# Benefit of tool vs. specific prompt instructions

**URL:** <https://community.openai.com/t/benefit-of-tool-vs-specific-prompt-instructions/583421>\
**Category:** API\
**Created:** [January 11, 2024, 11:22pm UTC](https://community.openai.com/t/benefit-of-tool-vs-specific-prompt-instructions/583421 "2024-01-11T23:22:00Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![nicktucker4](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/nicktucker4/32/147045_2.png) [@nicktucker4](https://community.openai.com/u/nicktucker4)\
**Post date:** [January 11, 2024, 11:22pm UTC](https://community.openai.com/t/benefit-of-tool-vs-specific-prompt-instructions/583421/1 "2024-01-11T23:22:00Z")

</div>

Hey all,

I’ve been reading about tools lately and wondering if there would be a major benefit vs. using a specific prompt to get my achieved output, which is what I’m doing now. My ‘tool’ would basically specify the kind of JSON output I want, which is what I’m doing via the following prompts now:

```
        {'role': 'system',
         'content': 'You provide information on incidents that occurred in the following article: ' + article_content},
        {'role': 'system',
         'content': 'There may be multiple incidents listed in a single article. When this is the case, you must use a list of JSON responses. Multiple names, ages, and addresses may be listed in a single incident.'},
        {'role': 'system',
         'content': 'All output must be provided as an array of objects in dictionary or hashmap format, with the following keys: accused_name, accused_age, accused_location, charges, details, legal_actions. All values must be strings. If the article is not about a crime, the output should be N/A.'},
        {'role': 'system',
         'content': 'The JSON output should be an array of objects, with each object representing a single incident. If there is only one incident, the array should contain only one object. Do not include a dictionary or hashmap as the root object.'},

```

What would the benefit of using a tool with a schema use versus my current approach, if any? What would be more efficient?