AI Hype — Straight Talk

Figure 1. A story from The Final Industrial Revolution series: Adventures.

 

92% of AI news is fabricated B$. Our aggregator service classifies this daily.

I’m selling premium vegetables at a cheap candy convention here!
The good news: 95% of you learned that sugar is poison.
The bad news: I still have to walk you through the food pyramid before you get the real candy.
Like fine Belgian chocolate — not what you’ve been eating at this sugar convention.

Let’s find the 8% that matters!

Where Business Value Actually Lives

Seven tiers. Most of you are stuck in the first two, convinced you’re transforming.

T0 — Buy AI Tools (Zero competitive advantage)

Microsoft Copilot. Google Gemini. ChatGPT subscriptions for everyone.

This is 77% of your happy AI news. It’s also Microsoft Office 2025. Necessary, but NOT Differentiation.

Vodafone gave 68,000 employees productivity tools. Congratulations - you bought Office.
Wells Fargo automated compliance. Still T0. United Wholesale Mortgage documented "productivity gains."

So did everyone who installed Excel in 1995.

T0+ — Integrate Tools Into Operations (Minimal value)

Support chatbots. Development acceleration with Cursor. Cost reduction bots.

Better trinkets. Still not transformation. Just shinier Office.

T1 — AI at Domain Boundaries (First actual disruption)

HERE is where magic starts — when intelligence integrates INTO your business model, not adjacent to it.

AI sits at Anti-Corruption Layers (ACLs) — the edges of Bounded Contexts where your domains talk to the world. This is architectural work. This requires foundations most of you don’t have or even understand.

Who’s here?

• DoorDash: AWS Bedrock optimizing logistics domains;

• Robinhood: 80% cost reduction through integration;

• GM: IoT sensor integration across 140+ manufacturing sites, 15% downtime reduction;

• JPMorgan Chase: 450+ AI use cases, data mesh architecture with domain boundaries;

• Walmart: Multi-agent AI across supply chain - eliminated 30M driving miles;

• Duke Energy: Grid domain integration, power flow sims from 6 weeks to 15 minutes.

Notice something? These aren’t chatbots. These are architectural increments to live business models. (You need to have them domain models first — right?)

95% of you fail here because your architecture wasn’t designed for this.

T2 - AI-Only Subdomains (Market-changing disruption)

New AI domains sitting BETWEEN your existing business domains. Behavior your business never had before.

Almost nobody’s here yet. Shopify Sidekick is approaching it - production agentic systems spanning domains. Maybe 3-5 others not talking publicly.

If you’re not solid at T1, attempting T2 will destroy you.

T3-T6: Read the 2023 playbook if curious. For traditional enterprises, T1-T2 is where 99% of value lives. You won’t see T3+ for years.

Why 95% Can’t Reach T1

I predicted this July 2023. MIT confirmed it August 2025.

Most enterprise software wasn’t designed - it evolved.

Like a coral reef.

Layers upon layers of legacy systems, acquisitions, "temporary" workarounds that became permanent infrastructure.

— Vadim Kuhay
July 2023

 

You can’t AI-enable chaos.

You are welcome to remember this concept.

T1 requires

• Clean domain boundaries (you have integration spaghetti);

• Unified data architecture (you have 12 systems with different "customer" definitions);

• Architectural foundations (you have 20 years of technical debt).

DoorDash didn’t get lucky. They had foundations before adding intelligence.

Your company is debating ChatGPT licenses while they’re doing actual integration work.

The 8% of News That Matters

Ignore completely:

• Funding announcements (their problems, not yours);

• New model releases (unless you’re T1+ already);

• Vendor "AI features" (T0 wearing a tuxedo);

• Consultant frameworks (they don’t understand YOUR domains);

• "AI success stories" from companies selling AI tools.

Watch carefully:

• Architectural integration discussions (rare as heck);

• Actual T1 implementations with architecture specifics (me 😃);

• Domain modeling approaches for AI (DDD Europe 2025);

• Failed pilots with honest postmortems (even rarer);

• Companies discussing Anti-Corruption Layers and Bounded Contexts
  (pure gold, if you ever find it).

The brutal numbers from our aggregation:

• 11% fully implemented AI (Salesforce research);

• 21% repatriated workloads from cloud (Flexera 2025);

• 9.7% enterprise adoption — barely moved from 3.7% in 2023;

• 95% pilots failing (MIT).

You’re NOT alone in the struggle.
You’re alone in admitting it.

What to Actually Do

If you’re stuck in T0:

Stop buying tools. Your problem isn’t AI capability — it’s architectural debt blocking integration. Fix foundations first or keep wasting money on chatbots. You don’t need to fix everything. Just move incrementally always showing value in production.

If you’re attempting T1:

Ask yourself honestly

• Can your systems support domain-boundary integration?;

• Do you have clean Bounded Contexts?;

• Documented boundary conditions in ACLs?;

• Unified data across domains?

No?

Then you’re not attempting T1. You’re attempting T0+ and calling it transformation.

If you’re planning T2:

You’re 2-3 years away minimum. Get T1 working first. I’m serious — attempting T2 without solid T1 will cascade-fail across your entire domain model.

You can’t build the third floor without the second.

One Question That Cuts Through Everything

Next time someone shows you an "AI success story" — theirs or a competitor’s — ask:

"What tier are they actually at?"

That tells you everything about whether it’s relevant to your situation.

Vodafone’s 68K Copilot deployment?
T0. Ignore it.

DoorDash’s AWS Bedrock integration?
T1. Study it!

Your vendor’s "AI-powered" feature?
Probably T0 in a nice box.

They can’t answer?
Talk about sports.

The Vegetables You Came For

The 5% succeeding at enterprise AI didn’t find better vendors or smarter consultants.

They built better foundations BEFORE adding intelligence.

That’s the vegetable. It tastes better than sugar once you try it.

Most of you won’t. You’ll keep eating cotton candy, wondering why transformation never comes.

The ones who do? You already know who you are. You felt the guilt in that opening paragraph.

Quick recap

• T0/T0+ — productivity tools, zero competitive advantage;

• T1 — domain boundary integration, first disruption;

• T2 — AI-only subdomains, market-changing;

• T3+ — read the 2023 playbook, not relevant yet.

Want to know what tier YOUR architecture can actually support?

Stop lying to yourself and find out.

P.S. I’m not writing carrot theory. I fix corporate architecture — decades, hands on, every day.