Turn one weak or inconsistent AI prompt into a clearer, testable instruction system.

Bring one existing prompt, system instruction set or assistant instruction framework. We review what is causing weak, inconsistent or overly generic outputs, rewrite the structure, improve the rules and give you a practical version to test and reuse.

Book for €20 See exactly what is included

Up to 40 minutesHuman-led reviewOne prompt or instruction system

Direct answer

What does prompt optimization actually change?

Prompt optimization is not about adding more words. The goal is to make the AI's role, context, task, boundaries, decision rules and output format easier to follow. We look for vague instructions, contradictions, missing information, weak fallback behavior and formatting problems, then restructure the instruction so it is easier to test and maintain.

Concrete deliverable

What you receive from the €20 session

This is a focused professional service for one prompt or instruction system, not a generic prompt pack.

Deliverable A · Review & rewrite

Improved prompt or system instruction framework

We take the instruction you already use and turn it into a clearer structure for the task it is meant to perform.

Understand the objectiveClarify what the AI should accomplish, who uses it and what a good output should look like.
Review the current instructionIdentify vague wording, duplicate rules, conflicting requirements, missing context and weak output constraints.
Restructure the promptOrganize role, context, task, constraints, output format and fallback rules into a cleaner instruction hierarchy.
Improve reliability rulesAdd uncertainty handling, clarification behavior, exclusions or escalation rules where they improve the use case.
Create test casesDefine normal, incomplete, ambiguous and difficult inputs that can reveal whether the rewritten instruction behaves better.
HandoverProvide the revised instruction plus concise notes explaining what changed and what to test next.

Typical outputs

Rewritten prompt or system instruction
Clearer role and task definition
Output-format and boundary rules
Fallback or clarification behavior
Small test checklist
Practical next-step notes

Fixed service: €20 · up to 40 minutes

Deliverable B · Example use case

Customer-support assistant instruction cleanup

A typical use is an assistant that answers correctly sometimes but becomes inconsistent when the customer asks vague, multi-part or unsupported questions.

Define supported topicsSeparate what the assistant can answer from what requires a human or additional information.
Set response behaviorDefine tone, answer length, structure, source boundaries and when the assistant should ask a clarification question.
Reduce hallucination riskAdd explicit rules for missing information, uncertain answers and unsupported requests.
Structure the outputStandardize useful sections, bullets, summaries or next actions where appropriate.
Test difficult questionsTry incomplete requests, contradictory instructions and questions outside the approved scope.
Document the final versionLeave the customer with a reusable instruction set and a simple method for future testing.

Good fit for

ChatGPT custom instructions
Claude projects or assistant instructions
Gemini or similar AI assistants
Chatbot system prompts
Agent role instructions
Structured content or research prompts

One focused instruction system per order

Scope note: this service covers one prompt or instruction framework. Complex multi-agent systems, large knowledge bases, API integrations or full workflow implementations may require a larger package. Third-party software and API fees are separate.

What we need from you

Bring the prompt and a few examples.

The best starting point is the exact instruction you are using plus examples of outputs you like and outputs that are going wrong.

Current prompt

Share the prompt, system instruction or configuration text you want reviewed.

Your objective

Explain what the AI is supposed to accomplish and who the output is for.

Bad examples

Provide a few weak, incorrect, inconsistent or overly generic outputs if available.

Good examples

If you already know the style or structure you want, examples make the target clearer.

Platform context

Tell us whether the prompt runs in ChatGPT, Claude, Gemini, a chatbot platform, agent tool or another supported environment.

Constraints

Share any tone, compliance, formatting, source or workflow rules the instruction must respect.

Scope boundaries

A stronger prompt can improve consistency, but it cannot guarantee model behavior.

AI Experts can improve the clarity, structure and testability of an instruction, but third-party AI models can still produce unexpected outputs. We do not promise perfect accuracy, guaranteed business results or complete elimination of model errors. Higher-risk use cases may require additional legal, security or domain-specific review.

Related services

Need more than one prompt?

Move to a larger implementation package when the work includes several prompts, a full workflow, integrations or ongoing testing.

Ready to improve one prompt?

Book the €20 AI Prompt & System Instruction Optimization session.

Bring your current prompt and the result you want. We will work from the real instruction rather than a generic template.

View pricing and book