Explore the technology behind our AI-Assisted Action Planning & Intelligent Nudges.


This article walks through:


Feature Overview


AI-Assisted Action Planning & Intelligent Nudges are available to customers who have purchased either the Activate or Grow packages as enhancements to the Perceptyx platform. Together, AI-Assisted Action Planning & Intelligent Nudges provide an intuitive, guided experience that helps employees understand the most important places for them to focus their action and naturally build new habits in the flow of work to drive impact where it matters most.

 

  • AI-Assisted Action Planning: When the Perceptyx AI Engine detects new survey feedback for a manager, it automatically analyzes the data and generates action plans for the manager’s review. These action plans can be modified by managers (including selecting different items to action). Then, the AI Engine proactively helps employees take action on these plans in their flow of work throughout the year.

  • Intelligent Nudges: Nudges are the main mechanism by which the AI Engine guides employees to follow up on their action plans. Each nudge contains a behavioral suggestion that helps the employee take action on plans and build new habits naturally in their day-to-day work. Every nudge is related to the org’s key business & talent priorities and personalized at the individual level. 


Our Technical Method


Throughout these features, Perceptyx utilizes narrow AI technologies as defined in our Perspective on Artificial Intelligence, Machine Learning, Natural Language Processing, and Generative AI (“Perspective on AI”). For AI-Assisted Action Planning & Intelligent Nudges, all data processing takes place within Perceptyx systems; no data is passed through external AI technologies from other vendors. No Generative AI is utilized by these features at this time. 


When new survey data is detected for an employee, the data is analyzed by our Most Actionable Items algorithm to determine which survey items are most likely to create business impact if put into action by that particular employee. This algorithm takes into account the following variables to recommend survey items for actioning:


  1. Drivers of Engagement: Determined by a proprietary algorithm that leverages positive divergence analysis, these are the survey items that are mostly likely to create an increase in overall employee engagement if their scores are increased. In summary, this algorithm identifies the most highly engaged employees in a given respondent group and compares their survey scores with other employees to determine the items that have the largest difference in percent favorable between the two groups.

  2. Significant trend down: Has the survey item trended down for this group of employees since the last available data?

  3. Bottom 5 below comparison group: Did this survey item receive low scores relative to other groups at the company? 

  4. Bottom 5 below global benchmark: Did this survey item receive low scores relative to external benchmarks from similar companies?

  5. Bottom 5 favorability: Did this survey item receive overall low scores from the employees included in this data set? 

Once the Most Actionable Items have been determined, Perceptyx uses a mix of proprietary automation technologies to proactively recommend items to each eligible employee (respecting existing “minimum N” privacy thresholds and customer preferences). Employees can choose to confirm the recommendations or to modify them. If they do neither, then Perceptyx leverages additional automation technologies to automatically create action plans for the employee. Once created, these action plans are visible to the employee on the Act page of the Perceptyx platform and can be modified or archived by the employee at any time. 


Once action plans are created, the AI Engine employs a separate set of algorithms to select Intelligent Nudges to send to each employee. These algorithms take into account the following data signals to personalize nudges at the individual level while ensuring that nudges relate to organizational priorities and employee listening data. 


  1. Survey data. Utilizing the Most Actionable Items algorithm, as described above

  2. Job level. We customize our recommendations based on employee’s job level (executive, manager of managers, manager, or individual contributor)

  3. Job category. Each employee is assigned one of ~30 job types, which vary from “knowledge worker” to “airline flight crew.” Job types are determined by analyzing 15 different job attributes, such as whether an employee interacts with customers, or whether they have autonomy over their calendar, in order to ensure that every nudge is relevant. 

  4. User-provided job attributes. Each employee can optionally provide additional information about their job that is not typically captured in HRIS systems, such as what kinds of meetings they routinely attend, or whether they typically work from home or in the office. 

  5. Personal goals. Each employee can select up to three personal growth areas, and receive nudges to help them grow in areas that matter to them. Personal growth areas are defined by the themes in our People Insights Model.

  6. Organizational priorities. Through the People Insights Model, we map our nudges to the talent and business priorities that matter most to each organization. 

  7. Various HRIS data. We account for attributes such as timezone, language, manager ID, and others to make nudges feel focused and personal. 


As employees engage with nudges, Perceptyx collects and analyzes various data signals, such as view rates for nudges, and the percentage of employees who “react” to nudges by clicking “yes”, “no”, or “I already do this.” We report some of this data to Admin users in the Act Metrics dashboard to help customers understand how their employees are engaging with nudges. We also use anonymized aggregate data to improve our nudges over time. Passive tracking of nudge view rates can be deactivated upon request. 


Future Developments


Perceptyx continually invests in new features to improve our product experience and impact. We are in the process of developing new algorithms that will leverage machine learning (ML) as defined in our Perspective on AI. For example, such algorithms may be used to provide more advanced nudge personalization based on past user activity or other employee attributes. 


As set forth in our Perspective on AI, our ML practices prioritize user privacy by excluding any customer-specific data from the training and refinement of our models. We adhere to strict anonymization and security protocols, ensuring that the development of our ML algorithms is conducted in compliance with the highest privacy standards.