Deep Research, Operator, and DeepSeek: The Datapoints That Redefine Work
The data points in one direction.
Welcome to FullStack HR, and an extra welcome to the 128 people who have signed up since last week.
If you haven’t yet subscribed, join the 8500+ smart, curious HR folks by subscribing here:
Happy Thursday!
In a month, I’ll be in Amsterdam and I hope to see you there.
If you are an HR Leader - tickets are free and I dare to say that attending is a must given what it’s happening (and given today’s topic as well).
Let’s dive in!
Back in the 90s, when I was in school, way in the back, on the top floor of Hagaskolan, there was this one room with a few computers.
In the beginning, only one of them was connected to the internet. We had to dial in—literally hear the modem screeching and beeping before we got online.
It was slow. It was unreliable. It was magic.
And I can still vividly remember the first time I used the internet at home, but I can’t tell you the exact moment when suddenly, everyone had it. It just crept in. One day, more computers were connected. Then, gradually, the internet became normal. And suddenly, it wasn’t just in that one computer lab—it was everywhere.
No fanfare. No momentous shift.
Just an inevitable, slow-motion change.
And AI? It’s the same thing but this time it’s faster.
The Problem With the On/Off Mindset
People keep talking about AI as if it’s a light switch: on or off, job loss or no job loss, disruption or no disruption.
But that’s not how this works.
It’s incremental. It seeps in. It becomes obvious in hindsight, but while it’s happening, we barely notice.
And that’s where we are now.
Instead of debating if AI will change work, we should be asking: Where are we already seeing it change?
If we start mapping it out, if we actually look at what’s happening—not the grand theories, but the real trends—we’ll see that the shift is well underway.
The Datapoints That Matter
ChatGPT Operator: A Glimpse Into the Future
If you haven’t seen my previous post or video on ChatGPT Operator, here’s a quick summary. Operator is OpenAI’s attempt at making AI more autonomous—capable of executing tasks without constant human input. Think of it like a virtual assistant that can handle complex workflows.
It’s far from perfect. It’s slow. It has quirks. But that doesn’t really matter. What matters is that it lifts the lid on a future where AI agents can act independently and integrate into workflows. It’s clunky now, sure. But fast-forward a few iterations, and it won’t be.
And that’s a datapoint - this is coming.
DeepSeek: It’s Not About Quality—It’s About Cost
Now let’s talk about another datapoint.
DeepSeek.
If you’re not familiar with it, DeepSeek is a Chinese company that a couple of weeks ago released their new AI model, R1 that’s positioned as an alternative to models such as ChatGPT o1.
It’s built with efficiency in mind, reducing the computational power needed to train and run large AI models.
Is it better than ChatGPT? No, not really. Despite what the benchmarks claim, in my own tests, it’s just as good—but not better. And that’s not the interesting part.
What’s interesting is that DeepSeek has managed to build a model that is significantly cheaper to train and run. And that’s the real takeaway: the cost of AI is plummeting.
Josh Bersin, a well-known voice in the HR space, recently made a post that from my point of view completely missed this point.
He focused on whether DeepSeek was safe. (Which it’s not and they very bluntly say so them selfes in their terms of use - they share everything you type in the web/app-version back to China) But that’s missing the forest for the trees.
The key takeaway isn’t about quality —it’s about cost.
And it’s yet another datapoint.
Deep Research: The Consulting Industry Just Took a Hit
Then there’s Deep Research from OpenAI. Deep Research is a tool that allows users to generate in-depth reports, insights, and strategies on virtually any topic.
Whether it’s leadership, recruitment strategies, or market trends, it pulls together an extensive research process into something that can be done in minutes.
I tested it.
An then Lars Schmidt posted on LinkedIn and asked if anyone had played around with it. And since I had, I shared the test above.
Then he asked —what if it could build a full-blown recruitment strategy?
Well, I had to try it. And guess what? It did.
It wasn’t perfect, but it was solid. (Especially the last part around new tools!)
And to be clear, this is work that companies pay consultants good money to do.
And now? You can get AI to do it at near-zero cost.
Again—by itself, Deep Research isn’t going to replace workers everywhere.
But in the grander scheme of things, it’s another piece of the puzzle.
Another datapoint.
Connecting the Dots.
Individually, none of these tools are immediate game-changers. (Well, Deep Research might be a game-changer…TBD)
But put them together, and the pattern becomes clearer:
AI is getting better.
AI is getting cheaper.
AI is getting more accessible.
And when those things, among the other things that is going on in the space, happen simultaneously, disruption isn’t a possibility—it’s inevitable.
White-collar work as we know it will change.
You have to present me with strong counter-arguments here to convince me that AI, sooner than we think, will impact work as we know it.
The standard phrase whenever I or someone else says that work will change is, “Things will work themselves out.”
Maybe. Maybe not.
But I don’t think we can afford to sit back and just let this happen to us.
We need to be active participants.
We need to test, experiment, and understand where this is going.
Joakim Jardenberg put it perfectly in a recent Facebook post. He transcribed a meeting and where he talked about how we need to shift from “AI helps me” to “I help AI.”
Sums it up perfectly.
That’s the mindset shift we need.
It’s not about AI doing things for us—it’s about us learning how to work with AI in new ways.
And let’s be clear—organizations won’t adapt overnight.
They never do. But they will adapt. And what happens then?
Let’s stop looking at AI developments in isolation.
Let’s stop pretending this is just another tech wave that might matter.
The trajectory is clear.
The trends are converging.
The cost of knowledge is approaching zero.
The accessibility of the tools are going up.
The autonomy of the tools are going up.
The only question is: Are we paying enough attention?