- With Agents
- Posts
- It's not a perfect way to think of personal AI Agents - but close
It's not a perfect way to think of personal AI Agents - but close
An insight, a use-case of agentic AI, three new shareable things.
Clay Griffith
January 23, 2025
Audio above is AI-generated, duh.
withAgents is for everyday people who want to personally keep up with this fast changing space and wield that knowledge to their advantage. The insights are casual and the use-cases straightforward. Glad you’re here.
An insight
Think of them like a VA.
And aren’t they?
Virtual Assistants. You know the deal. Hands-off, low-opinion-unless-asked, focused. Highly skilled and professional at administrative tasks while (hopefully) in low (to no) need of day-to-day management. VAs.
A lifeline for the Work Harder Not Smarter à la Four Hour Work Week movement, VAs have had a good ~20 year run helping executives, small business owners, creators and slackers get good value for their money.
“Time is the most value asset.” Or, as I say, “I’d rather be doing literally anything else.”
Delegate the easy stuff.
As we’ve dived into the budding AI Agent space over the past few weeks, a resounding question comes up in the threads and posts: “WHAT ARE THE USE-CASES!?”
Fair. The Vertical Agents and B2B use-cases are neat but not our main concern here. At withAgents, we’re focused on how individuals can or will someday use AI Agents to get more done, personally. So for real, what are those more personal use-cases?
Well, in theory, the only limit is our imagination. Agents could feasibly do countless things if properly built and well-inspired. But I’m feeling low imagination right now… Time to go to the internet. And after a quick roundup of the common VA job categories found online, we’re getting somewhere.
So, what do VAs do and how do current agent capabilities compare?
VA’s often do this | AI Agent Solution Available? | Notes |
|---|---|---|
Scheduling Meetings | Yes / Build Your Own | This is a super common use-case, even in low-code AI workflow builders. |
Managing Emails | Yes / Build Your Own | This is a super common use-case, even in low-code AI workflow builders. Beyond sorting, they reply and interact. |
Travel Planning | Yes / Slightly More Complex | Totally doable. Check out things like Wonderplan and Mindtrip. |
Social Media Management | Mixed Feelings, but Yes | This is another super common use-case with countless sub use-cases. Way beyond AI influencers. |
Data Entry | Yes / Build Your Own | Very common use-case as well. Many pre-built methods and tools and integrations here. |
Bookkeeping | Partially | Think of AI more as an accelerator to this work vs. fully displacing bookkeepers now. |
Research | Yes | See: Google Gemini Deep Research, ChatGPT Tasks, Perplexity, on and on. |
Customer Support | Yes | Another massive adoption area already. We still think ‘human in the loop’ matters, but in terms of doing what a VA does — mostly yes. |
Content Writing | Yes | If there is a use-case for outsourcing copywriting to VAs, it’s likely also a good Agentic use-case. Tons of tools in this space for Agents. |
Graphic Design | Partially | Our opinion here is that AI is good for research, look books, idea grouping, and other procedural tasks but leave good Graphic Design to people |
As you see here, AI Agents are already plenty capable in most of these VA categories. You can easily build solutions — often in no-code platforms — or find pre-built stuff.
It’s not even close to a perfect outline of Agent use-case for people like you — we all have unique workflows and needs. But perhaps this gets you thinking.
Low hanging fruits.
Shoutout to “Blake” (didn’t catch the last name) from Medici Coffee/ATX Writing Club on Friday for this newsletter topic inspo.
P.S. Wanna see what AI Agent builders are wielding to get more jobs done?
We put together a public spreadsheet of hundreds of integrations, tools, and triggers commonly found in AI Agent platforms. Get the gears turning, give it a look.
Open the research sheet here (Google Sheets)
With Agents, you can…
3 can’t miss demos and explainer vids, found this week
We’re starting something new for this section. Instead of a single demo, you get the BEST of the week. We doom scroll / brain rotted our way through a pile of clickbait low-quality videos to find the golden nuggets. You sit back and get the goods.
1.
Jeff Su gives practical wisdom for becoming more “AI Native”
Seriously ya’ll — a must watch, and watch it all. This one was so good it kind of actually hurt. Jeff breaks down some practical, actionable tips to just start getting good at using AI in your life; better. We’re all about it.
2.
Jack Roberts makes us want to try Lindy.AI
This covers a couple of AI workflow automation use-cases and is a good follow-along to familiarize with Lindy — which we’re hearing a lot about right now (and have yet to try). Skip to around 7:21 to see the email responder use-case in action.
3.
Ben actually does show what managing an Agent army is like
Watch the first 10 minutes and you might just bookmark this hour+ long video to dedicate an afternoon to. This is powerful stuff and Ben is using one of our favorite current platforms, Relevance.AI, throughout.
Sound smart
Three shareable things
Stargate
Man, what a name. In case you missed it, Trump, Sam Altman, Masayoshi Son and Larry Ellison just announced a massive project (called Stargate) aimed at building out data centers / AI infrastructure to the tune of $500B over the next four years. The first such data center is already underway in Abilene, TX — a famously boring place where my own wife went to school, just up the road. More here.
Why you should care: While some are casting doubt, this is a sign of the arms race for AI that’s already unfolding. Between companies, and countries. In this case, many power players are coming together on one big initiative.Your New “Digital Twin” Just Dropped
Standford researchers from the Human-Centered AI (HAI) institute have done an extensive study on training LLMs to mirror the personality, behavior and decision-making of real-life people. 1,052 people to be exact. The methodology seems strikingly simple and compelling for use-cases, all at once. Read here.
Why you should care: Imagine a simple 1 hour interview with 30 customers of a brand, becoming an always-usable cohort of AI Agents who think just like customers, on anything. Or, obviously, AI-grandma.Closed-Door Briefings on “Super Agents”
Sam and the team at OpenAI make our newsletter twice this week. There’s lots of hype going around about supposed government/official briefings around soon-to-come PHD-level AI-agent capabilities and their implications in our economy. Generally speaking, the reaction to this rumor is a mix of optimism and skepticism — “just more good marketing from Open AI.” Read here.
Why you should care: Where there’s smoke, fire. Expectations that Operator may launch this week (maybe before you read this) may be related and that could reveal more. Stay tuned.
Share these insights, or share the whole thing.
One request
This is the third ever newsletter for WithAgents. It’s a passion project of mine and a perfect way to focus my digging-into-the-space as we kick off 2025.
Like you (I imagine) I wouldn’t call myself an AI-community member or researcher.
I’m a young dad and career man in Austin Texas. In my day-time, I work alongside high-talent teams to develop and launch services and accelerators in the tech services space. I’ve recently joined a fantastic company called Wizeline to do just that.
Outside of that work, I like to share my ideas and build things that scratch my own itch — enter WithAgents.ai.
WithAgents was born from my own confusion with the Agentic AI hype-cycle, complex definitions and struggles actually adopting them into my own workflow without having to become an AI expert. So I decided to build this newsletter. I’ll learn, and share as I learn. Hopefully doing-so in a way that is digestible to the non-AI-geek.
I hope you’ll follow along and please share in your socials if you find value.
It’ll help this keep moving.
Thank you.