The City We Built: Bringing the Models Back Together

Opening Scene

Standing on a hill at dusk, overlooking the city introduced back in Article 1, you can finally see the whole layout at once. Near the center, the original zoning office where every blueprint began. Spreading outward: a neighborhood that looks like a family reunion hall, a library with rows of numbered lockers, a tidy pantry-style grocery district, a post office with its delivery routes, a department store with departments radiating from checkout, an old stone bank vault, a restaurant strip with kitchens and a cookbook shop, and a moving company’s loading yard. And at the city’s edge, lit in glowing teal, a brand-new district is under construction, still connected by light-trails back to every older part of town.

In Plain English

Every concept in this series — entities, keys, normalization, cardinality, dimensional models, data vault, OLTP/OLAP, schema flexibility — isn’t a separate, unrelated topic. They’re different rooms in the same building: the discipline of deciding, deliberately, how information should be structured so it can be trusted, found, connected, and used well. AI hasn’t replaced that discipline. It’s made first drafts faster and given the discipline a new, fast-growing district to plan for.

The Old Way

Touring the city as it stood before this new district arrived:

The zoning office (Article 1) is where every model begins — conceptual, then logical, then physical — the discipline of not building before you’ve planned.
The family reunion district (Article 2) is where entities, attributes, and relationships first get named and defined — the raw vocabulary every other district depends on.
The library locker district (Article 3) is where keys uniquely identify and link records — the glue holding separate tables together reliably.
The pantry district (Article 4) is where normalization keeps facts stored once, consistently, avoiding contradiction.
The post office district (Article 5) is where cardinality precisely describes how entities connect, protecting every count and report from silent duplication.
The department store district (Article 6) is where star and snowflake schemas reorganize data for fast, intuitive analysis.
The bank vault district (Article 7) is where data vault modelling preserves full history for auditability and resilience.
The restaurant strip (Article 8) is where OLTP and OLAP are kept deliberately separate — the fast-moving kitchen and the reflective cookbook.
The moving company’s yard (Article 9) is where the choice between schema-on-write and schema-on-read gets made, trading upfront discipline for later flexibility, or vice versa.

Every district exists because someone, at some point, faced a real trade-off and made a deliberate modelling decision in response to it. That’s the whole discipline, viewed from the hill.

What’s Changing (and Why AI Is the Reason)

A genuinely new district is being built — for embeddings and semantic models. Vector embeddings represent meaning as points in space rather than as neatly defined entities and attributes; semantic layers sit on top of traditional models to let AI systems query them in plain language. This district doesn’t replace the older ones — it’s built at the city’s edge, connected to them, because AI models still need clean, well-modelled data flowing in from the rest of the city to be reliable.
AI-assisted tools have sped up every district’s first draft, without removing the need for human judgment. Across this entire series, the same pattern repeated: AI can now propose entities, detect keys, flag normalization issues, catch cardinality mismatches, suggest dimensional structures, and help structure flexible data — faster than ever before. But in every case, a human still had to judge whether the AI’s draft actually reflected the real business, not just a plausible-looking pattern in the data.
Strong fundamentals are what make someone good at directing AI-assisted modelling tools, not obsolete because of them. An AI tool can draft a schema in seconds. Only someone who genuinely understands what a primary key is for, why a fact table’s grain matters, or why a data vault preserves history the way it does, can tell whether that fast draft is actually a good one — or just a fast one.

The Metaphor, Fully Extended

City District	Data Modelling Concept	Article
The zoning office	Conceptual, logical, and physical modelling layers	Article 1
The family reunion hall	Entities, attributes, and relationships	Article 2
The library locker room	Primary, foreign, and candidate keys	Article 3
The pantry district	Normalization and the normal forms	Article 4
The post office	Cardinality and relationship types	Article 5
The department store	Star and snowflake schemas	Article 6
The bank vault	Data vault modelling	Article 7
The restaurant strip	OLTP vs. OLAP modelling	Article 8
The moving company’s yard	Schema-on-write vs. schema-on-read	Article 9
The new glowing district at the city’s edge	Embeddings, semantic layers, and AI-assisted modelling	Article 10

For Beginners: What to Actually Do

Revisit each district’s article whenever you encounter that concept again in real work — this series is meant to be a reference map, not a one-time read.
When you’re handed an unfamiliar database or dataset, try to mentally locate it on this city map: is it mostly transactional (the restaurant kitchen), mostly analytical (the department store or cookbook), or something flexible and evolving (the moving company’s yard)?
Don’t rush to the new AI/embeddings district before you’re comfortable navigating the rest of the city — the new district’s value depends entirely on the quality of what flows into it from everywhere else.
Keep practicing the core skill underneath every district: asking “what trade-off was this design decision actually solving for?”

For Practitioners and Leaders: The Deeper Layer

Treat AI-assisted modelling tools as fast district planners, not city mayors — they can draft quickly, but accountability for whether a model truly serves the business still rests with people who understand the underlying theory.
As semantic layers and embedding-based systems get adopted, audit the traditional modelling foundations feeding them (clean entities, trustworthy keys, well-governed dimensions) just as rigorously as you would any other downstream consumer — garbage flowing into the new district doesn’t become trustworthy just because an AI model sits on top of it.
Use this series’ city framing as an onboarding tool for new analysts or engineers — walking someone through the whole map before they specialize in one district tends to produce more adaptable practitioners than starting narrow.
Resist the temptation to treat “AI-assisted” as a synonym for “no review needed” anywhere in this city — every district in this series showed a version of the same lesson: AI changes how fast a draft appears, never whether someone should check it.

Quick Recap

Every concept in this series is part of one connected discipline: deliberately structuring information so it can be trusted, found, connected, and used well.
Each “district” exists because of a specific, real trade-off someone had to design around.
AI has sped up first drafts across every part of this city, from entity discovery to key detection to schema suggestion.
A new district — embeddings and semantic AI layers — is being built at the city’s edge, but depends on the rest of the city being well-modelled.
Strong fundamentals remain the thing that makes someone good at directing AI tools, not something AI has made unnecessary.

Where This Fits in the Series

This article closes the series by reassembling every previous article’s metaphor into one connected city, and looking ahead to where data modelling theory is headed next as AI reshapes the discipline. If you’re returning to this series later, Article 1’s zoning office is the natural starting point for anyone new to the theory, and this article is the natural one to revisit whenever you need the whole picture at once.

Image Instructions

Image 1 — Header banner (~1600×600px, wide format): A wide aerial illustration of the full city at dusk, viewed from a hill. Older stone and brick districts are spread across the city, lit by ordinary warm lamps — rendered in muted gray/blue tones — recognizable in miniature as a family-reunion hall, a library, a pantry-style grocery, a post office, a department store, a bank vault, a restaurant strip, and a moving company’s loading yard. At the city’s edge, a new district under construction glows softly in electric teal, connected to every older district by faint glowing light-trails. Bitt the Beaver stands on the hill in the foreground, overlooking the whole city, holding its small T-square up toward the new glowing district. Flat vector illustration style, clean lines, minimal in-image text.

Image 2 — Supporting diagram (~1200×800px): Placed directly after “The Metaphor, Fully Extended” table. A simplified, abstract infographic styled as a city map, with nine small labeled district icons arranged in a rough circle (matching the nine prior articles), each connected by thin lines to a tenth icon at the center-edge representing the new AI/embeddings district. The nine outer icons and their connecting lines are muted gray/blue; the central new district icon glows electric teal, with soft light-trails extending outward to each of the nine connected districts. Bitt the Beaver appears small in the corner of the map, holding its T-square, pointing toward the glowing central district. Flat vector illustration, clean lines, minimal text, soft glow reserved only for the teal AI element.

Cross-series visual identity note: In every image across this series, muted gray/blue always represents “the old/traditional way” and electric teal/blue glow always represents “AI/the new layer.” Bitt the Beaver, the series’ recurring mascot, appears in every header banner and most supporting diagrams, with its prop or pose adapted to each article’s metaphor while its core design stays recognizable. Style is flat vector illustration with clean lines, minimal in-image text, and soft glow/gradient reserved exclusively for AI/new elements.