Over a quiet morning coffee, it’s time to think it over and ask some questions about:

• The real value of AI 
• Where it makes sense to use it
• Who can benefit of it (and who can’t)
• Future of learning 
• … and more

So join me for some morning reflections .. 

I. Everyone Has AI – So Nobody Has Competitive Edge 

AI promised to speed up and automate my work – with the idea that I would either have more free time or handle more work. In other words, I would either earn the same (spending less time) or earn significantly more (in the same amount of time).

That sounds nice. And it could be true. But only under the assumption that I am the only one using AI and no one else is. Meaning I have a monopoly on it. And thanks to that, I have an advantage in speed and intensity. But I don’t have a monopoly. Everyone has AI today – or at least everyone who needs it can use it as necessary. At that point, we are all in the same boat. And nobody has an advantage. The only change – we all work overall more intensely. But for the exact same money, because compared to others, we bring no extra advantage (meaning extra value). Isn’t that great?

Basically, AI has joined the ranks of other tools we use for work. Just as necessary and useful as Excel or a calculator used to be. And no, it won’t do the work for us – on its own, without control, guidance, and supervision. Or rather, what it can do on its own, everyone can do, so it brings no value – and nobody will pay us for it.

So what about it? The answer is pretty simple. Nothing. It’s just a tool. Yes, it’s stupid to ignore it and pretend it doesn’t exist. It’s equally stupid to think it will do my job for me. So use it, but without overblown expectations. And push it only where it actually brings clear, measurable results, profit, or improvement. Where it brings nothing provable, it has no place – it would only add unnecessary complexity and higher costs. So – use it with common sense.

—

If AI doesn’t offer a global competitive edge on its own, where does its actual value lie? In small-scale, local control.

—

II. Hey SaaS – We (probably) Don’t Need Your Tools Anymore

On a smaller local scale, AI can actually help. If I have good, sorted data, if I can precisely formulate what I want, and if I can check and verify the outputs, I don’t need to use (and pay for) external tools. Most of the work AI can help with like this is nothing new under the sun – generating reports from data, creating dashboards, various visualizations, database management, building single-purpose apps… So at this point, instead of adapting to paid third-party tools, I hand it to the IT section (or even handle it myself) with a clear request and get a tailor-made tool directly for a specific company and its environment. Which again won’t give me any significant advantage over other companies (because they will do the same thing), but at least it frees me from dependency on external tools and vendors.

On the other hand, this behavior may well contribute to a SaaSpocalypse, because who would pay for something they can have for free, faster, and customized to their needs. Well .. I will not cry.

Local (in-house, home) use of AI has clear advantages – I can easily describe what needs to be solved – and sometimes even how, I usually understand the issue fairly well, and I can evaluate the result. Plus, processing data locally eliminates the risks of data transport and storing it on someone else’s rented infrastructure, alongside giving me independence.

—

Building focused internal tools makes practical sense. But there is a fine line between solving real workflow problems and shoving AI into things that never needed fixing.

—

III. I Don’t Need AI Micromanaging My Life

When I think about the integration and “use” of AI in our world and daily life, I see one common denominator – useless waste. Just because AI exists doesn’t mean I have to push it into everything without thinking whether it makes sense at all or if it actually brings real value (and potential profit). Just like during the rise of digital watches, when it was good practice to stick them onto everything including pens and umbrellas. Today, if a company wants to show it’s ready for the future, it takes its (any) product, forces AI into it, pats itself on the back, makes a viral video, and hopes for success. Then we get masterpieces like an AI toothbrush with cameras or an AI-enhanced bird feeder. Mind you – these are not prototypes or PoC (Proof of Concept), these are real comercial products.

Some things I just want to work simply and “like always,” and there really is nothing to improve. I have no interest in an AI-powered-enhanced-extended flowerpot nagging me to water my plant. I don’t want a talking fridge that demands food scanning just to proudly announce that my yogurt is expired and I should go grocery shopping for the weekend. I’m not an idiot, I don’t need micromanagement – especially from a machine that is only expected to keep cold temperature. Do we really not have more important problems to solve than constantly creating these useless gadgets we proudly call IT/AI/Smart whatever gadgets?

Stone me if you want, but I just don’t think that to find out the expiration date on butter I need an internet connection, smartwatch notifications, a backup reminder in my email, messages on an OLED fridge screen, and an AI-generated task in my online calendar. And if I really need all this, why the hell am I even here and what good am I for?

Just imagine the technology used and the material and energy resources consumed to make this work for such a trivial situation. Shipping across half the planet (multiple times), energy-intensive manufacturing, a device that must be connected to data infrastructure running and updating 24/7. All for something that fundamentally doesn’t need to be solved because it was never a big or important problem.

There is no need to shove AI everywhere.

Even if I can.

Less is more.

—

But in the end all that matters is: Can AI actually do something useful for me?

—

IV. Can AI actually change (and ideally improve) my own work?

I think this is the most important question.

The best way to find out is to play around with it – just a light, simple experiment.

For the sake of objectivity, I will skip both extremes. No programming (where AI has a massive advantage), but also no home electrical wiring or dishwashing (where AI is completely useless).

I will pick a simple operation that I do regularly. Something that is a normal part of any office job and spans across different professions. Just something with broad appeal and universal use.

The operation: a simple search in a file – like a project name, keywords, etc.

Note: for our experiment, the operation doesn’t have to be “searching.” For the test it should be any action or task (or a few of them) we do commonly and repeatedly.

For my “operation” I’ve got at least 5 different ways to do this search. Let’s take a closer look:

1. Standard Ctrl+F in an open document (e.g., Word, Notepad, LibreOffice). It’s simple. I learn one shortcut and then just click through all the found matches.
2. Use bash knowledge and type: grep “what_I_am_looking_for” selected_file.txt. Basically also simple. I just learn to run bash and the right command.
3. Use a local AI environment (e.g., AnythingLLM + local LLM). A bit harder. I have to download the right program, pick a model, upload the document into RAG, and then prompt it. The advantage: it all happens locally on my machine.
4. Upload or copy the file into a cloud LLM (ChatGPT, Gemini…) and ask it to search. I need an account or a subscription, but I don’t have to remember any commands. I just need to formulate what I want and type it into the chat. The disadvantage: I am sending my data and documents “out there.”
5. Run an agent and give it a task via a prompt: “search this document for this.” I have to deal with installing the agent and connecting it to a model (local or cloud). But I get the good feeling of solving it on my own computer with my “personal” agent.

Some of them are with the use of AI some without it. All of them will deliver a result, and all will be basically correct. So, mission accomplished, everyone is happy.

But what was the real price? What does each option “cost” me?

• Learning something new?
• Spending time tweaking and setting up an agent?
• Dependency on an internet connection?
• Lengthy explanations to the AI about what I want?
• Or the need to have a subscription for cloud LLMs?

So, what is best for me?

Which method has potential? What will actually make my work easier? Or will I have to think, plan, and choose digital local and cloud tools and workspaces every single time based on the type of operation? Or should I just hardcode one way, simply because I’m used to it?

In my opinion, the answer isn’t “AI for everything” or master the command line but a sufficiently functional workspace. Meaning, an environment where I can actually get most of “my” work done with ease. Because I can run ChatGPT, alongside agents, do fun stuff in the command line, and still have several programs and apps open, “orchestrating” it all together – from Word to Canva to whatever else.

But from my own experience: if I have to deal with more than 2-3 different environments (agents, AIs, programs, or apps), my focus and ability to coordinate reasonably go out the window. Not to mention that every environment or app requires learning at least the basics of operation and customization. I spend more time switching, dragging information around, and maintaining context between them. In the end, it costs me multiple times more effort, time, and energy.

So for me, it’s about a workspace that allows me to handle the largest share of my work as easily as possible. Or a workspace that I can customize this way myself. And whether that is a command line, purchased software, a custom AI agent, or something else doesn’t really matter. Simply: What works for most of my stuff – and make it better.

Personally, I believe the future is having a single universal workspace that can adapt to my needs and continuously change and transform based on my preferences and projects.

And one more thing: to make this decision, I need to keep track of what’s possible and occasionally test new options and methods on my daily work… even if it’s just so I can throw them away in disgust later. :)

V. Delegating, Wasting, and Forcing

Between today’s first and second coffee, I found a moment for another AI mini-reflection. Today I don’t have any complete thought, rather just fragments and directions of thinking:

1. I get good and usable results where I know my way around the topic myself and ideally have some experience there. If I just delegate all the “work” to AI (without trying to understand), my result simply won’t be very good.. maybe usable, but nothing more.
2. Simply delegating work to AI (without any of my own knowledge and effort to understand) is otherwise personally about as useful for me as driving a car to train for a sprint – yes, I’ll get to the finish line faster, but with no effect on my own physical/bodily development.
3. In many ways, the current extensive use of AI is just a waste (which is unsustainable in the long term) – using advanced technologies, expensive infrastructure, and energy-demanding data centers for absolute nonsense. To put together a schedule, a shopping list, make funny pictures, check the weather forecast or the news, I really shouldn’t need an agent connected to the most powerful LLM with the monthly suscription fee the size of a budget of a small company.
4. Forcing AI integration into companies is stupid. Usually, it means forcing people to use AI “for everything”. This creates a very fragile environment with a high level of dependency. In the longer term, it will hit a wall economically – the budget will be burned through unexpectedly fast (80% of requests will be just fluff, but “since management pushed so hard, I used it to solve everything“) and the company will be left hanging in a vacuum: AI implemented, processes replaced, but they won’t be able to afford this operating model. So – at the cost of further expenses – they will return to the original ways of solving things or – again at the cost of extra expenses – buy the corresponding AI performance.

Note: The most fun part will be when, after the forced pressure to transition to AI, we find out that it actually isn’t that efficient, it’s too expensive for long-term operation, and the reliability isn’t high. And we will want to go back. 

In the meantime, however, we will lose working knowledge, we will lose capable people, and we will get lazy – we’ll get used to relying on and delegating everything to AI.. and the classic crash into a wall will come.

Now it’s time for the second coffee :)