🇬 🇱 🇪 🇳

Recruiting, Sourcing, Strategy, Innovation, Ethical AI, Data/Analytics, Search/Match, Lean, Agile, Social Engineering. Speaker, Blogger, INTJ, He/Him

[🇬 🇱 🇪 🇳 @GlenCathey]

If you're in sourcing & recruiting, I recommend you watch this video or listen to Lenny's podcast of Adam Ward at Cursor talk about how to build "high talent density" teams. As a sourcing/recruiting geek who loves the craft, it was great to hear Adam talk about what I've always loved about targeted & true passive talent recruiting. The episode is quite rich, covering the whole lifecycle from working with hiring managers on the role, performing the research, being incredibly thoughtful and strategic about outreach, first conversations, the courtship process, effective referral recruiting, etc. - all the way through to extending offers and pre-boarding to reduce/eliminate offer reneging. It was also nice to hear we're aligned on the pre-closing - something I don't think enough people appreciate or practice.

Worth a watch/listen if you're passionate about the craft of recruiting!

https://t.co/LF8fTumgAg

[🇬 🇱 🇪 🇳 @GlenCathey]

"Botsitting" AI is a feature, not a bug. It's what I've been calling the "Two Job Problem" for a while now. Multiple studies (BCG, Glean) are finding that nearly 50% of people working with AI spend more time "managing, directing, and overseeing AI tools than performing the underlying work tasks themselves."

But this should not be surprising to anyone.

AI is a work resource.

Like any resource, it needs direction and management.

Unlike most human resources, AI doesn't and in most cases can't fill in the gaps with basic instructions, so it requires more management.

I've been saying that AI skills are essentially management skills for about a year and a half now.

AI requires you to identify and explain the problem/work, provide appropriate context to enable the AI to do a decent job on the work (including what good looks like), review and provide feedback on the work, iterate and coach the AI to produce better output, etc.

This "botsitting" work is new for individual contributors, so I don't (and you shouldn't) expect most people to be great at it without training, practice, and experience. And who would expect it would not take time to manage AI as a work resource?

While managers have management experience, managing AI to perform work is a bit different than managing people, as I mentioned above. Most people carry quite a bit of context with them as well as an understanding of what good looks like, so people can take relatively lightweight instruction from their manager and fill in the gaps. But AI doesn't - unless you provide it/fill in the gaps for it.

Expecting to give AI to individual contributors, let alone an experienced people managers, and for them to get good results from AI without investing the time required to do so ("botsitting") is absurd.

I've coined the "Two Job Problem" to represent the fact that when you give individual contributors AI and expect them to use it to do work, you have essentially expanded their responsibilities and they now have two jobs: 1) doing the work, and 2) managing AI to do the work. They become producing managers, and for anyone who's had that kind of job (responsible for personal contributions as well as managing other resources - which I have), they know this is challenging.

I work with AI day in/day out, and while I can say that I can do more in less time with AI, it comes at a cost - "botsitting." Even with the frontier models, and with plenty of guidance, context, examples, etc., the first output is rarely usable. I often feel I am wrestling with the AI to get something I can use.

Just yesterday, I needed to create a single slide visual of a 47-step TA workflow. It took 4 different AI tools, 13 attempts and 2 hours before I got something usable. Granted, while the whole process was painful, it produced something I don't have the skill to create manually, and even if I did - it would have taken me a LONG time to do so.

Again, "Botsitting" AI is a feature, not a bug. Account for it.

[🇬 🇱 🇪 🇳 @GlenCathey]

Here's a prompt that turns any AI into an employer brand analyst. Give it a company + role + location → it searches Glassdoor, Indeed, Blind, Reddit & more, then synthesizes everything. Job seekers: use it before your next interview. Companies: your candidates already will.

https://t.co/wAlfDyfOb2

[🇬 🇱 🇪 🇳 @GlenCathey]

I'm sorry, but I can't understand having billions and being bothered by these kinds of taxes. You'd be surprised to know taxes on the wealthy in the U.S. peaked between the 1940s and early 1960s, with top marginal income tax rates reaching over 90% (peaking at 94% in 1944). These high rates, largely enacted to fund WWII and the subsequent economic expansion, remained above 70% until 1981. The taxes we have for top income brackets today are cute in comparison.

[🇬 🇱 🇪 🇳 @GlenCathey]

Check out my latest article: AI Fluency Isn't About Prompting - Anthropic's New Research Proves It https://t.co/ato8QXRDPp via @LinkedIn

[🇬 🇱 🇪 🇳 @GlenCathey]

@mikefutia REELS

[🇬 🇱 🇪 🇳 @GlenCathey]

You'll probably never see this, but I actually tested CoT with o1-preview and it worked well, at least back in September. I used a practical and measurable work example comparing 4o w/CoT, o1 preview w/standard prompt, and o1 preview wCoT. o1 preview w/CoT performed the best. Just ran it again - still works well. https://t.co/QuX5ZWGMla

[🇬 🇱 🇪 🇳 @GlenCathey]

Check out my latest article: Sourcing/Boolean Search Test: OpenAI o1-preview vs. 4o w/Chain-of-Thought Prompt https://t.co/mxDLG22gs8 via @LinkedIn

[🇬 🇱 🇪 🇳 @GlenCathey]

Check out my latest article: The use of generative AI by job seekers in the hiring process and what you can do about it https://t.co/TVVYmEGcnd via @LinkedIn

[🇬 🇱 🇪 🇳 @GlenCathey]

I appreciate this reference and meme - who can name the movie, the boy, and the horse? #talent42

[Andrew Ng @AndrewYNg]

1/ Thrilled to announce: Our new course ChatGPT Prompt Engineering for Developers, created together with @OpenAI, is available now for free! Access it here: https://t.co/OaIpa6L2jn

[Chris Hoyt - President of CareerXroads @TheRecruiterGuy]

We're opening our next @CareerXroads lecture up to ANYONE who would like to join us to celebrate #Nuerodiversity Week with our latest instructor, @GlenCathey.
You won't want to miss this awesome #DiversityandInclusion session!

https://t.co/T04TIP8iaN

[Andrej Karpathy @karpathy]

The hottest new programming language is English

[Paddy Jobsman @paddyjobsman]

I’ve used ChatGPT to write my last 5 client resumes. I can confirm you never need to hire a resume writer again. Writing a resume is now a solved problem.

[Jon Uleis @MovingToTheSun]

My new favorite thing - Bing's new ChatGPT bot argues with a user, gaslights them about the current year being 2022, says their phone might have a virus, and says "You have not been a good user"

Why? Because the person asked where Avatar 2 is showing nearby

[Ethan Caballero @ethanCaballero]

"This new Bing will make Google come out and dance, and I want people to know that we made them dance." - @SatyaNadella

[🇬 🇱 🇪 🇳 @GlenCathey]

Check out my latest article: ChatGPT: Amazing Free AI Tool with Many Uses - Including Recruiting https://t.co/91lo7hZIuD via @LinkedIn

[BBC Archive @BBCArchive]

What will the office of the future look like? 🖥️

#OnThisDay 1979: Judith Hann's look at homeworking technology gets a lot of things very nearly right.

[🇬 🇱 🇪 🇳 @GlenCathey]

@Metallica Hard to pick just one, but if I have to - Shortest Straw

[🇬 🇱 🇪 🇳 @GlenCathey]

@talentgenie @TangieRecruiter @SourcingRocks @peopleshark @jantegze @ohsusannamarie However, I could also argue that the need is dependent on what the sourcer will predominantly be focused on sourcing, as well as whether or not the employer/client is looking for the best that can be found or if they are happy with the best of what a single algorithm returns

[🇬 🇱 🇪 🇳 @GlenCathey]

@talentgenie @TangieRecruiter @SourcingRocks @peopleshark @jantegze @ohsusannamarie Blast from the past! From my experience evaluating multiple AI-powered search/match solutions with the same dataset, I think search/information retrieval is still important for sourcers for learn as any solution's algorithm is essentially a canned report...