Putting Agentic AI to Work
April 30, 2026 · GSM.elevate() Team · 8 min read
What Even Is Agentic AI?
If you’ve heard the term “agentic AI” thrown around and felt your eyes glaze over, you’re not alone. The term is everywhere right now, and there are a lot of definitions floating around. Let’s land on a simple one.
Agentic AI is AI that can take action on your behalf. Not just answer your question, but actually do something with the answer. Pull together research. Draft a mockup. Read your inbox. Plan your trip. The shift from “AI that talks” to “AI that does” is a big one.
.--.
|o_o | "What can I help with?"
|:_/ |
// \ \
(| | )
/'\_ _/`\
\___)=(___/The good news: you don’t need to be a developer to use it. You just need to know how to ask. Let’s walk through four real use cases.
1. Research That Goes Deep, Fast
Pick any topic you’ve been meaning to learn. The history of American manufacturing. How DNS actually works under the hood. The differences between database engines. The rules of cricket.
___________________
| ___ ___ ___ ___ |
| | | | | ||
| |___|___|___|___||
| | | | | ||
| |___|___|___|___||
|___________________|
research stackClaude can pull from a massive amount of context, summarize it for you, surface the parts that actually matter, and answer follow-up questions. It’s like having a research assistant who’s already read all the books. You just have to ask the right questions.
The trick is iteration. Start broad. Then narrow. Then narrow again. By the third or fourth question, you’re often deeper into the topic than you would have been after an afternoon of Googling.
It doesn’t stop at research, either. Once you’ve gathered the information, ask Claude to turn it into a study guide. Or a stack of flash cards. Or have it quiz you on what you just learned. You’re not stuck reading what someone else thought was important. You get a custom learning experience built around your topic, your level, and the way you actually learn.
2. Mockups and Design With Claude Code
This is where things get fun. Claude Code is the version of Claude that runs in your terminal or code editor. It can read files, write code, and build working mockups of apps and interfaces.
+-------------------+
| [x] [-] [+] |
+-------------------+
| |
| YOUR IDEA |
| GOES HERE |
| |
| [ Click Me ] |
+-------------------+
mockup.htmlWant to see what an idea looks like before committing real engineering time to it? Describe it. Claude builds it. You iterate. In an hour, you can have something clickable that would have taken weeks the old way.
In tonight’s class we put this to work. We had Claude generate three completely different design directions for the same app, all in minutes, and all drastically different from one another. Three real options, ready for review, while we kept moving on the rest of the build. The point isn’t that it replaces a designer. It’s that suddenly you have OPTIONS, and you can explore ten ideas in the time it used to take to explore one.
3. Daily Email and Calendar Summaries
Most of us deal with a steady stream of email and calendar invites, and finding the signal in the noise takes real time. Agentic AI is great at that.
Connect Claude to your email and calendar (with permission, of course), and you can ask for things like:
- “Summarize what I missed yesterday.”
- “What are my three most important meetings this week and what should I prep for?”
- “Are there any emails from this morning that actually need a reply today?”
+----------------------+
| Your Daily Briefing |
| ================== |
| 3 must-reply emails |
| 2 meetings to prep |
| 1 deadline today |
+----------------------+Stop doom-scrolling your inbox. Start your day with a briefing. The AI handles the triage. You make the decisions.
4. Trip Planning Without the Tab Sprawl
Anyone who’s planned a vacation knows the drill. Twenty browser tabs. A growing spreadsheet. A creeping sense of dread.
Agentic AI collapses that into a conversation.
/\
/ \
/ /\ \
/ / \ \
/ /____\ \
/__________\
trip plan“Plan a four-day trip to Asheville for two adults. We like coffee shops, good food, and short hikes. Moderate budget. Skip the touristy stuff.”
Claude drafts an itinerary. Then you iterate. “Swap day two for something with less driving.” “What’s a good rainy-day backup?” “Find a place with vegetarian options near the second hike.”
By the end, you have a plan. You didn’t need twenty tabs.
The PB&J Test: Why Specificity Matters
Here’s the most important thing about working with agentic AI: the more specific you are, the better the output gets. And it’s a lot harder than it sounds.
There’s a classic exercise that’s been used to teach programming for decades. The premise: explain how to make a peanut butter and jelly sandwich. Sounds easy.
Now imagine the person you’re explaining it to has never seen a sandwich, doesn’t know what bread is, and will do exactly what you say with no interpretation.
_________________
/ \
| PB & J |
| ~~~~~~~~~~ |
| on bread |
\_________________/
||||||||||||||“Get the bread.”
Where is the bread?
“In the cabinet.”
Which cabinet? There are six.
“Open the bag of bread.”
How? Where’s the opening?
“Put peanut butter on the bread.”
The whole bag? With my hands?
You see where this is going. By the time you’ve fully described how to make a sandwich, you’ve written a small instruction manual.
AI works the same way. When you ask for “a website for my coffee shop,” you’ll get something generic. When you ask for “a one-page website for a coffee shop in East Tennessee, dark mode, focused on showcasing our roasters and seasonal menu, with a section for booking the back room for events,” you’ll get something close to what you actually wanted.
The good news: you don’t have to be perfectly specific the first time. You just have to keep iterating. Each round, you get more specific. Each round, the output gets closer to what’s in your head.
A Few Important Things to Know
Before you dive in, there are two concepts worth understanding. They aren’t gotchas, but they matter.
Data training and your privacy. When you use an AI service, your inputs (the things you type, files you upload, conversations you have) may be used by the company to train future versions of their AI. This is called “data training.” It’s not inherently bad, but you should know about it.
Free plans often have data training turned on by default and may not give you the option to turn it off. Paid plans (Pro, Team, Enterprise) usually let you opt out. Different providers handle this differently. Before you put anything sensitive into an AI tool, read the privacy and data usage agreement, understand what’s happening with your data, and make an informed decision about which plan fits your needs.
AI can hallucinate. This is the term for when an AI confidently fills in a detail that turns out to be inaccurate. A made-up source. A misremembered date. A misattributed quote. A function name that doesn’t actually exist in the library you’re using. It happens, and the output can read very confidently.
Always verify the important stuff. If Claude tells you something that matters for a decision, check it. If it cites a source, click through and confirm. The tool is genuinely useful, and a quick verification habit makes it even more reliable. Treat it like a sharp, well-read coworker whose work you naturally double-check before shipping.
Where to Go From Here
If any of this got you curious, here’s what to try this week:
- Sign up at claude.ai if you haven’t already. The free tier is enough to get started.
- Pick one of the four use cases above. Just one.
- Try it on something real. A topic you’ve been meaning to research. A trip you’re planning. An idea for a project. Tomorrow’s email triage.
- When the first answer is generic, get more specific. Iterate.
Then come share what you find. You can connect with us on Discord or Facebook. The fastest way to get good with AI is to see how other people are using it.
/\
/ \
/ /\ \
/_/ \_\
/ GSM \
/ .elevate \
/____________\See you Thursday at 6:30 PM at the E.G. Fisher Library.