AI Agent Use Cases
Views —

AI Agent Use Cases

Added today

44All 189

Prompt to Prepare a Shopping Cart from a Recipe Photo

The author reports that Grok Bot receives a recipe photo, organizes the ingredients and quantities, and creates a shopping list by excluding ingredients already available at home.

Grok Bot
Life & Hobbies

1–20 of 189

Work & ResearchWalkthrough

Agent Team for Investigating Company Performance and Reviewing Report Sources

OpenAI's financial research example collects search results, calls analysis agents as needed, and then undergoes a separate review process.

OpenAI Agents SDK

How it works

  1. When the company to investigate and the questions are entered, the planning agent determines search terms to find news, earnings releases, and disclosures.
  2. The search agent gathers summaries using WebSearchTool. The writing agent calls basic analysis and risk analysis agents as tools when necessary and incorporates the results into the report.
  3. Finally, the review agent checks the report for contradictions or missing sources. The execution commands and configuration are available in a public repository.

Results

This is an example that creates a long Markdown report, a short summary, and additional follow-up questions for further investigation.

Notes

This is not a case that proves investment performance or analysis accuracy. The division of roles in the public example was confirmed, and real trading functions are not within the scope of this case.

Source

GitHub Original post ↗

Example prompt · OpenAI Agents SDK

Goal: Create a report investigating a company's performance and reviewing the supporting evidence.

Input & Environment: [Company name], [Quarter to analyze], [Disclosure URL], [Items to compare], [Python·Agents SDK environment].

Reference source: https://github.com/openai/openai-agents-python/tree/main/examples/financial_research_agent

Set up the financial_research_agent example to create a report explaining the specified quarter's performance. Connect the planning, search, basic analysis, risk analysis, writing, and review steps. For each figure, include the currency, unit, reference period, and disclosure source. Separate out any content that cannot be confirmed by search summaries as items for further verification. The deliverables are a Markdown report, a key summary, and a list of issues found during review.

Do not speculate about account connection methods or performance not present in the source. First, provide the necessary setup and execution order. Start from a working copy to preserve existing materials, and mark any items you cannot access as needing verification.

Verification: Check the units for revenue and profit, the year-over-year basis, and whether data from different quarters are mixed. Test whether the review agent catches figures without sources, and exclude unverified claims from the conclusion.

Along with the results, organize the inputs used, execution logs, and confirmed outcomes. Clearly distinguish between completed tasks and those that failed or have not yet been executed.