Where AI Actually Works — and Where It Fails Spectacularly
Weekly AI reboot: ultimate time to rewind, rethink, and reload—#17
I write about leadership from both sides of the desk. But, I also explore AI and why teaching critical thinking about it is more important than ever.
Please hit the heart ❤️, restack 🔄, subscribe 📨, and all that jazz to help spread the word! 🙌
Now that I've been diving into AI headfirst, I’ve been reading more, testing out tools, absorbing what I can, and figuring out how to use it in everyday situations.
Here’s what I’m noticing: on one side, you've got people yelling that AI is going to be the end of humanity. On the other, there are hype squads acting like AI cheerleaders, trying to convince everyone it can do way more than it actually can.
So just for kicks and giggles, I asked ChatGPT: what are the top most grounded ways to use AI? Not just good ideas in theory, but actual, useful things.
Here's what it came up with:
Writing and Learning Support
This one helps rephrase awkward sentences, brainstorm content, break down complex topics, and get feedback.Coding Support for Developers
These tools are used to save engineers time on routine stuff so they can focus more on design.Language and Access Tools
I know several people using AI to learn new languages. But it can also create real-time captions, generate natural-sounding voiceovers, and support immigrant communities, travelers, and ESL learners.Workflows and Summarization
This runs the gamut from summarizing Zoom calls and long PDFs to turning rough notes into usable outlines or action steps.Internal Search and Knowledge Management
Companies are training private GPTs to search internal policies. It’s the equivalent of a searchable, yet digitized filing cabinet.
Now, I bet there isn’t anything earth-shattering on this list. Most people reading it are probably nodding their heads going, “Yep, nothing new here.” And that’s exactly the point. This list isn’t flashy. It’s not sexy. It’s not screaming that it’s going to change your life and make you six figures.
It’s just showing where AI is already working and where it might make your life a little easier.
Now, for even more fun
I asked it to give me the five epic AI fails. Here’s the list that came back. I also double checked them all and provided appropriate links.
❌ 1. Klarna’s AI Customer Service Debacle (2024)
Klarna claimed its AI assistant was handling two‑thirds of customer service inquiries and improving resolution times. In reality, customers reported circular, inaccurate, or unhelpful responses, often requiring multiple follow‑ups. The company hasn’t abandoned AI altogether but they have rehired some of the human agents they laid off.
❌ 2. Air Canada Chatbot Gives False Refund Policy (2022)
Air Canada’s chatbot falsely told a customer they could buy a full-priced ticket and retroactively apply for the bereavement fare. However, the company denied the request. He took them to small claims court and won. The court ruled that the chatbot’s misinformation counted as official company communication.
❌ 3. CNET Publishes AI-Generated Financial Articles… Full of Errors (2023)
CNET quietly published dozens of finance articles generated with AI. Internal and external review revealed major factual errors, plagiarized text, and misleading claims. This forced corrections on 41 of 77 articles. The controversy led to credibility loss for CNET and a downgrade of its reliability rating.
❌ 4. Chegg’s AI Pivot Tanks Its Business (2023–2025)
Chegg invested heavily in AI-driven student services, but as free tools like ChatGPT surged, its user base declined sharply. Revenue plummeted, and the stock dropped dramatically. Students and educators questioned how good the tools were. Chegg was forced to rethink its strategy and lay off staff as its AI pivot backfired. The company continues to struggle financially.
❌ 5. Google’s AI Overviews Gone Wrong (2024)
Google introduced AI-generated search result summaries (AI Overviews), but they quickly faced backlash for absurd hallucinations such as suggesting users eat rocks or put glue on pizza to keep cheese in place. Public mockery forced Google to quickly pull it back and improve the safety features before releasing it again.
In the end, there are no real shortcuts
There is a gigantic difference between list one and list two. You're trying to scale something that, in so many ways, needs to be individualized. Clearly, anyone who thinks a chatbot can do customer service has never worked in the actual industry. It’s incredibly nuanced.
As for Air Canada: again, hallucinations. Yes, it was sloppy, but Air Canada was nuts not to just give this man the refund. His grandmother died. It was a $650 refund on the fare. Ridiculous.
I’ve also been saying this for months, and I’ll say it again: if you are not double and triple-checking the output, you are putting yourself at risk. These LLMs hallucinate. That’s just a fact. Will they ever fully fix that? Some people don’t think it’s possible.
Which means it’s never going to be 100% accurate. You can’t just publish sources or data without verifying them. That was the CNET problem. They stopped checking things they used to double-check.
What can you do?
My best suggestion for anyone out there trying to figure out what AI, ChatGPT, and all these other tools actually do is this: find people to read and follow who aren't trying to sell you some AI bridge to nowhere.
It’s not always the most exciting stuff to read, but if it’s honest, that’s what matters. Because there are people who will try to take advantage, and then there are others who never use the tools but still have all kinds of opinions on them.
Find someone you can trust to get advice from. I write about my weird AI and ChatGPT experiments all the time.
A few of the strangest but useful things I've used it for?
My hair is down to the middle of my back, and I haven’t had it professionally cut since the pandemic. I just do it myself. So I asked how to trim off as many split ends as possible without lopping off three or four inches, and I also asked for the best ways to add some layers.
Another time, I used it to deal with pest control. I asked about different things ants eat, how to make a DIY bait, and ended up learning way more about ant behavior than I expected. It was kind of fascinating.
The third one might be the most surprising. I used ChatGPT to diagnose and fix the evaporator fan in my freezer. I figured out the issue, bought the part, and fixed it myself. And I am not mechanically inclined at all. Just ask my brother.
TL;DR: If it sounds too good to be AI true, it probably is
In the end, there are no real shortcuts. You still have to put in the work at some level. Whether it’s double-checking, figuring things out on your own, thinking critically, or making decisions.
You can’t outsource it all.
Curious about Custom GPTs?
Send me a DM and we can talk about what they do and how to use them. Due to high demand, I’m no longer offering free access to the one I built for career exploration, but I’m happy to answer any questions ⤵
Welcome to The Spotlight Corner 📢
Here is a shoutout to my favorite piece of the week by
: I’m Losing All Trust in the AI IndustryI restacked this, and it’s worth shouting out again. It’s a nuanced view of AI, and it’s from someone like myself who genuinely wants to embrace and accept it. But there’s so much misinformation. People are stretching reality, saying it can do things it really can’t.
Then there are people on the other side shouting that we shouldn’t be using it at all, or worse, shaming those who do. Women are using it far less than men. We’ve got to get on the same page because we all need to become AI literate.
You can’t teach it if you don’t understand it. And if we don’t start approaching this realistically, we won’t be able to model it for the next generation.
📌 Still thinking about AI and what to believe 🤔
Tech Toolbox: Tools I’m Loving Right Now 🛠
My favorite tech tool this week: Tab Suspender
I recently cleaned out my computer. My CPU and memory usage were spiking. My system needed some serious resuscitation because it was gasping for oxygen.
One of the fixes was a Chrome extension called Tab Suspender and it can be used on both Mac and PCs. It automatically puts your open tabs to sleep after a set amount of time that you define. It’s fairly customizable, and it’s been extremely helpful because I always have way too many tabs running. Usually more than 20 at once. And it was draining the life out of my system.
The constant tab activity was setting off my antivirus, spiking CPU usage, and creating this endless loop of slowdown. This extension was one of several things I tried, which I shared in my Monday post if you’re interested. But if you’re someone who always has a gazillion tabs open like me, this might be worth checking out.
The best dictation tool I’ve used ⤵
Wispr Flow Referral Link: wisprflow.ai/r/WISPR6911.
(You get a $15 credit once you hit 2,000 words! Trust me when I tell you that will happen quickly once you get hooked on this thing).
In Case You Missed It! 🔙
👉 My posts 📝from last week:
➠ I Didn’t Replace My Slow Computer—I Did This Instead (How I used AI to fix my computer).
➠ Last Week’s Weekly Reboot: (I talked about some of the milestones I hit recently).
➠ check out my Instagram 😊
And until next week, “Don’t forget to lead with purpose in everything you do.”
© 2025 Bette A. Ludwig: All rights reserved
👉 Don’t Forget to Evaluate Your Leadership Approach with This FREE Assessment
If you're getting value from AI Can Do That? 🔍, I'd truly appreciate it if you could tap the like ❤️ button and share or restack 🔁. Thanks for reading and for your support! 🙌
Thanks for this levelheaded take. The last sentence rules - if it sounds too good to be true, it probably is.
Thank you for writing about your ongoing experiences to engage the broad area of AI with intention and curiosity. I appreciate that there are not shortcuts. In almost everything, shortcuts are costly long-term.