AI Can Do What? (and Why You Still Need a Human)

A colleague of mine recently needed to figure out what size sign would fit her new office window. The obvious solution: drive 40 minutes round trip, measure the window, figure out proportions, and stop at the printers — basically kissing her morning goodbye. Instead, she uploaded a photo of the building to an AI tool and asked for help. It estimated the window dimensions, recommended a sign size and proportions, and suggested local and online printing options with costs and delivery times. Total time saved: about three hours. Her takeaway: some of the most useful ways to use AI aren’t the obvious ones — sometimes it saves you from fighting traffic getting to another part of town with a tape measure (Verbrugge, 2026).

Stories like that illustrate how fast our information landscape is changing. Here’s the twist — and the reason I’m writing today.
AI is impressive; however, it may be telling you what it thinks you want to hear.

Professional researchers have been putting these tools through their paces, and their verdict is more nuanced than the hype. In her
Online Searcher column, editor Marydee Ojala notes that generative AI tools now market themselves as “answer engines” that jump straight from question to response — fine for simple factual lookups, but prone to missing the nuance of a complex research question (Ojala, 2025).

The mechanics explain why. These tools run on predictive analytics — predicting the most likely next word based on patterns in past data. The technology doesn’t actually “know” anything, and it doesn’t grasp the meaning of what it produces. Even experienced searchers hit walls: Ojala describes needing four rounds of prompting a major chatbot–her final attempt with what she calls a scolding tone–to get a decent result on a straightforward name search that a structured database handles in one query (Ojala, 2025).

The skill isn’t simply using AI; it’s having a toolbox.

Librarians know that search success comes from knowing which tool fits which job — pivoting from AI to targeted web search to specialized databases as needed (Ojala, 2025). VAs know that AI is a good tool for brainstorming, comparing options, and getting unstuck — but it can’t supply a real understanding of you and your particular business. It may help you map out a project; someone still needs to own it and follow through. Verbrugge offers the caution that’s easy to overlook if you depend on AI chatbots: always check the information, because it may be incorrect (Verbrugge, 2026).

In other words: the goal isn’t to use AI for everything: it’s to use the right kind of help for the right kind of work — if you choose to use it at all.

That’s exactly how I work with my entrepreneur clients. I can use AI tools to move fast on some aspects, such as drafting emails, first-pass research, and shortcuts like the window-sign trick. Then I apply human judgment, source-checking, and structured research to everything your business decisions depend on. You get the speed and the accuracy, without becoming a full-time AI fact-checker yourself.

This month, I’m offering a free 30-minute “delegate it” session: bring me your most annoying task and we’ll figure out a better way to handle it. Just hit reply and tell me what’s eating your weekends.

Sources cited in this issue:
Kelly Verbrugge, “I Didn’t Know AI Could Do That,” zenith-va.com.
Marydee Ojala, “Voices of the Searchers: Search Successes and Failures,” Online Searcher, Nov/Dec 2025 (Information Today).

Add consistency to your social media feed: Send me your latest podcast or webinar: I’ll extract 8-12 standalone moments, cut each to under 60 seconds, add captions, music, and branded graphics, and hand you back a month’s worth of Reels, TikToks, and Shorts. Get started today.