AI Insights: A More Consistent Way to Understand Survey Results

AI Insights helps move from survey data to understanding more quickly by translating key results into a concise, plain-language narrative. Rather than asking every user to interpret a complex dashboard on their own, AI Insights highlights the signals most worth paying attention to and explains them in context. The goal is not to replace analysis or human judgment. It is to provide a more consistent, defensible starting point for understanding what the data is showing, so users can move from data to insight, and from insight to action, faster.

Built on Perceptyx best practices

AI Insights is intentionally designed to be conservative in what it surfaces. The agent does not freely explore the survey or decide on its own what is important. It works from a defined set of results and applies rules based on Perceptyx analytical and consulting best practices.

That means:

• Insights must earn a mention. Items, trends, and groups must meet predefined criteria before they are highlighted.
• Multiple forms of evidence may be required. Statistical significance alone is not always enough. Where appropriate, results must also clear practical thresholds such as the size of a difference or change.
• Language reflects the strength of the evidence. A benchmarked strength, a relative strength, and a highest-scoring item are treated as different findings and described accordingly.
• Weak signals are not forced into the narrative. If an item or group does not meet the criteria, it is left out. If an insight cannot be supported with sufficient evidence, that section may not appear at all.
• Findings stay grounded in the data. AI Insights only summarizes information supported by the underlying reports and avoids unsupported causal explanations or diagnoses.

The standards behind each insight

Each AI Insight has its own criteria for what can be surfaced. The rules are designed to keep the narrative focused on evidence that is both meaningful and appropriate to interpret.

Executive Summary

The Executive Summary brings together the findings that qualify across the other insights. Response rate is only called out when it is notably strong or low, and the same standards used in the underlying cards still apply. The result is a concise synthesis rather than a restatement of every metric.

Recommended Focus Areas

Driver insights focus on the top Priority Focus Area drivers for the configured outcome measure. Additional characteristics, such as benchmark position, low favorability, or meaningful trend, are only referenced when they meet the required criteria. Driver findings are described as priorities for action, without suggestion causation or suggesting they fully explain the outcome.

Strengths

Strengths first prioritize items that are meaningfully above benchmark, defined as five points or more. If no items meet that threshold, AI Insights shifts to the items furthest above benchmark and frames them as “relative strengths.” If no benchmark is available, it instead draws from the survey’s most favorable items. The language reflects the strength of the evidence, and only a limited number of items are surfaced to keep the summary focused.

Opportunities

Opportunities follow the same stepped approach in the opposite direction. Meaningful negative benchmark differences (five points or more) are prioritized first, followed by relative benchmark gaps, then lowest-favorability items when benchmark data is unavailable. This prevents a low score alone from automatically being treated as a significant organizational concern.

Comments

When Narrative Analysis Agent or Comment Copilot is enabled, AI Insights uses the generated comment summary and surfaces a limited set of the strongest positive and negative themes. The intent is to provide a concise view of what employees are saying to supplement the quantitative story.

Change Over Time  coming soon

Trend findings will focus only on changes that are both statistically significant and large enough to be practically meaningful, at least three percentage points in either direction. The experience will compare against the most recent configured trend survey and limit the number of improvements and declines elevated into the summary.

Hot Spots  coming soon

Hot Spots will apply multiple criteria before identifying a demographic group as meaningfully different from the organization overall. Group size, statistical significance, and the magnitude of the difference all have to meet the required standard. This helps protect against overinterpreting small or unstable groups while also supporting confidentiality.

Behavioral science behind the experience

AI Insights is designed around a simple idea: the agent does not go searching for a story. It summarizes the specific results it is given, using interpretation rules defined in advance.

Three behavioral science principles shape that experience:

• Behavioral scaffolding: AI Insights embeds appropriate interpretation directly into the dashboard, helping users understand key signals without requiring deep statistical or employee experience expertise. The goal is to support, not replace, human judgment.
• Cognitive load reduction: Survey dashboards can be dense, and meaningful signals are easy to miss. AI Insights helps answer the “so what” quickly, making it easier for users to focus attention on what matters most.
• Narrative sensemaking: People make sense of complex information through coherent stories, not isolated metrics. AI Insights organizes findings into a contextual narrative that is easier to understand, discuss, and act on.

Those same principles also help address predictable ways people can misinterpret survey data. Negative findings tend to command more attention and urgency than equally strong positive results, and low-scoring items can easily be treated as the highest priority even when they are not the most important or impactful areas to address. People also naturally interpret new information through the lens of what they already believe about a team, leader, or organization, making it easier to reinforce an existing narrative than to reconsider it. Once a pattern appears in the data, there is also a strong tendency to move quickly from what happened to why it happened, even when the survey itself cannot support that causal explanation.

Complexity adds another layer of risk. When users are faced with many scores, comparisons, trends, and demographic differences at once, they may gravitate toward the most visible, familiar, or emotionally salient result rather than the most meaningful one. And because numerical differences are easy to notice, a score gap can feel important simply because it looks large, even when it is not statistically or practically meaningful. In each case, the risk is the same: giving a signal more weight or certainty than the evidence supports.

AI Insights is designed to counter these tendencies through consistent thresholds, balanced framing, evidence-based language, and limits on what gets elevated into the narrative. The result is a more disciplined starting point for interpretation, while leaving contextual judgment and decision-making with the human.

AI supports judgment, it does not replace it

AI Insights can help answer one of the first questions that comes up when reviewing survey results: “What should I pay attention to?” What it cannot provide is the organizational context behind those results. Leaders, HR business partners, and consultants still bring knowledge of the business, what was happening when the survey was fielded, what actions have already been taken, and what response makes sense now.

That is the role AI Insights is designed to play: a disciplined first layer of interpretation that helps users get to a defensible “so what” faster, while keeping human context and judgment at the center.