Stop Guessing: Use the Three Tiers of TPS to Improve LLM Adherence and Predictable Output

This content originally appeared on DEV Community and was authored by TeeZe

Have you ever tried to get a complex, multi-step output from an AI only to have it ignore your constraints?

The problem isn’t the model’s intelligence—it’s the ambiguity of human language. Without a shared, structured vocabulary, your conversations with an LLM will always be unpredictable.

We developed the TeeZe Prompting System (TPS) to solve this. TPS creates a simple, shared lexicon between you and the AI, dramatically improving reliability by translating your intent into a language the AI is optimized to understand.

The Core Solution: The Three Tiers of Communication

TPS introduces a methodology where you match your communication style to the task’s complexity using three Tiers:

• Tier 1: The Simple Request: For quick, single-shot tasks (e.g., “Translate this phrase”).

  • Structure: Instruction: [What to do]

• Tier 2: The Blueprint: For complex, multi-step projects (e.g., “Create a project plan”).

  • Structure: A formal framework including Task Definition, Process Steps, and Constraints.

• Tier 3: The Brainstorming Session: For open-ended, iterative, collaborative exploration (e.g., “Poke holes in this business idea”).

Example: Using the Tier 2 Blueprint

To get truly predictable, structured output, the Tier 2 Blueprint is essential. It forces the AI to adopt a role and follow a defined process:

• Task Definition: Create a 3-day workout plan for a beginner.
• Role Assignment: Act as a certified personal trainer.
• Process Steps:
  1. Define the goal for each of the 3 days.
  2. For each day, list 5 exercises with a brief description.
  3. Add a “Warm-Up” and “Cool-Down” section.
• Output Format: A simple, easy-to-read list for each day.
• Constraints: Do not include exercises that require specialized gym equipment.

The Built-in Ethical Guardrails

Beyond clarity, TPS incorporates a foundational Ethical Monitoring Function (EMF) into its Master System Prompt. This “Moral Compass” is a set of seven core principles that bind the AI, ensuring your interactions remain helpful, harmless, and fair.

If you’re looking for a structured way to professionalize your LLM interactions, start with the TPS Toolkit today.

Get the full templates, methodology, and Master System Prompt on GitHub:
TeeZe Prompting System (TPS) Repository

Methodology License: CC BY 4.0 | Toolkit License: Apache 2.0

This content originally appeared on DEV Community and was authored by TeeZe