PromptStructor

Inspiration

As AI tools become part of everyday work, users frequently submit prompts that are much longer and more repetitive than necessary. These extra tokens increase API costs, processing requirements, energy consumption, and demand for computing hardware.

We were inspired to create a practical tool that helps people communicate with AI more efficiently without requiring them to become prompt-engineering experts. PromptStructor aims to make every prompt cleaner, clearer, and more resource-efficient while preserving what the user originally wanted to achieve.

What it does

PromptStructor transforms long, repetitive prompts into cleaner, leaner, and more token-efficient instructions.

Users can enter their original prompt and receive an optimised version that:

Removes unnecessary and repetitive wording
Preserves the original intent and important constraints
Displays the original and optimised prompts side by side
Shows how many tokens have been saved
Recommends an appropriate AI model for the task
Allows the improved prompt to be copied instantly

The broader vision is for PromptStructor to work not only as a standalone platform but also through an API and browser extension, allowing users to optimise prompts directly inside the AI tools they already use.

How we built it

We began by identifying the main problems caused by inefficient prompts: unnecessary token usage, higher API expenses, increased processing demand, and a poor prompting experience for users who may not know how to structure instructions effectively.

We then designed a simple workflow:

The user submits an unstructured or repetitive prompt.
PromptStructor analyses its intent, requirements, and constraints.
Redundant wording is removed while essential information is retained.
The original and improved versions are displayed for comparison.
Token savings and a suitable model recommendation are presented.
The user can copy the optimised prompt with one click.

We focused heavily on making the interface straightforward and transparent. Rather than silently changing the prompt, PromptStructor allows users to compare both versions and verify that their intended meaning has been preserved.

Challenges we ran into

One of our biggest challenges was reducing prompt length without accidentally removing important context. A shorter prompt is not useful if it changes the user’s intention or produces a worse response.

We also had to consider that different prompts require different optimisation strategies. A coding request, research question, creative-writing prompt, and simple factual question should not all be compressed in the same way.

Another challenge was balancing token efficiency with clarity. The smallest possible prompt is not always the best prompt, so we had to prioritise meaningful optimisation rather than simply deleting as many words as possible.

We also explored how PromptStructor could recommend suitable AI models without overwhelming users with technical information.

Accomplishments that we're proud of

We are proud that we developed a solution that connects prompt engineering with cost efficiency and digital sustainability.

Our prototype presents token optimisation in a way that is easy for users to understand through live comparison and visible token savings. It also preserves user control by allowing people to review the improved prompt before using it.

We are especially proud that PromptStructor is useful for both individual users and organisations. Individuals can obtain better responses and reduce unnecessary usage, while companies using AI APIs could potentially reduce operating costs across thousands or millions of prompts.

What we learned

We learned that prompt optimisation is more complex than simply shortening text. Effective prompts must balance brevity, clarity, context, structure, and intent preservation.

We also learned that users benefit from seeing why an optimisation is valuable. Features such as side-by-side comparison and token-savings calculations make an otherwise invisible improvement measurable and understandable.

The project also helped us understand the wider environmental and economic effects of AI usage. A small reduction in one prompt may appear insignificant, but repeated across a large number of users and requests, the cumulative reduction in processing and resource demand could become meaningful.

What's next for PromptStructor

Our next step is to develop PromptStructor into a fully functional platform and improve the reliability of its intent-preservation system.

We plan to introduce:

A browser extension that optimises prompts directly inside popular AI platforms
An API that other applications and organisations can integrate
User-selectable optimisation levels, such as light, balanced, and maximum compression
Prompt history and analytics showing cumulative tokens and costs saved
More accurate model recommendations based on task complexity
Support for multiple AI platforms and model providers
Team and enterprise dashboards for monitoring organisation-wide AI usage
Estimated financial and environmental savings
Custom optimisation rules for different industries and use cases

Ultimately, we want PromptStructor to become a universal efficiency layer between users and AI models—helping people obtain better results while using fewer tokens, reducing costs, and minimising unnecessary digital resource consumption.