REAL Talk With Sam Holcman

In a recent CIO article, by Martin De Saulles, “How effective are semantic hubs in moving agentic AI forward?” the authors argue that semantics are now the backbone of enterprise AI, especially as organizations rush to deploy agentic AI systems at scale. They highlight a critical shift: the challenge is no longer just moving and storing data but ensuring that data means the same thing wherever and however it is used.

That is precisely the problem space Enterprise Augmented Intelligence (EAI™) was designed to address. And it is the space EACOE and BACOE have been working in for more than fifty years.

This topic is one that should sound familiar to anyone who has been around enterprise architecture, transformation, or banking, as one example, for more than a few years: everything old is new again. The latest version of the old story is being told through “purchased models,” “off-the-shelf ontologies,” and what some are now calling the semantic operating system for banking, as one example. Buying somebody else’s model of your industry is not a new idea. It is a very old idea in very new packaging. Twenty-five years ago, it was sold as reference architectures, industry models, enterprise blueprints, and packaged best practices. Accenture sold it. IBM sold it. The Big Four sold it. Entire consulting practices were built around the claim that if you adopted a prebuilt model of your enterprise, you could move faster, reduce risk, and leapfrog the painful work of figuring out your own organization. But here is the Real Talk.

There is a lot of noise right now about Artificial Intelligence (AI) strategy, and one phrase you have probably heard is this: “Your AI strategy will fail if you do not fix your data debt.”

Now, there is truth in that. Years of fragmented systems, inconsistent definitions, and patchwork reports really are catching up with us. AI will expose every weakness you have buried in your data.

But if you take that message into the C‑suite or the boardroom, and you lead with “data debt” and “process standardization,” you will lose the audience that matters most.

Today, I want to reframe the issue. Your biggest AI problem is not data debt. Your biggest AI problem is executive relevance.

The real issue: multiple versions of the truth.

I want to talk about legacy thinking.

Not old thinking. Not wrong thinking. But thinking that was right - sometimes brilliantly right - and then calcified into doctrine.

Look up at the sky. Something remarkable just happened in space.

Blue Origin flew New Glenn - their massive, orbital-class rocket - for the third time.
And they flew it on a booster they already landed and refurbished. The same first stage. Flying again.

That is not a space story. That is a legacy story. And yes, there was a glitch on cargo release.

The exact same pattern is playing out in enterprise architecture and business architecture right now. The certification became the thing itself. Organizations stopped asking whether the certification and framework served the mission - and started asking whether the mission was being performed according to the certification and framework. Do not keep flying Atlas V missions in a New Glenn world.

The new paradigm is not coming. It has arrived.

The only question left is whether your organization will be the one catching up - or the one leading the way.

Organizations are learning a costly lesson: AI does not fail first because of the model; it fails because of the data foundation beneath it. More than half of generative AI projects were abandoned after proof of concept by the end of last year, largely because organizations lacked the data readiness required to move from controlled pilots into production environments.

This is precisely why EACOE matters. The Enterprise Architecture Center Of Excellence provides the disciplined, practitioner-based framework required to turn scattered, inconsistent, under-governed enterprise data into an architecture that AI can trust, interpret, and scale against. The EACOE AI Data Modeling Master Class operationalizes that discipline by teaching organizations how to build the semantic, governance, and modeling foundation that production-grade AI now demands. This broadcast is based on the work of Joanne Carew titled How poor data foundations can undermine AI success, CIO – April 17, 2026

Let us be honest - most Enterprise Architects and Business Architects start the day the same way.

You roll out of bed, scroll through overnight emails, open at least six tabs of frameworks you will only partially read, warm up yesterday’s coffee because the meeting starts in five minutes, and think…Maybe today I will finally fix that capability model.

Because let us face it - every good architect knows their day does not really start until they have had that first cup of caffeine.

And speaking of coffee - let me share something interesting that caught my attention.

Despite what your friend who refuses to drink anything, but single‑origin pour‑over might tell you, becoming a real coffee expert is not easy. And oddly enough, that is a problem for Wall Street.

Starbucks may be on every corner, but qualified coffee graders - the folks who certify bean quality for the commodities market - are in serious short supply.

According to the Wall Street Journal, these graders endure a brutal three‑stage competency program: a written exam, a three‑hour coffee grading session, and then a live tasting in front of proctors, identifying defects right down to the bean.
Miss one - and you start over.

Only five to eight percent pass. That is tougher than passing the California bar exam – or any EA or BA exam I have seen.

And it got me thinking: imagine if our industry had that kind of rigor.

In Enterprise Architecture and Business Architecture, too many certifications promise instant expertise.

Take TOGAF®, for example - you memorize a framework, pass a multiple‑choice test, and suddenly you are called “certified.”

Or the Business Architecture Guild’s® CBA® certification - memorize definitions, answer questions, check a box.

It is like calling yourself a coffee grader because you can tell the difference between a cappuccino and a macchiato.

Look, learning theory matters. But real-world skill does not come from picking the right answer on a multiple-choice exam - it comes from building, evaluating, and adapting.

That is why I sometimes say: certifications that only test recall create “credentialed beginners.” You may have sixty percent of the vocabulary, but not the muscle memory.

And that is exactly where EACOE and BACOE take a different approach.

CIOs are under intense pressure to harness AI to drive efficiency, innovation, and competitive advantage - but as a recent CIO article states on reimagining business processes makes clear, the first step is not to “automate faster,” but to rethink how work actually happens before a single model is deployed. In this context, the EACOE Process Visualization approach emerges as a best‑practice methodology for reworking processes in preparation for AI, ensuring that organizations automate optimized workflows, not legacy inefficiencies.

If your AI and analytics investments still feel like disconnected projects instead of an enterprise capability, this episode of Real Talk with Sam Holcman shows how Palantir quietly turned ontologies into a strategic moat - and how the same ontology‑driven principles behind EACOE and BACOE can turn your architecture into an execution engine for real business outcomes.

If you are a CIO, CTO, or Architecture Manager, you are already paying an “architecture tax” you never approved.

It shows up as overlapping platforms, programs that cannot finish, and “strategic” projects that quietly die after burning millions. The surprising culprit: ten everyday EA/BA words that your organization thinks it understands - but does not.

When these words are fuzzy, your architecture is fuzzy. And fuzzy architecture is expensive.

In both EACOE enterprise architecture and BACOE business architecture, ontology is the backbone: it tells us what kinds of things exist in the enterprise, how they relate, and how those meanings stay consistent as we automate, integrate, and apply AI.

Today, that makes ontology not just a theoretical idea, but one of the most valuable, underused skills in the AI job market – and a critical success factor for serious Enterprise Architecture and Business Architecture work.

What an Ontology Really Is: Kinds of Things Together and Their Relationships In information and computer science, an ontology is a formal description of knowledge in a domain – the kinds of things (concepts/classes) and the relationships between them. It is more than a glossary; it is a structured model of meaning that both humans and machines can use.