Rather predictably, perhaps, people analytics has tended to focus almost exclusively on numbers. One exception is sentiment analysis but that’s relatively rare.
However, as historian, Yuval Noah Harari, often says, it’s ‘language’ that is the operating system of civilization and no doubt it is the operating system of organizations as well.
So, if language really is the operating system, shouldn’t we be analyzing words more than numbers?
I think we should.
Consider the analysis of meetings.
Sure, we could analyze some numbers – such as how many people were in the meeting, how long it lasted, and who spoke the most.
But compared to what a human might glean from actually listening to the words, and I think the insights are potentially far greater.
Words are getting easier to analyze
Until recently, the ability of a people analytics team to analyze words was very limited.
Now though, with large language models, a new world of opportunity has opened up.
The biggest limitation of analytics may not be the language analytics tools themselves but our imagination on how best to use them.
An example of language analysis
Let’s imagine our concern is whether a project team is likely to hit its milestones.
If it’s not, then the manager may want to intervene. Here are some things we might ask the large language model to look for in the meeting:
- Was it clear who was meant to do what by when?
- Did people address contentious issues or dance around them?
- Was there too much disagreement? (or too much agreement?)
- Did some team members appear to have “checked out” and were not committed to the project?
You can see how radically different this kind of analysis is from the number-focused analysis we are used to.
Instead of gathering numbers and trying to glean insights from them, we just ask the large language model some direct questions.
We use the large language model much like we would use a human observer who was giving us feedback about a team meeting.
How to get started
Since meetings are such a common occurrence and are full of language, then analyzing meeting transcripts would be a good place to start in shifting people analytics from focusing only on numbers to focusing on numbers and words.
The people analytics team could work with a manager who is responsible for several teams and experiment with different prompts to see what a large language model could glean about the state of the project from the meeting transcript.
It might not amount to much, but it might provide important insights, you won’t know until you start experimenting.
Postscript: Did I lack imagination?
I suggested several specific prompts that you might use to see if anything was going off the rails in a project team meeting.
However, large language models are so smart that you could potentially ask something along the lines of “Make a list of common indicators a team is going off track. Review this transcript of a team meeting and report on those indicators.”
Large language models truly are astonishing tools.
It’s time that people analytics teams reoriented their efforts to take advantage of this new world of analysis that focuses on words not numbers.