Back to all discussions

  •   24 days ago

Clarification on AI-Operated Business Criteria & Gemini Agent Architecture

Hi everyone,

I'm a non-technical founder participating in the challenge and building an AI-first startup using Gemini and Google Cloud products. My focus is on solving a real business problem and creating a viable business, while leveraging AI agents to operate significant parts of the workflow.

Since I come from a business and growth background rather than a software engineering background, I want to make sure I'm interpreting the competition criteria correctly and building in a way that aligns with the judging expectations.

My planned approach is to use multiple Gemini-powered agents responsible for different business functions such as customer analysis, engagement, content generation, operational reporting, and conversion workflows.

I'm planning to use a mostly free-tier stack initially and currently evaluating tools such as:

* Gemini API
* Google Cloud Free Trial Credits
* Firebase
* Supabase
* n8n
* Telegram
* Google Sheets
* Lovable

A few questions:

1. For judging purposes, what level of autonomy and decision-making should AI agents demonstrate to be considered as meaningfully operating a business rather than simply automating tasks?

2. Would a workflow where Gemini-powered agents analyze data, make recommendations, trigger actions, and manage customer engagement be considered sufficient?

3. Which Google Cloud products would you recommend prioritizing to maximize alignment with the competition objectives while keeping costs low?

4. If you were building an AI-operated startup from scratch for this challenge, which parts of the business would you automate first using Gemini and Google Cloud services?

I'd appreciate any guidance from organizers, mentors, or participants who have experience with similar AI-agent architectures.

Thank you.

  • 4 comments

  •   •   24 days ago

    1. most important part is the result can be monetize - no matter what your way, just use ONE Google Cloud service
    2. sure, focus on one problem with single solution, don't over engineered in this step
    3. Gemini - don't over thinking for price in this step, after win hire team to analyze whole system include costs
    4. Gemini - you can use another cheap cloud hosting

  •   •   24 days ago

    Okay Thank you for your input.

  •   •   21 days ago

    Please only accept valid answers from event managers and sponsors ONLY. Please avoid believing everything you read. @Muhammad Rahman, please verify your identity and STOP SPAMMING for your benefit

  • Manager   •   18 days ago

    Thanks for the detailed questions! Per the official rules, the project must be a business that operates with AI, and projects that include LLM functionality must use the Gemini API for at least one LLM call in the deployed application. Beyond that, the architecture and level of agent autonomy are up to you — judges will evaluate your submission against the published judging criteria.

    We can't advise on specific build strategies or tool choices, but the rules and judging criteria are the best guide to what judges are looking for: https://xprize.devpost.com/rules. The FAQ may also help: https://xprize.devpost.com/details/faq
    Good luck!

Log in or sign up for Devpost to join the conversation.