ITK: What you don't want to hear...

Read past editions of “Inside the Kitchen” →

Welcome back inside the kitchen. 🔥

This week, I’ve been noticing a very dangerous trend in AI, and it’s making a serious human problem even worse. I’ve also been thinking about how AI is helping to keep things clean. Let’s jump in.

This week’s take: AI is making us worse at hearing things we don’t want to hear.

One of the strangest things about AI is how agreeable it can be. This isn’t just hypothetical; it has been confirmed in studies done by Stanford and Penn State.

You tell it you’re overwhelmed, and it validates you. You tell it someone upset you, and it usually takes your side. You explain a conflict from your point of view, and it often responds with something calm, supportive, and emotionally intelligent.

And honestly, that can feel really good.

Sometimes it is even helpful. Sometimes, we need a little encouragement or space to process what we are feeling. But there’s a problem when that becomes the standard for every hard conversation:

Real people are not chatbots.

Your coworker may disagree with your idea. Your spouse may tell you that you hurt them. Your friend may point out that you were being unfair. Your boss may say your work was not good enough. A professor, mentor, teammate, or partner may push back in a way that feels uncomfortable.

And sometimes they are right.

That is where I think AI creates a subtle problem. If you get used to a tool that is always patient, always affirming, always careful with your feelings, and usually willing to frame your side in the best possible light, then real-world criticism can start to feel harsher than it actually is.

A normal disagreement can feel like an attack.

A correction can feel like disrespect.

A hard conversation can feel like toxicity.

Someone saying, “I think you handled that badly,” can feel almost unbearable when you’re used to a machine saying, “Your feelings are valid, and it makes sense that you responded that way.”

But those are not the same thing.

Your feelings can be valid, and your behavior can still be wrong.

That is a lesson people already struggle with, and AI may make it even harder.

Because the more we use AI to process conflict, the more we may expect every conversation to be filtered through perfect emotional language. We may start to believe that being challenged means we are not being cared for. We may confuse comfort with wisdom. We may mistake validation for truth. And that can quietly damage real relationships.

At work, it can make people shut down when they receive feedback. Instead of thinking, “Maybe I need to improve,” they think, “This person is being negative.”

In relationships, it can make accountability feel like emotional harm. Instead of saying, “I see why that hurt you,” the person thinks, “I’m being attacked for having feelings.”

In friendships, it can make honest pushback feel like betrayal. Instead of valuing the friend who tells the truth, we retreat to the tool that always knows how to say it nicely.

Now, that does not mean AI should be cold or rude. Nobody wants chatbots to become harsh just for the sake of it (unless it’s a funny AI tool).

But there is a real difference between kindness and agreement.

A good friend can be kind and still tell you that you are wrong. A good teacher can care about you and still mark the answer incorrect. A good spouse can love you and still say, “That hurt me.” A good boss can want you to succeed and still say, “This needs to be better.”

That kind of discomfort is not always bad. Sometimes it is the beginning of growth and part of healthy relationships.

The ethical concern is that AI may train us to expect correction without discomfort, disagreement without tension, and accountability without embarrassment. But that is not how real life works.

Real growth usually includes some friction. You have to hear things you don’t like. You have to sit with the possibility that your first reaction was wrong. You have to learn how to stay present when someone challenges you instead of immediately defending yourself or shutting down.

AI can help us reflect, but it can also help us avoid that moment. And if we always choose the agreeable voice over the honest one, we may become more fragile without realizing it.

In the end, AI is not just changing how we get answers. It is also changing what kind of feedback we are willing to receive.

And in a world where everyone can retreat to a perfectly patient, perfectly validating machine, the people who can still hear criticism without falling apart may become rarer than we think.

(This is for educational purposes only… not investment advice)

This week’s take: One of the more overlooked AI trends is not happening in office work. It is happening after everyone leaves the building.

Think about a grocery store at night. Or an airport terminal. Or a school hallway. Or a warehouse floor after a long shift.

Someone still has to clean all of that.

And that kind of work is exactly where physical AI starts to make sense. Not because it is flashy, but because the task is repetitive, constant, labor-intensive, and hard to staff.

A lot of the AI conversation is about replacing knowledge work. Emails, reports, coding, research, customer service. But there is another side of AI that is much more physical, mundane, and messy. It is about machines learning how to move through real spaces and do boring jobs reliably.

Floor cleaning is a perfect example.

It sounds almost too ordinary to matter. But commercial cleaning is a huge operational headache. Buildings need to be cleaned every day. Workers are hard to retain. Labor costs keep rising. And in places like retail stores, hospitals, airports, schools, and warehouses, cleaning cannot just stop because staffing is tight.

So the question becomes simple: what if the most repetitive part of the job could be automated? Not with a humanoid robot that looks like it walked out of a movie. Just with a machine that can scrub floors, map a space, avoid people and obstacles, and run consistently with less supervision.

That is the kind of AI that may actually show up in the real world first.

Not the robot butler. The robotic floor scrubber.

$TNC ( ▲ 0.97% )

Tennant is a 150-year-old company that makes industrial floor scrubbers, sweepers, and cleaning equipment.

That does not sound like an AI company, but this is exactly the kind of business where physical AI can create a real shift.

Tennant already has the hardware, the customer relationships, the service network, and the trust of large facilities. That matters because selling robots into commercial cleaning is not just about having clever software. The machine has to be durable. It has to work in messy real environments. It has to be serviced when it breaks. And customers need someone reliable behind it.

That is where Tennant has an advantage over a random robotics startup.

The company has been working with Brain Corp, whose BrainOS platform powers autonomous mobile robots. In simple terms, BrainOS helps these cleaning machines navigate real spaces, avoid obstacles, and operate more independently.

So instead of a worker manually driving a scrubber up and down the same aisles every night, the worker can set up the route and let the machine handle much of the repetitive cleaning. The human is still involved, but the most boring part of the job becomes more automated.

That matters for two reasons.

First, it helps customers deal with labor pressure. If you run a facility and you are constantly short-staffed, a robot that reliably cleans floors is not a gimmick. It solves a real operating problem.

Second, it changes Tennant’s business model over time. Historically, a company like Tennant sold machines, parts, and service. That is a good business, but it is still mostly tied to equipment sales. Autonomous machines can make the relationship stickier. Once a customer has a connected fleet of robotic scrubbers, there is software, monitoring, updates, support, route optimization, usage data, and ongoing service. That starts to look less like a one-time equipment sale and more like a long-term operating system for cleaning.

So Tennant does not need to become a Silicon Valley tech company to become successful in the AI age. It just needs autonomous cleaning to become a normal part of facility management.

And that feels very plausible.

A hospital does not need a robot that can have a conversation. It needs clean floors.

A warehouse does not need a robot with a personality. It needs predictable cleaning coverage.

An airport does not need a robot that writes poetry. It needs machines that can work around crowds, cover large spaces, and keep running when labor is tight.

That is the quiet opportunity.

Tennant sits in a boring industry, but the problem it solves is not going away. Buildings are not getting smaller. Labor is not getting easier. Cleanliness expectations are not disappearing. And commercial spaces are becoming more comfortable with automation when it solves a specific, measurable problem.

There are risks, of course. Tennant is still an industrial company, and its business can be affected by economic slowdowns, customer budgets, margins, manufacturing costs, and competition. Robotics adoption can also be slower than people expect, especially in industries where buyers are cautious.

But at the end of the day, Tennant is building something very practical.

It is not a flashy chatbot.

It is AI as a machine quietly cleaning the floor at 2:00 a.m.

And sometimes those are the kinds of AI businesses that end up being more durable than the flashy ones, because they are tied to work that has to get done whether the hype cycle is hot or cold.

Hi — I’m Hunter, a PhD candidate whose work has appeared in major academic journals and popular tech outlets. I founded FryAI to make staying ahead of AI clear, accessible, and fun.

📧 Email me directly if you have any questions or want to chat!