• On The Fly
  • Posts
  • #141 AI Agents = Personal Secretaries

#141 AI Agents = Personal Secretaries

Plus: What is Workslop?

Author

Dan Healy
14 Oct

Welcoming you back to our Productivity Edition 🕺 đŸ•ş đŸ•ş 

Loyal readers know I’m All-In when it comes to AI, and OTF’s persona in general is one of positivity and encouragement, but the more advanced and complex these tools become, the more I feel it’s our responsibility to talk about the CONS as well as the PROS.

Like every great development, we gotta take the good with the bad. But that doesn’t mean we ignore the bad completely.

Not to say that we’ve had our head in the sand - we’ve talked about downsides in the past, but this edition feels like we take it a little bit further:

  • Workin’ With AI dives into the pros and cons of AI Email Agents

  • Always Informed talks about the always-annoying generative AI hallucinations

  • Even Hustle Hub links to a really interesting article about “workslop” and what it’s doing to workplace productivity

This edition is informative AF, which is super important to us. Because to really get after it, we need to understand both the risks ⚠️ and the rewards🏆

AI HUMANS Prompted These Highlights!

  • Workin’ With AI: How To Build an Email Agent 🚀 đŸ’°ď¸

  • Always Informed: Why Does it Feel Like ChatGPT’s on Hallucinogens? 🚀 đŸ“ˆ 

  • Hustle Hub: Brand Elevation, Workslop, & ChatGPT Money Advice 🚀 đŸ’°ď¸ đŸ“ˆ

ALWAYS INFORMED 🗞️ 

An AI Flaw That Can’t Be Fixed?

For those that primarily use generative AI as a Google-replacement on steroids (which is probably a lot of us!), it’s incredibly annoying when the bot’s output is inaccurate or, worse, straight-up false.

The word used to describe AI’s confident false claims is hallucinations, and if you’re not careful (i.e not prompting properly), well, it’s like ChatGPT took some magic mushrooms…

This seems odd for such a powerful tool - why can’t you give me the right answer for a basic question? And if that’s not possible, why do you have to make sh** up sometimes?

Maybe it’s just a bug that’ll disappear as the tech evolves?

Well, according to a September article from Futurism, it seems like hallucinations are here to stay.

TL;DR Takeaways:

  1. Generative AI is “optimized to be good test-takers,” which means guessing when you don’t know the answer.

  2. The AI industry isn’t incentivized to eliminate hallucinations, because:

    1. It could “dramatically increase costs.”

    2. Having AI admit that it doesn’t know an answer could erode user trust and result in abandonment of the tool. (Basically, people prefer a confident wrong answer than an honest “I don’t know,” which unfortunately checks out.)

  3. The article’s conclusion? “The business incentives driving consumer AI development remain fundamentally misaligned with reducing hallucinations.”

O.T.F. 
O.F.T - One Final Thought

We are no experts, but based on everything I know from years of reading, trial, and error, your best bet to minimize hallucinations is to request sources/links in your prompts. And even that might not solve the problem unless you’re willing to vet those sources for yourself.

This is how AI is supposed to work though, right? It’s not supposed to replace all your work and effort; it’s just supposed to make it more efficient and effective.

WORKING WITH AI 🤖 

The Night I Watched How To Build an AI Agent For My Inbox…🤯 

One night, I decided to watch a workshop about building an Email Agent on the AI-powered platform Lindy.ai.

(I talked about this 2 months ago about how AI agents are basically digital employees. I went in-depth about what they are and what they can do; if you’re interested in that context, click above.)

Why do this? Because having automated systems to eliminate boring or repetitive work is how we boost productivity.

Example: Every morning, you log in at work and check your inbox, and you groan at all the generic responses you’re gonna have to send:

  • “Got it! Thanks for the update.”

  • “Sounds good, I’ll circle back once reviewed.”

  • “Thanks for sending this over — I’ll loop in [Name].”

  • “Approved on my end — proceed.”

  • “Thanks for your patience — I’ll get back to you shortly.”

Now, imagine you could build an AI agent that could read those emails, analyze the tone, and drafts responses in your style/tone. 

(Note: it wouldn’t automatically send the emails, (even though you can set that up…maybe for another edition), but it saves you plenty of thinking/drafting energy. Even if you have 30 emails to respond to and you don’t like 20 of the bot’s drafts, it still saves you 10 emails’ worth of time.)

Now, let’s go deeper into the PROS and CONS ⬇️

PROS:

  • It’s a Time Saver: You cut your email workload by letting an Agent handle the repetitive email responses. Freeing up…👇️ 

  • Mental Clarity: …your headspace for actual thinking.

  • Speed to Respond: Working off of drafts is faster and easier than staring at a blank screen thinking about how to begin.

    In short - saves you time, energy, and frustration. Win, win, win.

CONS:

  • Setup and Learning Time: The initial training and prompt-building takes real time and effort. Massive learning curve, but worth it.

  • Privacy Concerns: This is my main concern, seriously. Agents read your inbox, so you need strong data security protocols. Or - what I’m doing - making a separate email that only handles emails that aren’t deemed confidential.

  • Exceptions: Some emails still need your human judgment, empathy, and original response to sensitive matters.

For those still new to AI, this seems like a big leap. I’m still leery tbh, especially about those privacy concerns. But I’m sure people once felt similarly about putting their CCs on the internet when online shopping - it’s just gonna take time to get comfortable and understand how to do it safely. So if you’re not ready to jump in with both feet yet, I encourage you to at least dip your toe by learning about it.

HUSTLE HUB  đŸ“ˆ 

YOUR SUCCESS:

Be ELEVATED - One of OTF’s three pillars is about elevating your brand (both personal and professional); here’s how AI can help!

THEIR SUCCESS:

Be Inspired Warned - A writer asked ChatGPT about the worst things you can do with your money, and everyone’s favorite bot was on point with its response.

YOUR SUCCESS:

Be More Productive - AI can absolutely increase productivity, but if you’re not careful it can destroy it.

Before You Go

Enjoy this viral Instagram post - it’s so damn good 🤣 

Next week, we’ll be back with October’s MONEY EDITION. Prof Mike’s got that one covered, since I’ll be on my honeymoon! Can’t wait to see what he has in store…

As always, see you next Tuesday!

Find Dan on LinkedIn

You are now On The Fly & In The Know.