Agentic AI Competition

2. Task Description

Each team should build an agentic copilot that supports a real-world task in one of the following five domains:

Teams should identify a concrete user group and a practical problem. The copilot should help users complete a task more effectively by planning actions, asking for clarification, retrieving or organizing information, using tools, monitoring progress, and producing useful outputs.

The final system should go beyond a generic chatbot or a simple wrapper around an existing model. It should demonstrate an agentic workflow and show how the agent helps solve a real problem in a usable, novel, and cost-efficient way.

Example Topics by Domain

3. Competition Format and Timeline

Note: Shortlisted teams that wish to attend in person may receive at most one registration-fee waiver for one eligible full-time student member graduating in 2026 or later, subject to verification of full-time student status and the conference registration policy.

Computing Support

4. Submission Instructions

Teams should submit a complete demonstration package that helps reviewers understand the target problem, the agentic copilot design, and the practical value of the prototype. The competition is primarily evaluated through submitted videos and PPT reports. Up to 10 video submissions will be shortlisted for the final award review during ICCSE 2026.

Video Contents.

The video should present an agentic AI copilot related to one of the five competition domains: Education, Health, Law/Ethics, Sustainability, or FinTech. Teams may present a new prototype, an ongoing project, or a more mature application, as long as the video clearly explains how the system depends on Agentic AI techniques and how it supports a meaningful real-world task.

All submitted videos must be presented in English. To make the video easier to follow, teams are encouraged to organize the presentation around the following logic:

• Problem and users: introduce the target users, real-world scenario, and task difficulty the copilot is designed to address.
• Agentic workflow: show how the copilot understands the user goal, plans or coordinates steps, uses tools or skills, retrieves or organizes information, monitors progress, and produces useful outputs.
• Prototype demonstration: include a concrete interaction or workflow demo rather than only conceptual slides.
• Value and impact: explain how the copilot improves usefulness, decision quality, efficiency, accessibility, cost-effectiveness, or social benefit for the selected scenario.

There is no fixed script requirement, but the narration, captions, or on-screen text should be clear and understandable to an international review committee.

Submission requirements.

• Team participation and dual submission: Each participant may join only one team, and each team may submit only one competition entry. Submissions with substantially overlapping team members or substantially identical materials may be treated as duplicate submissions and may be declined at the discretion of the organizers.
• Submission platform: Teams must register and submit materials through TIANCHI.
• Submission package: Each team should submit one ZIP file containing the demo video and the PPT report.
  • File naming: Please name the ZIP package as TeamName_AgenticAICompetition.zip. Inside the ZIP package, please name the video as TeamName_Video.mp4 and the PPT report as TeamName_Report.pptx.
  • Video format: The demo video is recommended to be no longer than 5 minutes and should be submitted in a widely supported format, preferably MP4 with H.264 encoding. Other common formats may be accepted if they can be opened by standard video players.
  • PPT report: The PPT report should indicate one of the five domains, include team information and member contributions. It should summarize the target problem, motivation, target users, agentic workflow, skills and tools, practical usefulness, novelty, efficiency and cost-effectiveness, if any. A PPT report template is provided here: Download PPT template.
• Originality and permissions: Teams may reference existing tools, datasets, models, or platforms, but the submitted copilot should demonstrate original design, implementation, and scenario-specific value. Teams must hold the copyright for the materials used in the video and PPT, or have permission to use and distribute them.
• Content policy: Submissions must be relevant to Artificial Intelligence and must not contain offensive content, personal attacks, privacy violations, or inappropriate advertising or promotion of commercial products.

The organizers reserve the right to investigate and disqualify any teams involved if plagiarism, collusion, duplicate submission, or rule manipulation is suspected.

Review Process.

The committee will first evaluate the submitted video and PPT report. Up to 10 teams will be shortlisted for the final showcase and award review during ICCSE 2026. Shortlisted teams may refine their demo videos and submit a final video version before the conference for showcase and award consideration.

The review will be based on the following criteria:

1. Relevance to Agentic AI: The system should demonstrate meaningful agentic behavior such as planning, tool use, retrieval, multi-step reasoning, progress monitoring, or adaptive decision-making.
2. Novelty and innovation: The solution should show novelty in its scenario, workflow, interaction design, use of agentic skills, or integration of tools.
3. Practical usefulness: The copilot should be genuinely useful for target users. It should reduce workload, improve decision quality, save time, increase access to expertise, or help users complete a real task more effectively.
4. Presentation quality: The video and PPT report should clearly explain the target problem, motivation, user scenario, system workflow, key agentic capabilities, results, limitations, and potential impact.
5. Efficiency and cost-effectiveness: Additional consideration may be given to solutions that are cost-efficient and resource-aware, achieve useful outcomes with reasonable development and operating costs.

To ensure fair review and system authenticity, shortlisted teams may be asked to provide source code or other implementation materials. The organizers may reject any submission at any stage if it is found to lack a real implementation, be misleading, primarily promotional, off-topic, or inconsistent with the stated requirements.

Awards

Other teams will receive participation certificates. 👍