Part 3 of “The Myths and Realities of AI at Work”
Let’s be honest-AI feels almost magical sometimes. It remembers your favorite email phrasing, predicts what you want to say next, and can summarize a 10-page report before your coffee cools. It’s no wonder so many people start to trust it like a super-smart coworker.
But here’s the thing: AI isn’t a person. It doesn’t understand humor, stress, or that weird tension in the room when someone’s about to quit. It’s incredibly powerful at analyzing data, but it doesn’t grasp context.
In this final post of our AI myth series, we’re talking about trust-what AI can do brilliantly, what it fumbles, and why humans still need to stay in the loop.
If you’ve ever asked ChatGPT or Gemini for information and thought, “Wow, that’s exactly what I needed,” you’ve also probably had the opposite experience-where it confidently gives you an answer that’s completely wrong.
AI doesn’t know things. It predicts them. It uses patterns in its data to guess what words probably go together in response to your question. Sometimes those guesses are perfect; other times, they’re pure fiction.
Remember the viral story of an AI-generated photo of the Pope wearing a designer puffer jacket? Totally fake, yet it fooled millions. That’s what happens when AI makes something that sounds or looks plausible but isn’t grounded in truth.
In business, this can mean serious consequences-like sending clients inaccurate info, misquoting research, or even violating privacy laws. AI is a brilliant assistant, but you still need a human fact-checker on duty.
Real-world example: A marketing manager used AI to summarize customer feedback and present findings to leadership. The AI grouped some negative feedback as “positive” because it misread the tone of sarcasm. The result? The team celebrated a “win” that was actually a red flag.
Moral of the story: trust, but verify.
This one’s easy to fall for. AI can sound thoughtful. It can weigh pros and cons, list tradeoffs, and even “recommend” actions. But it doesn’t actually think.
When you ask ChatGPT to “suggest the best pricing strategy for a new bakery,” it doesn’t reason like a strategist in your industry or local. It pulls from patterns in its training data about bakeries, pricing models, and marketing advice and predicts what a smart-sounding answer should look like.
There’s no genuine understanding, just advanced mimicry.
That’s why AI can tell you what to do (e.g., “offer discounts to attract new customers”) but not why it matters to your specific industry or location (e.g., “you’re in a large town where coupons work better than word of mouth”).
AI is great at logic. Humans are great at wisdom. You need both to make real decisions.
It would be nice if AI were born ethical-but it’s not. AI learns from us, and humans are full of biases, blind spots, and conflicting values.
If you train an AI on internet data, it’s going to pick up everything-from brilliant insights to conspiracy theories. Without careful design and oversight, those biases leak into the results.
Example: facial recognition tools have repeatedly been shown to misidentify people of color at higher rates because their training data included fewer diverse faces. That’s not the AI “being racist”-it’s the data reflecting human bias. But the harm is real nonetheless.
Even simple tools can cross ethical lines. Imagine uploading a client’s confidential data into ChatGPT to “summarize notes faster.” That’s risky because not all tools guarantee your data won’t be stored or reused. Always double-check the tool’s privacy settings and terms of use before inputting anything sensitive.
Quick rule of thumb: if you wouldn’t post it on a company Slack channel, don’t put it in an AI chat box.
Here’s where a lot of small businesses get caught off guard. AI tools often store or share data-sometimes even using it to “train” their systems unless you opt out.
For instance, if you use a free AI tool to transcribe a user research interview, a meeting, or draft a proposal, you might unknowingly be feeding proprietary information into the model. Once it’s out there, it’s not coming back.
That’s why larger companies often have strict rules about AI usage and some go as far as creating their own proprietary AI they know that data privacy isn’t just a legal issue-it’s a trust issue.
So what can you do? • Check the settings: Many tools (like Notion AI or Zoom’s transcription feature) allow you to disable data sharing. • Anonymize sensitive info: Use initials or codes when testing AI workflows. • Educate your team: A quick “AI 101” workshop by a qualified person can prevent big mistakes later.
It’s not about avoiding AI-it’s about using it responsibly.
Installing AI isn’t like setting up a printer-you don’t just plug it in and walk away. AI needs ongoing tuning, review, and retraining to stay useful. Think about it: your business evolves. Your products change, your audience shifts, your employees rotate in and out. If your AI tools don’t evolve with you, they’ll start making outdated or irrelevant recommendations.
For example, a café owner might use AI to forecast inventory. Over time, the café adds new menu items and seasonal promotions. If the data set isn’t updated, AI keeps ordering based on last year’s trends-leading to waste or shortages.
Successful AI adoption looks more like gardening than engineering. You plant it, water it, prune it, and adapt it as conditions change.
AI is incredible, but it’s not a mind-reader, moral compass, or crystal ball. It’s a powerful mirror that reflects what we feed it.
If we feed it diverse, clean data and use it with care, it can transform how we work. But if we let it run unchecked-without oversight, ethics, or updates-it can amplify the very problems we’re trying to solve.
Applied anthropologists (like me!) love to remind people that only humans can analyze humans. AI can see patterns in behavior, but it can’t understand why those behaviors happen. That’s where context, culture, and empathy come in-and those can’t be automated.
So as you explore new AI tools for your business-whether it’s ChatGPT, Jasper, Notion AI, or even that chatbot on your website-remember: AI is a partner, not a replacement. It’s at its best when it works with your people, not instead of them.
The Wrap-Up: Making AI Work with You, Not on You
If you’ve made it through this three-part series-first off, congratulations. You’ve officially survived the marketing noise, tech hype, and sci-fi storytelling around AI and arrived somewhere much more useful: reality.
We’ve talked about what AI is (and what it’s not), busted a few myths, and hopefully helped you see that while AI is powerful, it’s not a magic wand. It’s more like a new coworker-brilliant at certain tasks, occasionally clueless, and definitely not ready to run the place unsupervised.
Here’s What We’ve Learned Together In Part 1, we untangled the biggest myths about AI’s capabilities-like the idea that it “understands” your business or that it works the same across every industry. Spoiler: it doesn’t. AI can analyze numbers, spot patterns, and churn out ideas, but it still needs a human brain (and a good dose of common sense) to turn all that data into meaningful action.
In Part 2, we tackled the money myths-the tempting belief that AI instantly saves time and cash. In reality, it takes planning, training, and a bit of patience to make it pay off. Think of it less like flipping a switch and more like onboarding a new team member who’s learning on the job.
And in Part 3, we got real about trust. We looked at how AI can be wrong, biased, or even risky if you’re not careful with your data. We also explored why AI isn’t “ethical” or “intelligent” on its own-it mirrors the data and people that shape it. Which means it’s our job to guide it responsibly.
The Real Takeaway: It’s About Partnership, Not Replacement AI is changing the way we work, but not the reason we work. It can help you write, plan, and automate-but it can’t think creatively, empathize with your customers, or make value-driven decisions. That’s still human territory.
If you’re a small business owner, freelancer, or community leader, the goal isn’t to “get ahead of AI”-it’s to learn how to work alongside it. When used wisely, AI can handle the routine so you can focus on the meaningful. Imagine your AI tools managing your scheduling, summarizing meetings, or organizing data-while you handle the people, ideas, and vision that actually make your business thrive.
That’s the sweet spot.
]]>Part 2 of “The Myths and Realities of AI at Work”
AI feels like a cheat code sometimes, doesn’t it? You can input your data set, type a single prompt aaaannnnd boom! A blog post, a financial summary, or a customer service script appears like magic. It’s easy to think, “Why didn’t I start using this sooner?”
But as with most shiny new tools, there’s a catch. The convenience that makes AI so appealing can also make it easy to overlook the hidden costs, both financial and human. Some costs show up on your credit card bill; others show up in your workflow, your brand voice, or even your decision-making.
So before you hand your business over to an algorithm, let’s unpack what really happens when you let AI take the wheel.
This is the marketing dream AI companies want you to believe: pay $20 a month for ChatGPT Plus, and suddenly you’ll triple your output and cut your labor costs.
Reality check: you might save time, but not necessarily money. Here’s why. Every AI tool comes with setup time, training time, and integration effort. You don’t just plug it in and watch your profits rise-you have to teach it how to work for your business.
For example, imagine a small marketing firm using Jasper to write ad copy. It can crank out ten drafts in minutes, sure-but then the human team still has to edit them for tone, accuracy, and brand consistency. The time saved on writing is often spent polishing the results.
Over time, the ROI shows up, but it’s gradual-like hiring a new employee who needs a few months to learn your business. AI isn’t a cost-cutter on day one; it’s a long-term investment in efficiency.
We’ve all been there: you try a free AI tool and think, “This works great-why would I pay for the upgrade?”
Well, free AI tools are like free apps: they get you in the door, but you’re often the product. Some limit features, store your data for training, or expose sensitive business info if you’re not careful.
Let’s say you use a free AI writing tool to draft proposals for clients. You might not realize that the text you input could be stored on external servers-or even used to train the model itself. That means your private business data isn’t so private anymore.
Paid tools, like ChatGPT Plus, Notion AI, or Microsoft Copilot, often offer enterprise security and privacy options. If you’re handling customer data or intellectual property, those features are worth every penny.
So yes, “free” can help you experiment-but when it comes to real work, think of AI like business insurance: you get what you pay for.
Here’s the tricky part: AI can actually slow you down-especially at first.
Why? Because AI is only as good as your prompts, and learning to prompt well is a skill. You might spend 10 minutes rephrasing your question until the tool gives you something usable. Multiply that by every task, and suddenly the time savings start to evaporate.
Example: you ask ChatGPT to “write a blog post about digital marketing trends.” It gives you 700 words of generic content-technically fine, but not specific to your business. Now you have to rewrite it, check the facts, and make it sound like you. That’s time wasted that could’ve been spent writing it the way you wanted the first time.
Over time, you’ll get better at prompting. For example, learning to say, “Write a friendly e-mail for people who need a reminder to put their laundry in the dryer, include reasons, and keep it under 500 words.”
But early on, AI can feel like a toddler: eager to help, but always needing supervision.
Some tasks are AI’s sweet spot, like drafting outlines, generating captions, or summarizing long documents. But others? Not so much.
AI tools often struggle with context. If you run a landscaping company and ask AI to “write a client proposal,” it might give you something that sounds good but doesn’t reflect your location, services, pricing, materials, etc. You would still need to edit and ensure that it includes everything you want it to…and nothing that you don’t. So while AI can jumpstart a lot of your processes, it can’t eliminate them completely.
This one’s sneaky. Many people assume that since AI is “smart,” it’s plug-and-play. But in practice, the people who get the most out of AI are the ones who invest time in learning how to use it well.
Training your team, or even yourself, on prompt writing, ethics, and data privacy makes a huge difference. For instance, tools like ChatGPT, Claude, or Gemini can all do similar things, but each has different strengths. Knowing when to use which saves frustration and improves your output.
And here’s the part most people miss: your staff needs to feel confident about using AI, not just compliant. Many of the things that AI can automate for you are things that people don’t like to do anyway, like schedule meetings or creating templates from scratch. When you invest in training your employees to understand how AI works, how to ethically use it, and what AI can do for them (rather than to them), adoption becomes a lot smoother.
The Hidden Costs: Time, Oversight, and Trust AI doesn’t just cost money-it costs attention. You have to manage how it’s used, review what it produces, and ensure it aligns with your business values. That’s not a bad thing. In fact, this oversight is what makes AI valuable. But it’s important to recognize that “automation” doesn’t mean “no supervision.”
For example, if you use an AI chatbot like Intercom or Drift to handle customer service, you’ll need someone checking responses, updating scripts, and intervening when the AI gets things wrong. The human-in-the-loop model is what keeps customers haappy, the system reliable, and everything on-brand
AI can absolutely make your life easier-but it’s not the instant miracle people think it is. It’s a partner that needs onboarding, boundaries, and a bit of patience.
The real cost of AI isn’t just dollars-it’s how thoughtfully you integrate it into your workflow. When used strategically, it can give you back hours of your week. When used carelessly, it can eat those hours right back up.
The trick is balance: use AI to speed up the work, not to skip the thinking.
In the next post, we’ll dive into one of the most interesting parts of this conversation: trust. Can we really rely on AI to make decisions, give advice, or represent our businesses? Spoiler: only if we know how to keep it honest.
]]>It’s me again with another blog about AI. Welcome to the first post in a 3-part series. During this short series, we’re going to pull the curtain back and give you a look at the great and powerful Oz. And like the Wizard of Oz, there’s a lot of mystery but not a lot of magic. In this post, we’ll talk about what AI really does, what it definitely doesn’t, and why you (yes, you, the human) are still the most important part of the process.
It seems like AI is everywhere and is the magic pill for getting ahead, doesn’t it? After all, it writes emails, designs graphics, summarizes meetings, and recommends what to binge-watch next (I swear, I started watching Frieren and the next thing I know, half of my recommendations are for anime!). It’s touted as the coworker who never sleeps, never takes vacation, and, if you believe the hype, might soon be taking your job or-even running your business for you.
But here’s the thing: most of what people believe about AI is only half true. Behind the flashy AI startups, corporate tech demos, and buzzwords, AI is still just a set of math equations doing its best impression of human intelligence. It’s useful and worth learning about, but it’s not going to be solving all your problems anytime soon.
Nope. Not even close.
AI doesn’t “understand” anything. Instead, it looks for patterns in whatever data you feed it. Think of it like a very fast, very cocky intern: it can confidently sort information, make predictions, and give you answers that sound like they’re factually true. However, it doesn’t know your customers, your company culture, or that one tricky client who always needs a personal touch.
For example, ChatGPT or Gemini might help you write a sales email or brainstorm social media ideas. That’s great. But if you ask it “explain my customers’ pain points,” it’ll base its answer on the general data it’s been fed, not your actual business in your actual market.
That’s where humans come in. You have the lived experience and emotional intelligence that AI lacks. AI can suggest a marketing idea but only you can decide if it makes sense for your audience or your business. You are the expert when it comes to your business, not AI.
If only that were true! AI isn’t a one-size-fits-all t-shirt. Think of it more like a tailor-made custom suit. How well it fits depends entirely on the measurements taken (or in this case, the data input). And the measurements (data input) must be accurate, constantly updated, and complete. Imagine making that suit with last year’s measurements, which are a bit off anyway and missing the inseam length. Now you understand why you can’t just depend on whatever output comes from your prompts.
The AI that helps a restaurant predict food inventory won’t perform the same way for a law office tracking contracts. Why? Because each industry speaks its own language, has its own standards, and maintains its own best practices. A phrase like “drafting” means something different to a lawyer than to a graphic designer.
Real-world example: Grammarly and Jasper both help people write, but they were made for different purposes. Grammarly focuses on grammar and tone while Jasper focuses on brand style and marketing context. Each AI will give you different answers to the same prompt because they were each trained with industry-specific data, not because it understands the writing or marketing industry.
So before you use an AI tool, ask: Was this built for my kind of work? Does its data set have the dynamic, most up to date information that is crucial for my work yet? If not, while it will definitely give you answers that sound smart and appear knowledgeable, it won’t actually fit your needs. Much like that intern.
This is the big one that I hear all the time. The enthusiastic C-suite executives, the large number of AI start-ups, and AI evangelists on social media endorsing (and forcing) the everyday usage and automation of AI means that a lot of workers are worried about being replaced-with good reason! But as companies who have laid off workers and are now scrambling to hire them back know: it’ll automate some tasks, but it can’t replace humans.
Only humans can read body language in a client meeting, or sense when your team’s burned out. It can’t have an off-the-record conversation that rebtilds trust after a rough week. What it can do is handle routine work: scheduling meetings, organizing receipts, transcribing notes (somewhat accurately), etc.
Picture this: your office uses Microsoft Teams, Otter.ai, or Fireflies.ai to transcribe a meeting, which you ask ChatGPT to summarize and list in bullet points. Of course that seems like it will save time! But only humans can tell when the summary is correct or whether something important has been missed. And when it comes to deciding what actions to take based on those notes, like how to handle a frustrated client or restructure your workflow, that still requires human judgment.
AI is an assistant, not a replacement. You still need people to check for accuracy, tone, and fit. You need them to interpret qualitative data, such as understanding why one team is eager to adopt a new software and another is resistant, be able to understand team dynamics and use that knowledge to facilitate efficient work.
Please, just…don’t. Remember when I said that the dataset it trains on needs to be complete, accurate, and up to date? Let’s talk about “hallucinations.” That’s the term for when AI just… makes stuff up (now would be a great time to read my blog “AI and the Second Law of Robotics). AI is programmed to obey your prompts and give you want-and what you want is an answer. That doesn’t mean it’s going to be a complete, up to date, or even correct answer. AI doesn’t know anything; it predicts the most likely sequence of words that combine to create a convincing answer.
You’ve probably seen examples of AI inventing fake citations that lawyers confidently used in court cases, or a chatbot telling someone their airline refund was approved when it wasn’t. So when you ask it your research questions you will get an answer but it may be only partially correct. Again, it guesses based on patterns contained in the data set it was trained on and it’s highly unlikely to include any recent ground-breaking research studies or all journal articles published. So while some of those patterns it guesses are accurate; others will be total nonsense.
That’s why human review is essential. Use AI to brainstorm or get unstuck, but always check its facts before acting on them or including them in research. AI is programmed to give you answers, they just might not be the correct ones.
Despite the sci-fi, SkyNet vibes the hype is creating, AI isn’t self-aware. It doesn’t think, feel, or have opinions. It doesn’t “know” it’s helping you; it’s just crunching data and predicting what to say next.
If AI were a person, it’d be the friend who’s really good at trivia but has zero emotional intelligence. You can ask it for a list of ideas or stats, but don’t expect it to understand why those things matter to you.
That’s why pairing AI with human creativity works best. Let’s say you run a small home décor shop and ask ChatGPT to “write a post about fall decorating trends.” It’ll give you a solid draft, full of ideas like “cozy neutrals” or “layered textures.” But it doesn’t know your brand tone, your customers, or that your top-selling items are handmade textiles. That’s your part-adding the soul and context that AI can’t generate.
So What Can AI Do Well? Quite a bit actually, when you use it for the right things. • Routine work: summarizing meetings, generating outlines, or drafting emails. • Brainstorming: jump-starting ideas when you’re stuck. • Organization: helping structure data, lists, or workflows you’ve already mapped out. • Customer service: answering common questions with well-written templates.
Think of AI as a coworker who’s terrible at reading the room but lightning-fast at crunching numbers. In other words, you give it the information; it gives you structure and options.
AI is a fantastic tool but that’s all it is: a tool. It doesn’t dynamically update itself, manage a team with empathy and wisdom, and it’s definitely not a magic bullet. The more you understand what AI can (and can’t) do, the better you can use it to make your workday smoother, your business more efficient, and your decisions more informed.
In the next post, we’ll look at why “AI saves money instantly” is another myth-and how the hidden costs of convenience can surprise even the savviest small business owners.
Until then, this is your friendly, neighborhood anthropologist. Stay critical, my friends.
]]>I’m going to say those two words…or two letters. You know the ones I mean. The ones that everyone is sick of hearing about right now.
Did you roll your eyes a bit? I feel you. If you’re anything like me, you’ve heard more than enough about it from LinkedIn bros, social media influencers, news anchors, and vibe coders. Your CEO thinks that they have to incorporate it into all your workflows so the company can stay in competition with everyone else. LinkedIn bros need something new to talk about. News anchors regularly give conflicting information and generally have no idea how it works. Vibe coders understand how it works but don’t care.
And then there’s me. Social media influencer (nah, not really). Why am I bringing up AI when I’m so ambivalent about it? Well, it’s because I’m so ambivalent about it. Plus, I’m a researcher so that means I go down every rabbit hole I can find - especially when someone’s paying me to - so I can translate for those of you who have just enough time to read this, but not enough time to do the research and parse the data yourself. Just your friendly, neighborhood anthropologist here doing my good deed for the day.
Just to let you know, this is only one part of an ongoing series. Bite-size for your brain digesting pleasure. While there are a few different things that the term “artificial intelligence” can refer to, right now let’s focus on the kind that so many organizations are trying to integrate. During the rest of this post, whenever I mention “AI”, that’s the one I’m referring to.
AI is built with basic rules in place that it is supposed to obey. Remember Isaac Asimov’s Three Laws of Robotics from I, Robot? For non-scifi nerds, in I, Robot there are three laws that are inherent to robots’ programming and they cannot deviate from. First, to not harm humans through either action or inaction. Second, to obey all orders given to it by a human except when it conflicts with the First Law. Third, it must protect it’s own existence provided that doesn’t conflict with the prior Laws.
Why is this important for you to know? AI is built, first and foremost, to give you an answer to any question you ask, something like the Second Law. It cannot deviate from this. If you ask AI a question, it must give you an answer. Here’s the catch-
The answer doesn’t have to be correct.
You read that right. It’s job is to give you an answer, not necessarily a correct answer. So, why would AI answer your question incorrectly? Because it’s inherent programming that it cannot deviate from dictates that you receive an answer to your query even when the information necessary isn’t available to the AI you’re using.
Which leads to the next, trickier, question: Why doesn’t AI have all the information at it’s fingertips, so to speak? Google does. Ah, but therein lies the fundamental difference. AI are programs developed by companies to formulate answers to questions based on statistical probability using information that it has been “trained” on.
Google is completely different. Google is a system that indexes webpages that are published on the internet. Basically, it works by pulling up webpages, etc. whose keywords roughly match what you’ve typed in the search bar. Then it’s up to you to decide which sources are legit and truthful, to the best of your ability.
Because AI formulates answers based on data that it has been fed, it cannot give you information that it doesn’t have. It’s widely known that AI organizations are being sued for intellectual property, copyright, and trademark infringement because the programs are being fed art, books, and other things that require purchase in order to use. However, these companies acknowledge that if they had to pay authors, artists, and everyone who’s work they need to use in order for the AI to work they wouldn’t make any money.
And there, folks, is the answer to your question. Why does AI give you wrong answers? Why is it “hallucinating” law cases that don’t exist when asked to support a lawyer’s case? Why does it XXXXX? Because it’s fundamental Law is to give you an answer, no matter what that answer is. Organizations make money from repeat customers and those who develop AI are no different. If you were to receive a response that indicated the program didn’t have the information it needed in order to answer your query then what are the chances that you’d use it repeatedly? None.
Does this mean that they’re useless? No, of course not, and I’ll be diving deeper into that later. But what is does mean is that each of us has to use common sense. We can’t depend on a computer program to give us the answers we need, whether that’s to create a speech, do coding, perform research (oh, please don’t do this!), or anything else completely on its own.
So can AI be reliably used for anything? I’m glad you asked - yes! And I can’t wait to tell you about it…
…next time.
Until then, this is your friendly, neighborhood anthropologist. Stay critical, my friends.
]]>