725 billion for AI: a number that’s hard to grasp

725 billion US dollars. That’s how much Alphabet, Amazon, Meta, and Microsoft plan to pour into data centers, chips, and AI infrastructure in 2026 alone. That’s 77 percent more than the year before — and it’s private money, not a government stimulus program.

Against this backdrop, there’s an appeal you hear a lot right now. Leonard Schmedding sums it up in the very first sentence of a recent video: “You cannot invest too much in artificial intelligence” (0:01). Companies should buy their employees unlimited access to AI tools, invest in hardware and training, and stop debating data privacy for months on end.

The question that immediately comes to my mind: does that also apply to a dental practice with eight employees? To an HVAC contractor, a law firm, a painting business?

My answer is a clear yes — with one important caveat. And that caveat is not “be careful.” It’s this: the decisive question is not how much you invest, but whether you can even say what for.

The AI gap among small businesses is real

Run the numbers and the gap turns out to be bigger than the public debate suggests.

KfW, Germany’s state-owned development bank, studied the German small and midsize business sector. The result: 20 percent of these companies use AI. That’s a fivefold increase in six years — and it also means four out of five businesses haven’t even started.

Bitkom, Germany’s digital industry association, arrives at 41 percent in its survey from early 2026. The difference isn’t a contradiction — it’s a question of sampling: Bitkom surveys companies with 20 or more employees. KfW covers the entire sector, micro-businesses included.

And that spells out the real news: the smaller the business, the bigger the gap. Exactly where relief is needed most urgently, the least is happening.

A prominent example shows this isn’t just about small operations: in spring 2026, Siemens announced it would deploy the lion’s share of a billion-euro budget for industrial AI in the US and China first — citing regulatory hurdles in Europe. If a corporation with its own legal department decides that way, you can imagine how it feels for a business without one.

When adopting AI, budget is rarely the bottleneck

And here is where my experience parts ways with the advice in the video.

The suggestion goes: buy your people unlimited access to the leading AI models. For a software company, that’s probably right. But in the businesses I talk to, it doesn’t solve the problem — because the problem lies elsewhere.

To be fair: Schmedding anticipates exactly this objection. He knows the counterargument that it’s not about how much you invest but in what — and considers it redundant (7:18). His reasoning is linguistically elegant: investing isn’t just spending. Purposefulness is baked into the word itself; you don’t need to spell it out. Accordingly, he also says “everything with sense and reason” (9:22), and not every business needs its own high-performance computers.

I agree with the definition. What I dispute is the assumption that it goes without saying in practice.

Because purposefulness is exactly what’s missing in most small businesses. Almost everyone now uses ChatGPT: drafting copy, polishing emails, sorting out an idea. That’s useful, and many already count it as “we work with AI.” What never comes into view: that AI in a business means something entirely different from opening a chat window.

A bigger subscription doesn’t change that. As long as nobody in the company can name which specific process should be automated, it stays at writing copy — just with more budget behind it.

Then there’s a second trap I see all the time: as soon as businesses get started, their attention goes first to what’s visible and entertaining. A video for social media, a new logo, pretty pictures. It’s fun and it’s showable. But it’s almost never where AI moves the most money in a business.

Where AI actually moves money in a business

The value almost always sits in processes that have three properties at once:

They repeat — daily or weekly, in a nearly identical form.
They are easy to describe — you could explain to a new temp in ten minutes how it’s done.
They are currently blocking a person who has more important things to do.

Run these three filters over your daily operations and you’ll land in the same places in almost any business:

Where it typically jams	What’s actually being lost
Calls that go unanswered during work	Jobs and appointments nobody ever counts
Scheduling and rescheduling appointments	Working hours of your best-paid front-desk staff
Reminders for due appointments and follow-up care	The most predictable part of your revenue
Inquiries outside business hours	Prospects who call the next business instead
Inquiries waiting days for a quote	Jobs your competitor has won in the meantime
Quotes nobody follows up on	Deals that were within reach
Materials and job scheduling in production	Downtime, rush surcharges, and dead capital in the warehouse
Documentation and transferring data	Evenings lost to paperwork

The difference from playing around is measurable. A business that knows every third call goes unanswered on Mondays between 8 and 10 am can calculate what that costs. And only once it has done that math can the question “is the investment worth it?” even be answered.

And that’s my actual point: you can’t take “sense and reason” for granted in a small business — you have to spell it out. Whoever skips the question of “what for” will most likely put their money into the visible instead of the effective, and walk away convinced that AI doesn’t deliver.

Data privacy and the AI Act: what applies to AI assistants since August 2

The second brake is real. Bitkom names data privacy requirements as the biggest digitalization hurdle of all, at 77 percent; legal uncertainty affects nearly half of companies.

The video essentially treats this as an excuse, noting that the AI Act is being softened anyway. That’s exactly half right — and the other half affects small businesses in particular.

What was softened: With the so-called Digital Omnibus, in force since late July 2026, the obligations for high-risk AI were postponed to December 2027. That’s the most burdensome part of the rulebook.

What was not postponed: The transparency obligations under Article 50 have applied since August 2, 2026. They cover AI systems that interact with people — chatbots, voice assistants, and AI phone assistants. Under EU rules, that’s exactly what a small business deploys first.

Before you read this as a new reason to hesitate: the obligation is much smaller than the term sounds. At its core, it requires that people can tell they’re talking to an AI. For a chat window, that’s one sentence at the start of the dialog covering two things — that an AI is answering, and how to reach a human. For a phone assistant, it’s half a sentence in the greeting.

That’s not a project. That’s a line of text.

What remains and genuinely demands care is handling the data itself — for medical practices and law firms, professional confidentiality under Section 203 of the German Criminal Code; for every business, the question of which data leaves the building. These questions need answering. But they are answerable, and they’re no reason to wait three years.

A note on my own behalf: I fact-checked the numbers from the video before using them here. The 725 billion is exactly right. For two other figures cited about data privacy concerns among small businesses, I couldn’t find a reliable source — so they’re not in this text. On a topic moving this fast, going back to the primary source is always worth it.

Why the AI experience advantage is being built right now

If there’s one point where I agree with the video without reservation, it’s the urgency — just with a different rationale.

What’s being created right now is not primarily a cost advantage. It’s experience. A business that starts today will spend the next twelve months learning how these tools work in its operation: what they can do, where they fail, how customers react, how to fit them into existing processes. You can’t buy that knowledge later, and you can’t catch up on it. You can only earn it by doing.

Whoever waits until everything is settled isn’t waiting for certainty. They’re waiting for others to build the lead.

Where I don’t follow the video: AI doesn’t replace staff

The video ends with the advice to replace employees made redundant by AI — and if necessary lay them off — to free up money for AI.

For a business with eight people, that math doesn’t work. There, staff isn’t a cost line you swap for technology — it is the capacity. And the skilled-labor shortage, which Bitkom names as the second-biggest hurdle at 70 percent, ensures nobody gives up people voluntarily.

The realistic effect is different: the front desk spends less time on the phone and more time on the people standing in front of it. The skilled worker no longer does paperwork in the evening. That’s less spectacular than “employee replaced” — but it’s what actually happens.

Common mistakes with AI investments

Starting with what’s visible. Social media videos and image generators feel like progress. The value almost always sits in the invisible, repetitive processes.
Mistaking the chat window for a strategy. Having copy drafted is a good entry point, but it’s not AI in your operations.
Wanting everything at once. One process that demonstrably takes pressure off is worth more than five half-finished ones.
Starting without a baseline. If you don’t know how many calls you’re losing today, you can’t tell afterwards whether anything changed.
Inflating the legal framework into a project. The transparency obligation in effect since August is one sentence in the greeting. The real thinking belongs to the data, not to the labeling.
Waiting until it’s mature. The lead being built right now is experience — and experience only comes from doing.

My take

Yes, small and midsize businesses should invest more in AI too. The appeal is right, and the numbers on how far behind German small businesses are speak clearly enough.

But the size of the check is the wrong dial. A business that puts €500 purposefully into the one process that blocks it every day is better off than one that puts ten times that into something that looks good. Investing is not spending — I agree with that distinction in the video without reservation. It just has to be carried through to the concrete question: where exactly in my business?

My assessment: I believe the main thing missing is a concrete idea of how AI actually cuts costs or grows revenue. Most people use ChatGPT but have no idea what AI could do for their company — if that were clearer, they’d surely invest right away. GDPR concerns are of course a big factor too. Right now, the advantage goes to those who dare: they’re building an invaluable head start in experience, and they’ll come through the AI revolution — and even profit from it — while the worriers smother everything before it starts. And yes, companies should invest more — but purposefully. Playing around is often more tempting than truly looking at where the biggest value lies.

If you don’t know where to start: take a week and write down every process that repeats and holds a person up. That list is your investment plan. Everything else is accessories.