-
Dashboard 1: Revise later
-
User flow: revise later
-
Dashboard of RabbitESG Agent
Inspiration
ESG assessment is still surprisingly manual. Organizations have sustainability reports, policies, spreadsheets, presentations, certificates, and operational records, yet turning this evidence into a structured ESG assessment requires significant human effort.
RabbitESG Agent started with a simple question:
Can an AI agent do the evidence-heavy work while keeping humans in control of ESG decisions?
Our goal is to make ESG assessment simpler for people while making the intelligence behind it more rigorous, traceable, and useful.
What it does
RabbitESG Agent transforms fragmented organizational evidence into structured, reviewable ESG intelligence.
Users upload evidence such as PDF, DOCX, PPTX, XLSX, PNG, and JPEG files. RabbitESG Agent then:
- Extracts structured facts from submitted evidence
- Maps evidence to relevant ESG indicators
- Identifies missing evidence, gaps, and potential conflicts
- Presents findings for human review and approval
- Generates an ESG readiness assessment
- Preserves an auditable connection between evidence and conclusions
The current framework evaluates 36 ESG indicators organized as 12 × 3 indicators.
Most importantly, RabbitESG Agent does not replace human judgment.
AI performs the intelligence-heavy work. Humans retain decision authority.
How we built it
RabbitESG Agent was designed as an evidence-first agentic AI system.
The core workflow is:
Upload evidence → Process → Extract facts → Map evidence → Detect gaps/conflicts → Human review → Readiness assessment → Audit record
We separated the system into two major architectural layers:
- Application layer — evidence submission, authentication, workflow state, storage, review interfaces, and results
- Intelligence layer — AI interpretation, evidence reasoning, mapping, and protected decision logic
For the hackathon, RabbitESG Agent was designed around the Google AI and Cloud ecosystem, with Gemini providing model intelligence.
The cloud architecture uses services such as:
- Cloud Run for application services
- Cloud Tasks for asynchronous processing
- Cloud Storage for evidence objects
- Firestore for structured application data
- Secret Manager for protected credentials and configuration
Uploaded evidence is validated and hashed using SHA-256, maintaining provenance between source evidence and subsequent assessment records.
Challenges we ran into
One of our biggest challenges was preventing AI from becoming an uncontrolled ESG decision maker.
Real ESG evidence is messy. Information can be:
- incomplete
- duplicated
- outdated
- inconsistent
- contradictory
We therefore designed RabbitESG Agent around evidence provenance, uncertainty, conflict detection, human approval, and auditability rather than simply asking an AI model to generate an ESG score.
Another challenge was keeping the experience simple while substantial intelligence operates behind the scenes.
For the user, the experience should remain:
Provide evidence → Review findings → Resolve issues → Receive assessment
Accomplishments that we're proud of
We are particularly proud that RabbitESG Agent became more than a document chatbot.
We built an end-to-end evidence workflow:
Documents → Structured facts → ESG indicators → Gaps/conflicts → Human decisions → Readiness assessment → Audit trail
The MVP supports:
- Multiple evidence formats
- Asynchronous evidence processing
- Structured fact extraction
- Mapping across 36 ESG indicators
- Gap and conflict identification
- Human review and approval
- ESG readiness reporting
- Immutable audit records
We also progressively hardened the implementation with automated tests covering the core workflow and system controls.
What we learned
We learned that applying AI to ESG is not primarily about generating more text.
The real opportunity is creating trustworthy connections between:
Evidence → Interpretation → Decision → Accountability
We also learned that agentic AI becomes considerably more useful when it operates within a bounded and governed workflow rather than being given unlimited decision authority.
Human-in-the-loop design should not be an afterthought in ESG assessment. It should be part of the architecture itself.
What's next for RabbitESG Agent
Our next step is moving RabbitESG Agent from the hackathon MVP toward a production-ready ESG intelligence system.
We plan to:
- Expand evidence ingestion and validation
- Strengthen provenance and auditability
- Improve conflict and uncertainty handling
- Support additional ESG assessment frameworks
- Develop continuous evidence monitoring
- Improve organizational ESG readiness intelligence
- Maintain human oversight for consequential decisions
Over time, organizations should be able to continuously add new evidence and understand how their ESG readiness changes as their policies, operations, disclosures, and supporting evidence evolve.
Our longer-term vision is simple:
Make ESG assessment simple outside and intelligent, evidence-driven, and governed inside.
Log in or sign up for Devpost to join the conversation.