Stop Chasing AI Tools, Build Systems Instead
The rapid evolution of artificial intelligence has created a divide: those leveraging AI for significant gains and those endlessly searching for the next ‘magic bullet’ tool. The key to unlocking AI’s true potential lies not in the tools themselves, but in building robust systems around them, according to a new framework outlining 12 principles for success.
Many individuals find themselves overwhelmed by the constant influx of new AI applications, feeling that their existing knowledge quickly becomes obsolete. This feeling of falling behind often stems from a flawed approach: focusing on individual tools rather than constructing comprehensive systems designed to achieve specific goals. The prevailing advice is a significant mindset shift: instead of shaping your goals around specific tools, you should shape the tools around your system and your objectives.
The Trap of Tool Chasing
The current AI landscape often leads people into a trap where they become fixated on the latest software or application. This ‘shiny object syndrome’ distracts from the fundamental process of identifying a need or goal and then determining how technology can serve it. This is particularly evident with the emergence of powerful new models like OpenAI’s models, which can induce panic and uncertainty about how to integrate them effectively.
The core problem, as highlighted by the 12 principles, is the lack of a structured system. Without a system, new tools don’t enhance productivity; they often add to the confusion and stress. The goal is to build systems once, allowing them to operate efficiently and continuously, much like standard operating procedures in a business.
The 12 Principles of AI Systems
The framework breaks down the process into four phases, each with three principles:
Phase 1: Foundation (Principles 1-3)
This initial phase is about defining the core elements before any building begins.
- Principle 1: Map Out a Clear Goal. Every system needs a destination. Without a clear objective, the system lacks purpose. Identifying your ‘north star’ is crucial, as you cannot hit a target you cannot see. For example, a ‘second brain’ system’s goal might be to quickly capture ideas and receive future reminders, rather than just experimenting with AI for its own sake.
- Principle 2: Define Inputs and Outputs. A system is only as good as the data flowing through it. Understanding what information goes in and what results should come out is non-negotiable. Just as good food (input) leads to feeling better (output) in the human body, a well-defined data flow is essential for an AI system. The output could be a concise daily brief designed to jog memory without overwhelming the user.
- Principle 3: Reduce Decisions Required. Every decision point introduces friction, slowing down the system and potentially leading to abandonment. The easiest way to ensure a system works is to minimize the need for human intervention. This can involve automating decisions through encoded rules or smart routing. For instance, a simple voice dictation button on a phone that automatically captures and files ideas minimizes user decision-making.
Phase 2: Execution (Principles 4-6)
This phase focuses on the practical implementation and mindset shift required for effective AI system use.
- Principle 4: Make Steps Obvious. If a system’s flow cannot be understood within 60 seconds, it’s too complex. Clarity is paramount; confusion kills systems. Simplifying processes relentlessly, often through elimination, is key. Think of traffic lights: red universally means stop, a clear and obvious instruction.
- Principle 5: Automate Plus Escalate. AI excels at automating tasks, handling up to 80% of the workload. However, the remaining 20%, which requires judgment, empathy, or handling edge cases, needs human oversight. This means building escalation paths, not walls, so humans can intervene when AI is unsure. A system can flag inputs with low confidence scores for human review, creating a feedback loop for the AI. Tools like N8N can help manage these workflows and error handling.
- Principle 6: Build Feedback Loops. Data is the lifeblood of AI systems. Tracking metrics, such as model confidence in classifying information, allows for continuous improvement. If an AI’s classification confidence drops below a certain threshold, it should be flagged for human input. This human judgment then feeds back into the system, improving its accuracy over time. Measuring these actions, rather than just the final outcomes, is crucial.
Phase 3: Optimization (Principles 7-9)
This phase focuses on refining the system for compounding results.
- Principle 7: Default to Consistency. Systems thrive on rhythm and predictability. Small, consistent actions performed daily are more effective than large, sporadic efforts. Chaos is the enemy of compounding results. Focus on small, quick actions that make the system operate smoothly, saving time and energy.
- Principle 8: Remove Friction Early. The initial user experience is critical. The first 10% of a system’s flow dictates 90% of its adoption. If the entry point is difficult, users will not engage, regardless of how sophisticated the rest of the system is. Simplifying the start of the process is paramount.
- Principle 9: Measure Actions, Not Outcomes. Outcomes are lagging indicators, while actions are leading indicators. Tracking the behaviors that drive results—like the number of videos filmed for a content system—is more valuable in the early stages than tracking views or revenue, which may take months to materialize. This focus on actions enables iteration and improvement.
Phase 4: Growth (Principles 10-12)
The final phase focuses on scaling and improving the system.
- Principle 10: Improve After Starting. It’s more beneficial to ship a functional system and improve it based on feedback than to endlessly refine it before launch. Continuous iteration and adjustment based on real-world usage are key to long-term success.
A Practical Example: The ‘Second Brain’ System
The principles are illustrated through a practical example of a ‘quick capture second brain’ system. This system, built using a tool called Lerty, aims to capture ideas and thoughts effortlessly. Users can speak or type an idea into the system, which then automatically files it into appropriate categories (ideas, tasks, people, projects) within a database. Agents within Lerty process this data, and the system can generate daily or weekly briefs to remind users of their captured information.
Lerty offers features like customizable agents (akin to personalized ChatGPT or Claude instances), database integration for data storage and referencing, and dashboards for visualization. The system demonstrates how to define clear goals (idea capture and recall), manage inputs/outputs (voice notes to categorized data), and reduce decision friction (a single button press to initiate capture). The automation and escalation principle is shown through confidence scores; if the AI is less than 60% confident in classifying an input, it escalates to a human for review, simultaneously training the model.
Why This Matters
Adopting a systems-thinking approach to AI transforms it from a source of anxiety into a powerful lever for productivity and goal achievement. By focusing on clear objectives, well-defined data flows, and minimizing friction, individuals and businesses can build AI solutions that are not only effective but also sustainable and scalable. This shift moves users from being passive consumers of new tools to active architects of their own AI-powered workflows, ultimately leading to more consistent and significant results.
Source: 90% of People Use AI Wrong (12 Principles of AI Systems) (YouTube)
