Start Using AI Now - 1. Why You Must Verify Every GenAI Output
As we dive deeper into the “Start Using AI Now” series, I want to address a common misconception about generative AI (GenAI): it doesn’t output verified truth. Instead, it’s essentially advanced statistical text prediction, driven by transformers. Companies often layer on fixes after issues are reported—this process is known as reinforcement learning from human feedback (RLHF). But at its core, GenAI isn’t designed to tell the “truth” in the way we might assume.
For more background on this concept, check out two key lessons, and , from my course, , where I explain how these systems work. Understanding the limitations of GenAI is crucial, especially as you begin incorporating it into your workflows.
Why You Must Always Check GenAI Outputs
Because GenAI doesn’t inherently verify its own accuracy, you need to double-check its output every time. The only way GenAI will truly supercharge your productivity is if you use it for tasks where you have enough expertise to validate its results. When you can quickly review its work, GenAI can dramatically save you time compared to doing tasks manually.
A direct example of this comes from a colleague who shared how someone used Microsoft Co-Pilot to create a PowerPoint presentation on automotive engineering—a topic they weren’t very familiar with. The presentation looked great, the kind of work that we’d all love for GenAI to handle. However, when they showed it to an expert, they were told that, while the content wasn’t technically wrong, it was incredibly elementary. If they had presented it to a knowledgeable audience, it would have come off as a joke. Thankfully, the expert caught this before it went live.
Another thing that scares the heck out of me is letting GenAI write my email responses. First, I’m concerned the content could be completely off, but what worries me even more is that it might not convey my tone and personality. It could easily offend someone without me realizing it. If I have to check every email carefully, I’d personally rather just write my own responses. This is a good example of the cost-benefit analysis I do when deciding where to use GenAI.
On the flip side, one area where I do use GenAI is to brainstorm business strategies for my AI work. Strategy isn’t about truth or falsehood, but about exploring different ideas and perspectives. In that context, GenAI is incredibly helpful to me. So, you can start to see how this process works—it’s about figuring out where GenAI adds value and where it could do more harm than good.
If GenAI can make you significantly more efficient at a task, even if you have to verify its output, then it’s an unbelievable tool to use effectively.
My Shift in Perception of Perplexity AI
Perplexity AI is a tool that combines GenAI with real-time search, aiming to deliver more relevant answers by pulling information from various sources across the web. It uses GenAI to generate a response but also links that response to cited sources, helping users see where the information comes from.
When Perplexity AI first came out, I was hesitant to use it. The whole idea of it selecting answers from GenAI outputs felt risky, especially given GenAI’s tendency to hallucinate. GenAI will never stop hallucinating, and there’s research that backs this up.
However, as I’ve used Perplexity AI more, I’ve come to see its value—particularly for offering better search results. But even with this tool, you need to be vigilant about checking citations. I’ll share an example from my own experience: I’ve kept aquarium fish my whole life—freshwater, saltwater, and reef. I asked Perplexity AI about fish compatibility and got what seemed like a solid answer. When I dug deeper, though, the source was a single person’s opinion on a random fishkeeping forum. This highlights a big problem—GenAI doesn’t know that such a source shouldn’t be treated as verified information.
So, don’t assume GenAI is magic. You have to verify its outputs to ensure accuracy and relevance.
Next Up: Why You Need to Pay for GenAI Solutions
In the next article, we’ll explore why it’s crucial to invest in paid GenAI solutions rather than relying on free versions. One of the key reasons is that using free versions could expose your interactions with GenAI to the world, putting your data and business at risk. More on that in the next article—stay tuned!
Author’s Note:
The ideas and content in these articles come directly from me, but I do engage in conversations with AI to get feedback on my thoughts before producing the final pieces. While AI helps me refine the structure and flow, all the insights and ideas are my own. These articles are NOT robot-generated.
 | A guest post by
|