The Compass And The Road: Rethinking How We Build Products

Instead of clicking through menus, users talk to software and describe what they need. Instead of writing thousands of if-then rules, AI interprets, generates, and executes dynamically. Instead of...

I still remember when I built my first product. It was a classic three-layer app:

• A database where all the data was stored.

• A business logic layer to process the data.

• A user interface so users could interact with the system.

At the time, that’s just how you built software. Everything was structured, deterministic, and predefined. You had to know exactly what the system needed to do, then code every rule, every interaction, every user flow.

Fast forward to today, and this model is breaking apart. AI isn’t just enhancing software—it’s fundamentally rewriting how we build it.

Instead of clicking through menus, users talk to software and describe what they need. Instead of writing thousands of if-then rules, AI interprets, generates, and executes dynamically. Instead of hardcoding UIs, software builds the right interface on the fly.

This isn’t just a new technology trend. It’s a complete paradigm shift in how software is built, used, and designed. And if you’re a product manager, this means you have to rethink everything—from data strategies to business logic to UX.

How We Built Software Before AI

Data Layer → Structured & Deterministic

Traditionally, data was stored in structured databases like Oracle and MySQL, accessed through predefined queries. Later, data lakes and Customer Data Platforms (CDPs) emerged to break silos, but querying remained deterministic—users still had to know what they were looking for.

Business Logic Layer → Hardcoded & Workflow-Based

This layer transformed raw data into actions through predefined rules and workflows.

• In QuickBooks, business logic mapped transactions to a chart of accounts.

• In CRMs like Salesforce, rules linked customers to accounts and invoices.

• Workflow engines extended this logic, allowing automation—but everything remained scripted and predictable.

UX Layer → Component-Based & Static

The user interface evolved from simple web forms to dynamic, component-based UIs.

• Platforms like Shopify introduced pre-built templates and themes to make UX customizable.

• Low-code tools like Webflow and Retool made UI design easier, but the fundamental structure remained static and predefined.

Despite advances in data processing, automation, and UI customization, the old software model always required structured, deterministic execution—until now.

Transition: The AI Paradigm Shift

The core shift AI introduces is how we interact with software.

You don’t just click buttons anymore—you talk to it.

Instead of navigating structured UI screens, users describe what they need in natural language. AI understands intent and dynamically retrieves, generates, or performs actions.

This fundamentally changes how software is:

• Designed: From predefined UI to fluid, on-demand experiences.

• Built: From hardcoded rules to AI-driven adaptability.

• Used: From structured workflows to dynamic, real-time interactions.

This isn’t a small evolution—it’s a complete paradigm shift.

How AI is Reshaping Software

Data Layer → AI-Interpreted & Quality-Focused

• AI doesn’t need structured data lakes—it thrives on unstructured data (text, images, video).

• Big data is no longer a competitive advantage—what matters is curated, high-quality data that AI can process effectively.

Business Logic Layer → AI-Driven & Non-Deterministic

• Instead of scripting every rule, AI interprets intent dynamically.

• Some deterministic API bridges will still exist to provide structure, but the core logic will be adaptive.

• Testing AI-driven logic is different—you can’t just validate fixed workflows, as responses may vary based on context.

UX Layer → AI-Generated & Personalized

• No more static UIs—AI will generate the interface dynamically based on user needs.

• Apps will be AI-first, meaning users interact directly with Large Language Models (LLMs) or agents, not pre-built screens.

• AI will understand and act on data without manual queries, making navigation nearly invisible.

For instance:

• Instead of clicking through dozens of pages in a CRM, users ask AI to retrieve customer insights, and it dynamically presents the data in the best format (charts, summaries, action buttons).

• Instead of manually building reports in spreadsheets, AI analyzes and visualizes the data in real-time. ChatGPT and Claude can already do that!

Implications for Product Managers

1. AI-First Thinking is Required

• AI-first apps will be built around natural language interactions, not fixed UI structures.

• Product managers must shift from structured workflows to intent-driven design.

2. Testing & Validation Will Change

• Traditional unit tests won’t apply—AI-driven logic is contextual and non-deterministic.

• Product managers must focus on confidence thresholds, AI guardrails, and adaptive evaluation.

• In highly regulated industries, uncertainty may not be acceptable; be aware!

3. The Role of UI is Diminishing

• The future is conversational, adaptive, and visual—not rigid navigation flows.

• Product managers must think about how AI interacts with users, not just how users interact with screens.

• The role of UX, Research, Usability is going to “change”

You Will Need The Whole Village

This shift is already underway. AI-first experiences are rapidly integrating into products, from LLM-powered assistants to AI-generated content and decision-making tools.

But making this transition isn’t just a product challenge—it requires the entire company, from leadership to execution, to embrace the shift. Without full alignment, no product manager can drive this change alone.

I have no doubt that product managers are ready. What about their organization? More importantly, will they have the courage to make the tough business decisions needed to reinvent their offering? That’s the real question.