Beyond Bias: AI's Promise for a Fairer Workplace
Introduction
This LinkedIn post by Allie Miller shows how “pacesetter” organizations are outstripping the adoption of AI when compared to their less-advanced peers. It breaks out which functions within organizations are using AI, and at what rate. (link to original article).
What can we learn from re-shaping and examining this data? I call this “napkin analytics” because you could do it on one or two napkins.
Here’s the focus: where is AI being adopted more, and less, within organizations? A deeper dive into this data reveals a fascinating outlier: AI is being most aggressively used in one key People-led function.
Creating a New Metric: Outstrip Factor
In reviewing the data, it’s clear there are different uptakes of AI. I created an Outstrip Factor metric - divide the % of adoption in the Pacesetters by the equivalent value in the Others.
It’s unsurprising that in a Data-focused function, there’s a lot of uptake - Pacesetters have an 80% higher adoption rate than Others.
By contrast, Hiring amongst Pacesetters is about the same as Others. That might suggest on its face that People teams aren’t incorporating AI, but I’ll explain in a minute why that’s not true.
I then stack-ranked the report by Outstrip Factors. The graph below shows where Pacesetters are more comfortable relying on existing methods of operation and where they are investing in AI
It appears that for the most part, Pacesetter organizations are adopting AI in the most technical functions. However, the fascinating outlier is Performance Management. I’ve grouped it in the “People” business area, since this task, while conducted throughout companies, is typically led by the People organization.
So, AI is being used here at a rate that outstrips even some of the technical functions. That’s amazing. What’s even more fascinating is that this function is the place where AI Pacesetter organizations have adopted it by such a wide margin (the Outstrip Factor) as compared to the Others - 1.9x.
To summarize:
Pacesetters are adopting AI for Performance Management at almost twice the rate of their Others counterparts
As well, within those Pacesetter People organizations, AI is being used in Performance Management at a rate that’s almost 2.5x that of Hiring.
Here’s where the data stops telling the story and I’m prompted to ask one of my favorite experts, Cynthia Abbott, why this might be the case.
“Over the decades, we have compiled data that supports the statement “Performance reviews are heavily biased and say more about the person delivering the review than about the individual receiving it.”
What this means is that as humans we suffer from recency bias. The person who made me happy yesterday, or the one who just gave me a nice compliment, will receive the best review when I go to write it. Often, this bias is unconscious; most people aren’t malevolent, just ill-informed. While this behavior can create a turbulent employee experience, biased decisions regarding performance – decisions that can also impact pay – can also have major legal implications.
Thus the introduction and heavy integration of AI (Artificial Intelligence) in the performance management process. The AI can be trained to be unbiased whereas humans, even fully trained, still carry bias in their decision making. (Can we actually untrain racism, sexism, anti-intellectualism?)
As we continue to learn more about and lean into AI, the accuracy around performance calibration and performance expectations will become more exact, less emotional, and more equitable – a better experience all around.
Conclusion
There’s a lot of energy - much of it negative - about AI in the workplace and our lives. If, as the data from this report suggests, it’s being used to improve a critical People/HR function, that’s very encouraging. That said, this data doesn’t allow us to draw any conclusions about the outcome of this work.
Pacesetter organizations have identified Performance Management as being one of the areas where the investment of AI brings with it the most promise.
I’m looking forward to seeing what next year’s Enterprise AI Maturity Report tells us.
Appendix
Supporting Data
You can find the supporting data here, in my repository.
ChatGPT Transcript
I used ChatGPT to create and manipulate these data tables from the source image. Welcome to no-code analytics!
User: Please pull the figures out of this image and turn it into a markdown table.
Image uploaded
Assistant: Provided a markdown table with the data extracted from the image.
User: Please add a column where you divide the pacesetters # by the Others #. Call it "Outstrip factor". For example, 76/42 = 1.8.
Assistant: Added a new column called "Outstrip factor" with the ratio of Pacesetters to Others for each category and displayed the updated table.
User: Please round the outstrip factor to one decimal place.
Assistant: Rounded the "Outstrip factor" values to one decimal place and displayed the updated table.
User: Please sort this table in ascending order of outstrip factor.
Assistant: Sorted the table in ascending order of the "Outstrip factor" and displayed the updated table.
User: Please get rid of that first, unnamed column.
Assistant: Removed the unnamed index column and displayed the cleaned table.
Gratitudes
Thank you to Christine Potts and Rohn Jackson for feedback on early versions of this draft. Hats off, though, to Cynthia Abbott, whose domain expertise unlocked the story behind the data. Cynthia’s hosting a conference about the future of work, on October 23rd, 2024. Follow this link to learn more.
Thoughts? I’d love to hear your perspective. Ping me at davehk@gmail.com or let’s have coffee; real or virtual: https://tidycal.com/davehk/30-minute-coffee