AI Agent Hackathon - Problem Statement Document

Event Overview

The AI Agent Hackathon, hosted by Accenture and Google Cloud, aims to foster business

innovation and customer success through the creation of AI agents. Participants are expected to

submit ideas that address real-world business problems, show measurable impact, and leverage

Google Cloud AI services. Shortlisted teams will receive training, support from Google Cloud, and

opportunities to explore AI-driven business solutions.

Problem Statement Format

Idea Title

ReflectAI - Personal Knowledge Reflection Agent

Idea Description

ReflectAI is an AI-powered personal knowledge management (PKM) assistant that helps users

reflect on and generate content from their existing notes, journals, and documents stored in

Obsidian. Built using Retrieval-Augmented Generation (RAG), the agent leverages Google Cloud

services like Vertex AI and BigQuery to enhance writing and productivity. ReflectAI integrates

semantic and keyword search to retrieve relevant chunks of user notes, enabling generative AI to

provide contextual drafts, summaries, and weekly reflections.

The idea aims to improve knowledge reuse, reduce writer's block, and support personal productivity

through proactive and intelligent content assistance within a privacy-respecting, local-first

framework.
Problem Statement & Solution

In an age of digital overload, users accumulate vast amounts of notes and journal entries but

struggle to meaningfully reflect on or extract value from them. Traditional note-taking apps lack

intelligent summarization or content-generation features that are context-aware.

ReflectAI addresses this by implementing a RAG-based AI agent that works directly with users'

Obsidian vaults. It indexes and chunks note data using semantic embeddings and keyword-based

techniques. Upon a user prompt (e.g., "Write a weekly summary" or "Draft a section on AI agents"),

the agent retrieves relevant notes and uses Vertex AI to generate intelligent, context-aware drafts.

This agent demonstrates proactive behavior, such as suggesting summaries for periods with high

activity or prompting reflections at the end of the week. It interacts via a conversational interface and

operates securely on user data.

To extend functionality, a second agent-TeamMind-can be developed to summarize shared notes

and uncover recurring patterns across team documentation, allowing organizations to discover

collective intelligence. Integration with Google Calendar and Gmail will allow ReflectAI to detect

events or meetings and prompt contextual journaling based on schedules.

Expected Business Impact

ReflectAI benefits knowledge workers, students, researchers, and enterprises with

documentation-heavy workflows. It enhances:

- Productivity: Reduces time spent drafting by reusing existing content intelligently.

- Insight Discovery: Helps uncover patterns and ideas buried in unstructured notes.

- Knowledge Retention: Encourages consistent reflection, improving learning and memory.

- Collaboration: TeamMind supports shared knowledge visibility and insights across projects.

- Contextual Awareness: Calendar/Gmail integrations increase relevance and automation.

documentation systems.

Quantitative Impact:

- Achieves a 30-50% reduction in time spent drafting weekly summaries and project reports.

- Increases idea reuse and knowledge retrieval by 20-40% through smart, context-aware search

capabilities.

- Enhances team collaboration efficiency by up to 25% via automatic shared knowledge

summarization.

- Drives a 3x increase in journaling and reflective writing frequency, powered by proactive AI

prompts.

Qualitative Impact:

- Users experience improved mental clarity and emotional insight through guided, reflective prompts.

- Students and researchers accelerate idea-to-draft workflows with tailored content generation.

- Teams report stronger alignment and fewer knowledge silos, enabled by TeamMind's collective

memory surfacing.

- Users demonstrate higher trust and engagement due to the privacy-centric, local-first architecture.

Unique Value Proposition

ReflectAI is unique due to its hybrid retrieval approach (semantic + classical search), agentic design

focused on autonomy and reflection, and privacy-first integration with local note systems like

Obsidian. Unlike most cloud-first or SaaS writing tools, it emphasizes user control and extensibility.

TeamMind adds an enterprise-level capability for organizational intelligence and collaboration.

Calendar/Gmail context makes it actionable in real-time workflows.

- Number of AI Agents: 2

- Title of AI Agents:

- ReflectAI - Vertex AI Agent - Generates Contextual Reflections from Notes

- TeamMind - Vertex AI Agent - Surfaces Collective Knowledge from Team Notes

- Type of Agents: Vertex AI Agents

Technical Approach & AI Capabilities

- Google Cloud Services Used: Vertex AI, BigQuery, Cloud Storage, Cloud Functions, Google

Calendar API, Gmail API

- AI Model / NLP Requirements: Embedding generation (e.g., textembedding-gecko), fine-tuned LLM

for text generation, keyword extraction, RAG pipeline, named entity recognition, sentiment analysis

- Architecture Diagram: [To be attached]

- Scalability & Security: Local-first indexing; scalable to team use; optional cloud sync;

GDPR-compliant design principles

Development & Execution Plan

- Development Phases:

- Phase 1: Vault ingestion, chunking, and retrieval setup (BM25 + Embeddings)

- Phase 2: Integration with Vertex AI for draft generation

- Phase 3: Add TeamMind agent and GCal/Gmail context detection

- Phase 4: UI/UX layer for prompting and reflection interface

- Phase 5: Testing, feedback, and deployment

- MVP Timeline: 2-3 weeks

- Barriers & Risks:

- Embedding quality for varied note formats

- Adoption among non-technical users

- Calendar/email access and permissions

- Required Support from Google: Technical Mentors, API Credits, Documentation, Access to

Calendar/Gmail API sandbox

Judging Criteria

- Innovativeness: Combines semantic and keyword search in a reflective AI workflow

- Solution Value Proposition: Reuses personal and shared data to generate insights and productivity

- Productivity Impact: Enables faster content generation, reduces overload, improves collaboration

- Presentation: Aligns tightly to PKM and collaborative knowledge challenges with a clear,

autonomous agent narrative

Theme: AI Agents for Functions

Sub-Theme: Human Potential & T&O

Deadline for Submission:

14th April 2025

Submission Platform:

[Form Link - to be provided by the organizers]

Note: Supporting documents (PPT, Word) and architecture diagrams can be attached to strengthen

your submission.