To coincide with Workday Rising EMEA in Amsterdam today, Workday has unveiled AI-powered features for Workday Peakon Employee Voice, its employee feedback application. These new capabilities help managers in large organizations get early warning of emerging issues or drill down into specific geographies, departments or other workforce segments. Another new feature automatically triggers tailored surveys at specific junctures in the employee lifecycle, such as onboarding or exit.
The new features in Peakon, due to roll out next year, reflect its growing role in collecting employee feedback in addition to the numerical sentiment rankings the product initially focused on -- all of it anonymized to encourage frankness. Indeed, the new tailored survey feature is in part aimed at collecting more of this qualitative feedback, which is particularly valuable in smaller teams that otherwise may not generate many comments. But in large organizations, there's the converse problem that team leaders may not be able to read all the comments that come in, and therefore the new functionality gives an AI-generated summary that, for example, lists the top positive and negative issues mentioned by employees. Anne-Sofie Nielsen, VP, Product and Engineering, Workday, who was CTO of Peakon before Workday acquired the company in 2021, explains:
With AI summaries they get a summary of: the people that are very positive right now, they appreciate these things; and the people who are really negative, these are the things that are top of mind for them. So they get a really quick summary to try and understand what's working well in my organization, and where is there [a need for] improvement...
If you think about Peakon over time, we came from initially just focusing more on scores, and where are you scoring high and low? We have a lot of built-in analytics to help you understand which parts of your organization might be scoring higher or lower on certain aspects. But what we're seeing is more and more of the gold really is in those comments, where people leave more detail for you to help action what's coming out there.
Whereas managers of smaller teams will likely respond to comments directly, this feature is aimed at those who manage larger numbers or sit at higher levels of an organization, particularly where comments may be coming in multiple languages -- the new capability supports more than 60 languages. An automatically generated digest can help managers discover and act on issues as they arise, rather than only finding out about them later or not at all. She goes on:
I think the people who get the most value from these summaries have so many employees that it's not really realistic [to respond individually] -- I'm [personally] not getting a lot of responses to my feedback directly from the Workday CEO. He probably has other things on his mind. But I hope that my feedback goes into one of the themes that the company actions. One of the things we see with the companies doing this really well, is that they're good at also communicating back to employees, [such as] 'We've changed our flexible work policy based on this feedback that we've seen.'
Workday emphasizes that these AI capabilities are informed by Peakon's dataset of more than one billion aggregated employee responses and 200 million pieces of written employee feedback. Nevertheless, Nielsen concedes that it's still important to validate the findings being reported by the AI before embarking on significant changes. She goes on:
Before taking any big actions, we would, of course, always recommend that people go through and check, maybe use our semantic search capabilities or something else to understand more deeply those topics. But I think this helps people become aware of what might be some things to try and tackle. Typically, then, if you see, 'Okay, there's an area here that's trending,' you dig a bit deeper and you look at, 'Okay, what are maybe some associated scores? Where is it scoring lower or higher in the organization? Is this a local problem? Is this a global problem?'
We don't typically see people just reading a summary and then going out and doing something in the organization. Typically, they will take more steps. They might use that to say, 'Okay, let's set up a focus group to talk about how we could improve this,' or something else.
A related new capability provides the ability to request comment summaries from specific subsets of the workforce, based on factors such as location, department or engagement levels. These on-demand summaries can give a real-time view of employee sentiment in these specific subsets. For example, says Nielsen:
A leader... can go in and say, 'Okay, for my London office, what are people currently saying about health and well being?' That might still be a large number of responses to go through. An HR business partner or just a leader of that organization could say, 'Summarize that, and let me understand what's top of mind for those people.'
Because Peakon is part of the Workday stack, responses and summaries can be classified by any metric that's stored in Workday, including gender, ethnicity, function, role, or geography. Alternatively those categories can be implemented directly in Peakon.
Some Peakon customers also correlate responses with other data -- with appropriate consents in place -- by analyzing it in the Workday Prism data analytics tool. Nielsen says:
We see people really making some very interesting analysis with pulling in other data sets, whether that's more business-related data, like analyzing the output of restaurants or something else based on different employee engagement scores, but it could also be correlating with 'Who's taking this training program? Do we see that it's having an effect?'
[Such as] leadership training -- 'Do we see that those managers are then receiving better engagement scores from their teams post this training?' So it gets cross-referenced with lots and lots of other data points.
Analysis at scale is one of the things AI excels at. Whereas in the past it rarely made sense to manually analyze hundreds or thousands of individual comments and feedback to track engagement, wellbeing and other metrics across a large workforce, it's a relatively simple task for AI. Although, as Nielsen cautions, the AI analysis should never become the sole source of evidence, the probabilistic nature of generative AI is less likely to introduce serious errors when analyzing this kind of 'soft' behavioral data than it is when dealing with hard transactions. Bringing these new capabilities to Peakon is therefore an obvious application of the technology.
