Will AI Destroy HR Consulting?
1.76 trillion data points vs. some random stranger
Welcome to FullStack HR, and an extra welcome to the 238 people who have signed up since last week.
If you haven’t yet subscribed, join the 5000+ smart, curious HR folks by subscribing here:
Listen to the episode on Spotify or Apple Podcast.
Happy Monday (!)
I was off last week, spending some family time in Cyprus. It was a blast!
But now I’m back, and I'm launching a Generative AI and HR course on Sunday!
It's aimed at HR professionals, and its main focus is on how to work with AI in HR and, more importantly (in my opinion), how to help your organization work with AI.
It’s practical, down to earth, and built by me, who also happens to work with HR every single day.
When I turned 39 this summer, all I wished for from my wife was to get 4 days off from the family to go into my home studio and record this…
If you're a subscriber to FullStack HR, you've received an email with a discount code. Check your spam folder or search for "Johannes Sundlo" and you should be able to find it. Email me if you don't find it, and I'll fix it for you!
With the discount, the course is cheap! (Perhaps a bit too cheap… but it is what it is now.)
If you want to see the curriculum, head over to learnhrandai.com
But enough of the commercial break - let’s get to it.
I admit I'm part of way too many HR networks on Facebook.
And I kind of like it.
It lends me a sneak peek into what's cooking in HR and what we in HR are currently struggling with. It's a real-time snapshot of where the community is at.
After summer, I've experimented with one thing in several communities I'm part of. When someone has asked a question, I've taken that question into ChatGPT and then relayed the answer back to the person.
Or, I've just asked ChatGPT a similar question and compared the results from ChatGPT with the results that other humans are answering. Essentially, is ChatGPT answering in a similar way that a human would?
Take a moment to guess the outcome of these two small experiments before I give you the answer.
In the first instance, it's rare that I get taken seriously. That might have to do more with how I answer and that I just relay the answer provided by ChatGPT by providing a link to what ChatGPT answered. People usually comment on this, saying, "Haha, nice try and fun to use AI, but I actually need an answer."
In all instances, I've vetted the answer before submitting it, and I deem the answer, based on my own HR knowledge, to be correct. (But I have intentionally not presented it that way.)
In the second scenario, I've deemed ChatGPT to come to the same answers as humans have done; as long as it's not relying upon updated rules and regulations, then it gets it wrong for obvious reasons. This is most likely a diminishing problem as tools such as Bard and Bing, both connected to the internet, continue improving at similar tasks.
One of the examples I tested - first, the question
This is in Sweden so please take Swedish labour law into account. A question about interrupted sick leave. I understand that the substitute is pushed out, but I looked in our collective agreement to see if there is anything about the notice period to notify the manager. Am I remembering correctly that a person on sick leave does not need to notify x number of days in advance if they want to return to work? That is, the person on sick leave can come back on the day.
And then the answer:
Yes, you are correct. In Sweden, a person on sick leave does not need to notify their manager x number of days in advance if they want to return to work. They can come back on the day, even if their substitute must be pushed out.
This is because Swedish labor law gives employees the right to take sick leave when they are unwell and should not be penalized for doing so. If an employee's return to work disrupts the workplace, that is the employer's problem, not the employee's.
Of course, giving your manager as much notice as possible of your return to work is always good practice, especially if it is likely to cause disruption. However, you are not legally required to do so.
Here are some tips for notifying your manager of your return to work in Sweden:
Give your manager as much notice as possible, especially if your return will likely cause disruption.
Be clear about your availability and any restrictions that may apply. For example, if you return to work on reduced hours, let your manager know.
Be prepared to discuss any changes that may have been made to your workload in your absence.
Ask your manager if you need to do anything to prepare for your return.
If you are returning to work from interrupted sick leave, you may also want to discuss your return with your doctor. They can advise you on whether to return to work and what restrictions may apply.
It goes above and beyond to answer the question (even if it thinks I’m the one being sick…) and give practical advice on how to return to work. This is the correct answer and also the answer this person got from several people in the forum.
Since this week's update on Bard, it would be very interesting to perform the same test with Bard over the coming weeks, where you can also double-check the results through the G-icon.
So?
I have several thoughts about this.
But the first is an obvious one—we still lack trust in these models.
With "we," I mean HR.
That's both good and bad.
The good part is that these models still hallucinate, and the results are sometimes wrong.
The bad part is that, in most cases, they are right, which could save us a tremendous amount of time, effort, and energy if we trusted them.
(It's also a bit ironic that we trust complete strangers to answer our questions in a forum but not a large language model with 1.76 trillion data points.)
A recent and, from my point of view, insanely interesting paper (yet to be peer-reviewed) shows that consultants from BCG using AI finished 12.2% more tasks on average, completed tasks 25.1% more quickly, and produced 40% higher quality results than those without.
Most prominently, people who were less skilled increased their productivity the most, leveling the playing field among employees.
It’s an article worth reading as a whole; it’s packed with insights around the ins and outs of productivity gains and pitfalls - I hope we see more similar studies soon!
But to summarize it - overcome your trust issues. You’ll most likely benefit from it.
Secondly, I have a hard time seeing that this won't impact HR consultants who do short-term work, helping an organization with a specific problem. It has to, right?
If you have this on-demand HR consultant who can answer all of your questions at any given time and knows the ins and outs of the latest labor laws,—my bet is that we'll see an impact in the HR consultancy space in the next coming 3-5 years, as people adopt these new technologies.
Not saying that all HR consultants will go out of business, no, but some of them where you do short-term assignments on a particular set of problems will be impacted.
(Please email me in 2027 and remind me of this, and we'll see if I was right.)
But honestly, this is just a long text to say that I'm still puzzled that people are using random forums when looking for answers and not large language models.
Try Bard or ChatGPT next time.