more Answers would add value. Test periodically with HR, talent acquisition, and people analytics partners. Ask them what they want to know, and confirm that their questions route correct.

Always test each question after configuring it. Try several natural phrasings to confirm the Assistant interprets your intent consistently, and adjust the wording until you get a reliable, clear response.

Example question packs

Example question packs help you get started. Each pack contains twelve sample questions that demonstrate strong patterns, effective wording, and Series Types that work well with Answers. Use them as starting points and customize them based on your data, priorities, and language.

Available packs include Workforce, Spans and Layers, Compensation, Performance and Progression, Gender, Age, and Diversity. Recruiting Lifecycle packs are coming soon.

• One AI Assistant Answers - Imp Packs (1).pdf5 MB