Samuel Holcman: Thank you for tuning in to Real Talk. Be sure to join your host Sam Holcman again for another edition of our program. We'll have more Real Topics of discussion then. How do you sort out the so-called jargon from real-world practices that work? Do the members of your organization find some business or technology advice utterly confusing? Welcome to Real Talk with Sam Holcman. In this program, we set the record straight and in terms that business people and technology people can understand. Now, here is your host, Sam Holcman. Purchase models, off-the-shelf ontologies, and why everything old to be new again. You're listening to Real Talk with Sam Holcman. I'm Sam Holcman. Where we cut through the buzzwords and talk about what actually works in enterprises today. Today's topic is one that should sound familiar to anyone who has been around enterprise architecture, business transformation, or for example, banking for more than a few years. everything old is new again. The latest version of the old story is being told through purchased models, once again, off-the-shelf ontologies as the new phrase, and what some are now calling the semantic operating system for an industry. For example, banking as one example. Now let's be fair before we get a bit provocative. Off-the-shelf ontologies make a technically, technically serious argument. Banks, for example, have spent decades investing in data quality, master data management, governance, and reporting. And that without explicit semantics, clean data remains inert and governance becomes bureaucracy. That's a valid point. There's a number of ontologies that are looked at as reusable semantic models for concepts such as legal entities, instruments, contracts, obligations, and regulations. Yes, there is some real value in formal semantics. There is value in shared definitions. There is value in making domain concepts machine interpretable. In heavy regulated industries, that matters, of course. But here is the real talk. Buying someone else's model of your industry is not a new idea. It's actually a very old idea in very new packaging. 25 years ago, it was sold as reference architectures, industry models, enterprise blueprints, and packaged best practices. These names will sound familiar to you for those that have been selling those. Accenture sold it. IBM sold it. The big four firms sold it. Entire consulting practices were built around the the claim that if you adapted a pre-built model for your enterprise, you could move faster, reduce risk, and leapfrog the painful work, the work of figuring out your own organization. But what actually happened? in most organizations, exactly what always happens. The purchase model looked elegant on paper, expensive in the contract, but once it hit the real enterprise, the hard part begins. Teams had to map, map the company's products into the vendor products, actually increasing complexity. The company's channels to the vendor's channels, the company's definitions of customer, account, case, use, risk. an exception to somebody else's definition. The project that was sold as acceleration became a long exercise in reconciliation. This is the part executives rarely told clearly enough. That integration costs never disappear. That interfacing costs never disappear. It just moves. It is common knowledge that models have to be adapted. to each operating model, that semantic mapping from source systems are required, and that the learning curve and governance burden are non-trivial. That sentence alone should be enough to trigger memories in every executive who had ever funded a reference model initiative. Because this is the recurring cycle, expenditure first, reconciliation second, Sorry, disappointment third. Money is spent acquiring the model and then consulting support around it. Time is spent trying to make the model fit organization reality. Then people discover the model either has to be customized so heavily that its standard value has disappeared or else the enterprise has to bend awkwardly to fit the purchase abstraction. Let me repeat that last part of the sentence. The enterprise has to be bent awkwardly to fit the purchased abstraction. Yes, your enterprise needs to fit the model rather than the model fitting your enterprise. I don't think anyone really thinks that's a good idea. This is why everything old is new again. We are replaying the same movie with better technology. Still doesn't look good in 4K. Now here's where the issue becomes even more important in the age of artificial intelligence. Purchase models mostly focus on one dimension of the enterprise knowledge, what things are. They're strong on nouns. They can define a loan, a counterparty, a legal entity, a contract, an obligation, a reporting threshold. In general, that's useful, but it's not enough. A real enterprise runs on at least four kinds of data. First, structural data. What things are. The nouns, as I mentioned before. Categories, hierarchies, ontologies, taxonomies, all those words that you've heard before. This is where purchase ontologies, if they are strong, are actually strongest. Now here's the difficulty. Second, behavioral data. What things do. How work actually gets done. how customers actually behave, how employees route around systems, how exceptions are handled when the formal process breaks. Third, temporal data, when things change. Products evolve, rules evolve, processes drift, regulations shift, and every enterprise carries this residue of prior decisions, prior systems, and prior definition. Fourth, contextual data. Why things matter. Why a specific product matters strategically. Why one customer segment is treated differently from another. Why management tolerates one risk and not another. And why one metric matters more than the other and that enterprises specific decision making approach. No off the shelf model can ship all four in a box. None. At best, it can give you a standardized starting point. for structural data, one of the four. But the behavior, the time dimension, and the enterprise specific context are where most of the value, risk, and differentiation live. And that leads to one of the biggest misconceptions in this entire market since its beginning. The misused phrase best practices. Purchase models do not provide best practices. They provide published practices at best. That's a very important distinction and it matters. Best practices are the practices that produce superior outcomes in real organizations. They are proprietary to that organization. They are embedded in the organization's operating discipline, management methods, customer treatment strategies, and risk decisions of organizations that actually outperform others. They are not donated in public standard bodies, vendor packages, or industry consensus artifacts. What gets published is what the market can agree to and document. This is not the same as what's produced and does produce the best result. Public models and standards often represent consensus practice, unfortunately lowest common denominator practice, or regulatory reporting practice. And these can be useful. It may even be necessary, but it should never ever be confused with actual way, the best way your enterprise could or should run. Think about a typical recommendation. The winning strategy is to adopt standards and extend deliberately. Exactly, exactly. Adopt where standardization is required. Extend where your operating model, your products, your customers, and your competitive differentiation require it. What should people do? What should leaders do? Use shared ontologies where they are generally useful. And genuinely useful. Regulatory language, external reporting, interoperability, common industry definitions as required. But please, please stop believing that buying a semantic model means you have acquired an understanding of your own enterprise. As we like to say, your own DNA. That understanding still has to be built. It has to be grounded in the reality of your data, your behavior, your history, and your context. It has to be architected deliberately. And if you want AI to deliver real returns, it has to be based on an ontology-driven understanding of your enterprise. Not on the fantasy that a vendor package has already done that thinking for you. Everything old is new again in enterprise architecture because the temptation never goes away. We understand that. Leaders still want acceleration. Vendors still sell abstraction as certainty. And organizations still pay dearly when they mistake a published model for the enterprise truth. This is Real Talk with Sam Holstman. Do not buy someone else's picture of your enterprise and call it transformation. Build the understanding of your own enterprise first, then use standards and ontologies where they help, not where they replace judgment. Thank you for listening. Please reach out to us at www.eace.org and www.bace.org. for a no-fee discussion. It will be our pleasure to speak with you. Once again, thanks for listening.