Google to MIT: Traditional Business Done For. Oh Really?! — Straight Talk

More than 90% of top analytics put AI-Crisis as the most destructive economic transformation in history. "Doom and Gloom," if you ask me. Not without its merits, though. What’s important in all this noise — facts and rules of inference. What does this all really mean?! Where are these sentiments coming from? What numbers say? And for that I am starting a new tag on my blog: "Straight Talk." Let’s have a calm and sober look — shall we.

What’s the Message Range In The News?

In total we now have more than three thousand top publications actually proving the coming disruptive impact of AI Integration. And our analytics engine is pulling and analyzing a steadily increasing volume of these public and private works. I’m not going to bore you with the digest, for that one can subscribe to ASE Analytics and drink yourself to bed. Instead we will use the wealth of this information to highlight the most important messages of today:

Google Researchers Warn of Looming AI-Run Economies — "Without urgent intervention, we’re on the verge of creating a dystopian future run by invisible, autonomous AI economies that will amplify inequality and systemic risk."

Where is this sentiment coming from?: Google DeepMind researchers (Tomašev, Franklin, Leibo, Jacobs, Cunningham, Gabriel, Osindero) introduce "virtual agent economies." They argue we’re on track for a spontaneous, highly permeable agent economy (AI agents transacting with each other and with humans) unless we intentionally design guardrails. They propose design ideas: fair auctions, "mission economies" (coordination toward explicit goals), and socio-technical infrastructure for trust/safety/accountability.

So, Google discovers The Stage 4 of AI Adoption I’d written about years ago. This analysis, albeit naive, unearthed a single valuable and timeless fact in business automation — if it’s profitable — it shall be. The researchers are elaborating on the emerging trends to the Left of the Technology Adoption Curve. Google observes this in the Crypto space. And we have observed it in all market segments already. I will immediately tell you that what’s seen on the far-left isn’t automatically a trend. Moreover, attempting to regulate progress is a "Fool’s Errand."
Hold this thought for now.

MIT report: 95% of generative AI pilots at companies are failing — I’d written more about this here ("MIT Says Your AI Stinks. Here’s Why. And how to fix.) This is the current, far more important and timely concern. Whatever the distant analytics think of the far future — it all starts here, with the root cause for the current stage, Stage 1, failures.

Now that I’d given you the range of prognosis let’s start with our "Straight Talk."

How did I press these Sources and What have I learned!

To understand the current media trends versus the realities in the corporate hallways I needed to get at the authors. And that took a sweet minute. My approach was this:

1. Consider the source: What are the merit and competence levels of top voices.

2. Extrapolate Facts: Compare argument facts against ALL facts.

3. Profile the Argument: Is it even sound or a form of Modus- fallacy.

The key question I had: "Why is everyone so obsessed by Stage 4 and AGI while 95% of American companies failed Stage 1?!" Why are top researchers worried about "systemic risk" from AI economies while ignoring the actual systemic risk — "previously assumed good" enterprises burning billions on AI that can’t even read their own databases properly because nobody knows where the data actually lives.

What I’d discovered through few-too-many Socratic Dialectics with popular people is less than disappointing. Highest paid researchers are viewing things from the perspective of an "Ivory Tower" haven’t got a clue of the "Boots on the Ground." Top voices are slinging highly opinionated takes on the topics I could actually prove they don’t understand. And we, the plebs of all kinds, readily eating this up with complete disregard to the practical value of the content.

The current problem is "over-specialization."

We’re looking to AI specialist to solve our AI problems. And very few of our CTOs actually understand that they need to talk to a seasoned Systems Engineer! Systems are made of "infrastructure," "firmware," "system software," "business software," and most importantly "people!" It is a systems engineer who designed, redesigned, implemented, reimplemented, aligned and segregated all of these aspects through and through — with a single master goal: "make your business be," whatever the tech.

Did you know that the first discovery OpenAI made is:
"we need better systems engineering?!"
Having solved that — they’ve got their moat.
Even Google noticed that — people who Trunk with Borg.

Instead, many CTOs today will just read Fortune MIT article and walk away with …​ nothing to act on.

The Current mindset of CTOs — The CTO Avalanche

So popular media is of little use for actual AI Adoption problems. Thus, now back to our "Peanuts." What’s the problem that’s causing this single greatest "CTO Avalanche" of all time?!

I have started a sales funnel and talked to many capable officers living this hurdle today. Folks are in denial! And there are many flavors to this denial, almost as many as the officers that I meet.

The psychological trap: CTOs know their architecture is chaos but can’t admit it publicly, so they keep buying "AI solutions" that compound the problem. Each failed pilot makes the next one harder because now you have more wrapper layers. But the alternative requires courage to step back and pull the carpet off of that steaming heap of $#1T-and-sticks they call "Domain Model." If they even know what that is. For the most part, they’ll just say "dirty data."

Default decision: Just wait for that magical AI tool that will make everything alright.

And the conventional wisdom is that the tool is indeed coming. Question is — when is it coming? Such tools only arrive at the "commodification phase" — when everyone else has already moved on. The confusing part here comes from the era of Digital Transformation, when the "Early Follower" model actually worked. And that approach worked precisely because Cloud is fully decoupled from the company’s business model. It came in at best as a cross-cut concern.

Now the situation is drastically different. During the Digital Transformation CTOs were expected to Distribute their business models producing "Context Boundaries" — precisely what did NOT happen for 90% of the American companies. The idea of "Clean Domain Models" as a future dependency for AI (or any progressive evolution) fell on deaf ears, so most just lifted-and-shifted monolith into the cloud. And that dependency is killing all AI initiatives today. Yet few executives I interview can actually make the connection by themselves.

We made our beds. We didn’t think we’ll need to sleep in them so soon — did we now?

The Real Problem Underneath

Let’s dig a little deeper — shall we?

For a couple of years now I harp on "Domain Architecture" with my customers. And sure, fixing the boundaries unclogs everything and initiatives take off. Only now I realized that the Boundary Problem is yet another Effect and not a Cause.

The actual root cause for many is: "I knowingly cut that corner."
Maybe, I didn’t see the value in that hefty effort.

So, what do you do with that?!

My methods are simple.
Admit it and move on.
I fix my own mistakes.
Whatever my original reasoning was.

The Real Cost

Pondering over all the possible risks and costs I come up with one that is terminal — time.

Recently I shared an overwhelming success by a long-ago customer of ours — a Japanese hydraulics manufacturer. These people move through stages of AI adoption like hot knife through butter. In the linked articles I explained exactly how that happened and the unexpected rewards unlocked. A single decision made all of it possible — distributing the business, albeit for reasons other than the AI evolution. During my media research quest I asked these people what do they read about AI. They read no hype or any populist/futuristic AI stuff.

Why would they — they’re living it.

But the real realization of "value" hit me when an American precision hydraulics manufacturer reached out to me. It was not a typical two-hour assessment call but a much longer story. This business is highly biased, structured, top-down, and rigid. Everything is a one big brick and it’s all tied together by 3rd-party "platforms."

I didn’t take this customer unsure what can I actually do for them. The only thought I could not shake — "Origami will buy them soon enough." Because the future under AI is anything but rigid.

Here’s the brutal math:

Origami deploys AI improvements weekly. Their competitor needs 6 months for any change. That’s a 24x velocity difference. In 2 years, Origami will have iterated 100+ times while the American company managed 4 pilots.

At some point — and it’s sooner than anyone thinks — the performance gap becomes insurmountable. Origami’s AI-optimized precision tolerances will exceed what the American company can achieve with human processes. Their pricing will be 30% lower with 50% better margins.

The American board will call it "market disruption." Origami will call it Tuesday.

This is what architectural debt actually costs: your company.

The Real Solution

Executives need to learn to see through the noise. And there is a lot of noise this time around. I’ve built systems through many disruptions: Dot-Com till now. And I don’t remember this much hype and noise over anything emerging before.

Having decades of good and bad systems to my name I’d classify structural problems into the following three categories:

• Difficult: Irreversible impact;

• Moderate: Wrong Problem/Solution;

• Easy: Something that wasn’t done.

Most of the current AI Adoption impediments are in this third category - easy. We did not model our domains, demarcate our boundaries, and taxonomy our data.

So why won’t executives do the "easy" work?

Because "easy" doesn’t mean painless.

Modeling your domain means admitting you don’t know your own business structure. Demarcating boundaries means telling Product Owner Bob his kingdom needs to be split. Taxonomying data means discovering that your "single source of truth" is actually seventeen sources of lies.

The real barrier isn’t technical - it’s ego.

Executives would rather buy another $2M "AI Platform" than spend $200K documenting what they actually have. Because buying things looks like progress. Admitting architectural debt looks like failure.

The board asks: "What’s our AI strategy?" Nobody wants to answer: "First, we need to understand our own systems." That’s career suicide in quarterly-driven America.

So they buy wrappers. They hire consultants who promise magic. They keep compounding the chaos.

Meanwhile, companies like Origami just…​ did the work. No drama. No excuses. They modeled their domains years ago for operational excellence. Now they’re implementing AI features in weeks while their competitors are still arguing about data governance.

What would be the logical step forward?

Stop hiding from the truth about your architecture.

If you gotta fable — then fable.
If you gotta spin — then spin.
  — Whatever your culture requires.
Just get the darn thing sorted.

The Inevitable Outcomes

While everyone seems to agree that AI revolution will be as impactful as each industrial revolution beforehand, few can name the actual impact. This is expected.

But let me be specific about what’s coming.

The divide won’t be "AI-enabled" vs "traditional."
It will be "architecturally coherent" vs "architectural chaos."

Companies with clean domain boundaries will implement AI capabilities in weeks. Companies with wrapper-upon-wrapper chaos will still be "piloting" in 2027.

The coherent companies will do three things the chaos companies can’t:

1. Iterate at AI speed - Deploy, learn, adjust in days not quarters;

2. Compound intelligence - Each AI feature makes the next one smarter;

3. Buy their competitors - Not for market share, but for customer lists.

That American hydraulics manufacturer I mentioned? They don’t know it yet, but they’re already for sale. Not because they chose to be. Because Origami’s AI-powered precision will make their entire business model obsolete.

The board won’t even see it coming. One quarter they’re "exploring AI opportunities." Next quarter they’re 40% less competitive. Quarter after that, private equity is circling.

This isn’t a prediction. It’s already happening.

We’re watching the largest transfer of market power in corporate history. Not from human to AI. From architectural chaos to architectural clarity.

The companies that treated Domain Driven Design as "nice to have" in 2020? They’re about to learn it was actually "exist or exit."

Choose accordingly.

P.S. I’ll tell you a little secret about choosing:
Your training materials say: "Know your customer."
I say: "Know thyself!"

You can’t fix what you won’t acknowledge.