AI is MAGIC in procurement. Or is it?

AI is MAGIC. Or is it?

At least, that's what everyone seems to believe right now.

✨ I've been building AI into Hermix for years. Using it daily. Testing it across dozens of real procurement scenarios with real companies. And I can tell you: Arthur C. Clarke had it right. Sufficiently advanced technology really is indistinguishable from magic.

The problem is when people stop there.

Because AI is also not magic. And in public procurement, confusing the two is expensive.

✔️ Here's what I've actually seen work. GPT models reading 300-page tender documents and producing structured summaries in minutes. Machine learning cleaning up years of fragmented procurement data across 250,000+ European authorities. AI Chat answering "which experts are required?" or "what's the minimum turnover?" directly from the document, in seconds. GraphRAG and Deep Reasoning handling the complex multi-hop analysis that standard models get wrong. SearchGPT building buyer profiles that would take a junior analyst two days to assemble manually.

This is real. The numbers back it up. 87% reduction in market research time. 75% reduction in tender analysis time. 60% less time writing proposals. Our users at Capgemini, Accenture, Unisys, Fujitsu, and 50+ other companies work this way every day.

❌ But here's what I've also seen fail.

Relying on AI to generating "original" ideas. AI missing cultural, technical or commercial subtext - that changes the entire meaning of evaluation criteria. Non-supervised AI producing output confident enough to submit and wrong enough to lose a bid. These failures are real too. And they happen more often when teams treat AI as a black box rather than a tool that requires judgment and supervision. Then comes disillusionment.

AI in procurement compresses effort dramatically. It does not replace strategic thinking.

The companies winning consistently in public sector are not the ones using the most AI. They are the ones using AI for the right tasks, at the right stages, with human judgment at the decision points that matter.

Finding the right tenders. Understanding buyers before the RFP drops. Qualifying opportunities with real competitive data. Reducing the manual burden of documentation. These are the jobs AI does well.

Deciding whether to bid. Understanding what the buyer actually wants versus what they wrote. Building a differentiated technical approach. These remain human work.

❓ There is also a question worth sitting with. If AI writes the procurement. Writes the proposal. Evaluates the submission. What exactly are we measuring? I don't have a clean answer. But I think it matters that we're asking it.

I've written a longer piece on this. What works, what doesn't, the specific techniques we use at Hermix (GraphRAG, Deep Reasoning, SearchGPT, citation support), and the adoption barriers nobody talks about honestly.

https://hermix.com/ai-is-magic-or-is-it/

AI is magic?

BUBBLE 1. AI is magic, everybody is replaced by AI, software engineers disappear, people don't need to work anymore, we are all gonna have fun, everything is great.

BUBBLE 2. AI is magic, everybody is replaced by AI, creativity is killed by GPT, software engineers disappear, everybody loses their jobs, the stock market crashes, AI will take over the world, the world is doomed. 

MIDDLE GROUND: AI replaces only people who do repetitive boring tasks, and cannot adapt. AI cannot replace human creativity. AI doesn't replace software architects and engineers. Vibe coding is still a myth. We will benefit from AI to improve our lives, society and economy. 

Negativity and opportunities: again

Most people have a negative attitude.

We see problems and risks, rather than opportunities.

And negativity increases with age.

As we get older, our appetite for risk decreases. We become more conservative. We are afraid of change. We see fewer opportunities, as we have more to lose. Loss aversion bias appears. Potential losses seem more important than gains.

And we tend to see only problems, only the negative side of things. “The world is doomed, nothing works, education collapses, things were better in the past, today’s immoral youth, technology is dangerous, AI will destroy humanity.”

Of course, looking for problems is also a pragmatic, engineering mindset. You need to see problems - to start looking for solutions. In fact, every opportunity starts with a problem. Solutions rarely appear before problems (although it happens, and occasionally even disruptive).

I pushed myself for so long to think in terms of opportunities rather than risks, that it became a reflex.

I am conservative by nature, and I had to train my sense of opportunity.

Now I have an allergic reaction when I see fear of technology, fear of new, risk reports that block opportunities.

AI replaces software programmers and composers, yet?

After three years of AI revolution, I still believe that major technological revolutions create jobs and businesses. They don’t create unemployment. The invention of the plow didn’t put gatherers out of work. The printing press sparked a phenomenal boom in the book industry and jobs: printing houses, publishers, libraries, writers, translators.

AI won’t lead to the disappearance of software and technology companies either. It will multiply them. 

(Rant triggered by very cool discussions with Vlad Coroama, Mihai Amariutei, Alexander Kruel).

As for how much software programming will actually change, I have reservations. 

I first said it 25 years ago: Software developers are cheap, and replaceable. But software is much more than that. We need architects and engineers; not coders. We need people who understand mathematics, algorithms, data, patterns, design, methodology. "Writing code" is a low-level job, and requires low-level qualifications.

This is even more true today - when AI replaces more and more low-level jobs.

But:

DOES AI REPLACE REAL DEVELOPERS?

=> NO.

We write code every day, at Hermix. With AI. 

From our experience, AI isn’t miraculous, yet. Surely it helps. But it doesn’t replace us - it just helps us. It is a fast, limited assistant.

SO WHAT ABOUT ALL THE STORIES ABOUT VIBE CODING?

Every day, we read about a new AI agent writing an entire application from scratch. 

An AI agent wrote a C compiler. Another migrated an Nintendo game to run in the browser. A non-tech guy built an entire web app in 10 minutes. An AI invented a new game (ask for links).

Are these true?

Well, some are true, but they are anecdotic, exaggerated, not replicable.

If you examine them closely, none is a genuine example of truly innovative real-world application, built from scratch without human help.

It is trivial to write a C compiler in 2026. We already have thousands of open-source C compilers. 

It is trivial to rewrite an application from one language to another.

AI-generated code has security / architecture /design problems, needs huge corrections and rewrites to make it viable.

I keep an open mind, I read and follow developments, I test new stuff (never enough). But I haven’t seen a reliable case of a real, original application entirely generated by AI - yet.

That being said:

I STRONGLY BELIEVE IN AI and the AI REVOLUTION.

It is already transformative, it revolutionizes the society and economy.

As of now, AI doesn't operate independently. Doesn't replace teams of developers, nor music composers, painters, translators.

It only helps them.

AI remains an assistant, for those programmers / architects / composers / managers who are smart enough to control an army of AI agents. Essentially the same visionaries who already knew how to coordinate teams of human agents, to achieve original, impressive results.

The reality is that we now have (many) agents that can quickly solve deterministic computational problems. But we’re still talking about armies of junior agents — unoriginal, not very smart, and very fast. 

American checks and balances: the next world order

 The American democracy is a special construction. It is based on a complex set of checks and balances, built into the Constitution from the very beginning, with the precise purpose of preventing a single man or an interest group from seizing power.

This worked very well until now.

During Trump’s first term, it worked surprisingly well. American judges were still independent. Civil servants, police officers, and FBI agents were still independent. Prosecutors opened investigations against their own superiors.

Parliament blocked military or budgetary decisions of the presidency. The army said that the laws do not allow the execution of just any order given by the president.

Now it no longer works. I am shocked by this simple sentence: the American system of checks and balances, which was supposed to protect American democracy, collapsed after Trump’s second election.

The American democratic system has collapsed entirely. There are no more checks and balances.

It is somewhat similar to the rise of Octavian Augustus.

Caesar still pretended to listen to the Senate, but Octavian put an end to those foolish ideas, declared himself a God, and stopped the nonsense about Democracy, the Republic, and the Senate.

This is where we are now.

Trump has put an end to the nonsense about Democracy, the Republic, and the Senate.

I must underline that Trump will not be the first imperator of America. He is too old, too senile. He does not see far enough ahead. He is happy just making a bit more money with trumpcoin or penguicoin.

Not even Trump’s children will be the first American dictators. They are incompetent.

The next imperator of America will be a grandson of Vance, Thiel, or Musk. Someone who will get hold of an artillery division at the right moment, like Napoleon or Octavian, and will declare “the Consulate.” And then will occupy the Senate entirely, to the cheers of the masses.

The American dictatorship will most likely bring global stability in the medium term. The supremacy of America at the global level, as a world empire.

Much like the supremacy of the Roman Empire.

We are heading toward a Pax Americana lasting several hundred years.

I believe this will be a good thing for our children and grandchildren.

Their lives will be characterized by peace, under the umbrella of an American dictatorship and supremacy over the entire world.