Hiring In This Day And AIge

👋 I write weekly on AI development and its impact on society and culture. If you want to stay in the know, consider subscribing to this newsletter:

Hiring

I’ve been hiring for new roles and vendors for the past month and I myself have moved jobs in the last six. While I was working on data & AI projects previously, I didn’t have the time (or clearance) to try using AI tools in my day-to-day work. Most importantly, the tools I did try that existed at that time were just not cutting it for me in terms of performance and value. My day was consumed in meetings and emails on topics with a lot of history and context, and whenever I tried to use them to speed up work (e.g. to write an email), it would fall flat and I’d end up rewriting it from scratch myself (“whoohoo, I’m a better email writer!”). All that is to say, I thought it was a gimmick.

Those days are long done. I’ve been using AI tools extensively in my new job, not just to get sped up on the knowledge I need to execute, but also producing at scale even with time and skill constraints. I leverage them in two ways:

1. As a multiplier of existing skills (where I’m driving, not starting from 0)

2. As a senior collegue in new skills (where I’m taking a back seat and collaborating rather than coming in with a clear view of what I want as the end result, oftentimes starting from a blank page)

Most activities fall into both categories for me. Take coding for example: I could build an app on my own, but with my skill level, it would take me days or weeks. With these tools I might spin up one in hours.

It’s become easier to be a manager of one, and one feels like an understatement now.

Sure, there are still things I might do better but it would take me a whole lot longer to do than if I use an AI tool (hours compared to days). And that’s the whole point (at least for now): LLMs are multiplers of human performance, not (always) a substitute for it. Now, what they also allow me to do is be at least a junior in most fields. I’ve been using Canva a lot, and the feeling I get through it is very similar: where maybe 10 years ago we’d have needed to hire a designer immediately to get designs, now I can get by for some time before deciding to onboard someone new. The development is progressing at a faster rate than expected too: just a few months ago I would have to ideate creatives for ads on my own, and build them out in Canva. After GPT-4o, that doesn’t have to be the case anymore. Now I can just brainstorm with GPT and a full image creative comes out in a minute or so. It’s probably only a matter of time until that’s possible with video and other media too (I’ve already written about NotebookLM’s podcast-generating feature). I can then do a bit more work (or use existing tools) that allow me to e.g. look daily for inspiration based on system prompts, generate 10 creatives and launch new campaigns or generate landing pages to test out at scale. Introduce some agents to monitor campaign performance and iterate on results and soon you have a fullstack PPC manager. Almost all of this is already possible. Yes, some parts are clunky and not highest quality yet, but looking back it feels unreal compared to where we were just 2 years ago. Dharmesh Shah, co-founder at Hubspot, commented on his intuition that at some point we’ll be witnessing a “Linkedin for AI Agents” and I can totally see that happening. An economy of specialized agents that have the autonomy to contract each other seems like a possibility now.

Going AI-First

Before I get too philosophic about this, let’s get back to hiring. Duolingo CEO announced this week in a memo that they are going AI-first. That’s just weeks after Shopify did the same. What it means is best presented by these constraint points shared in the memo:

In summary:

• if you want to hire additionally, you have to prove you cannot “just” automate the work

  • not only hire in-house, but contractors too!

• you will be evaluated based on how well you can leverage AI in your work

The funny thing is this happened in parallel with OpenAI having some trouble with its latest update to GPT-4o. After the update, the model started being a bit sycophantic, which is a fancy way of saying people pleaser, which is a nice way of saying a suck-up.

OpenAI has in the meantime rolled back the changes and provided some information on what happened.

“If I wasn’t laughing, I’d be crying” is a nice way to describe this situation. Good thing is nothing (that) bad came out of this apart from a few annoyed users, but it could’ve gone a different way if not spotted sooner (thankfully it was so obvious about it).

Imagine having a colleague like that! Imagine, moreover, we were much more reliant on GPT that we are today and this happened.

Don’t like the new hire’s culture fit? Just roll back the change.

AI Race

Other stuff happened in AI too, here are some that caught my eye.

Getting Physical

AI is evolving into physical space. P-1 AI launched an AI engineer that specialises in industrial design. They’re hoping to deploy it to every industrial company on earth. Your next iPhone might be designed by AI?

Hugging face had an adjacent but less conquer-the-world-y announcement: they launched a $100 robotic arm that can pick up LEGOs and throw stuff in the bin. Affordable but not that exciting—let’s see where it goes. Hugging Face has open sourced some getting started materials, which means people will very soon be building cool and weird stuff with it.

Payments

For some reason, when I pictured what real progress in AI would be, it was always connected with making payments. “Sure, they’re smart and all, but can they order my Temu for me?” With the latest development from MasterCard, this will now be closer to reality. Agent Pay will “enhance generative AI conversations for people and businesses alike by integrating trusted, seamless payments experiences into the tailored recommendations and insights already provided on conversational platforms.”. Speaking in terms much closer to my heart:

Commerce

While we’re at shopping (the most prescient issue to be distrupted by AI), ChatGPT has integrated shopping options into its search. We’re probably months away from paid ads too, although Sam Altman has spoken in a recent interview he’s not the biggest fan of that route.

Coding

Anthropic’s Economic Index team analyzed 500k interactions with Claude relating to coding to arrive at the conclusion that it’s used mostly for automation (read: Claude does everything, you just sit back and eat cereal), by startups and building mostly user-facing apps.

Recruiting

If you do decide to hire, Mistral AI has cooked up a recruitment agent you can use to help you go through those applications at the speed of light.

That’s it for this week. Hope you find someone who likes you as much as ChatGPT likes humans.