DECIDE — Autonomous Business Decision Engine

• 

  Evidence-backed supplier recommendation and multi-agent execution pipeline

• 

  DECIDE dashboard and autonomous business decision workflow

Inspiration

Business decisions often depend on comparing large amounts of messy information from multiple suppliers, quotations, tenders, and documents. Doing this manually takes time and makes it easy to miss important costs, delivery risks, payment terms, or missing requirements.

We wanted to build an AI agent that does more than answer questions. DECIDE — Autonomous Business Decision Engine takes a real business decision, works through the required steps, analyzes available evidence, and produces an actionable recommendation.

What it does

DECIDE is an autonomous business decision agent for procurement, supplier comparison, and tender/RFP evaluation.

A user can provide supplier quotations, tender documents, and a business goal. DECIDE processes the information through a multi-agent workflow to:

• Extract relevant information from business documents
• Normalize supplier and commercial data
• Calculate effective costs and compare offers
• Identify hidden fees, delivery risks, and compliance issues
• Score alternatives using evidence-grounded criteria
• Produce a clear recommendation and next action
• Maintain an auditable decision trail

How we built it

We built DECIDE as a multi-agent application using Gemini and Google Cloud technologies. The workflow separates document intake, classification, extraction, analysis, risk evaluation, decision scoring, and action generation into specialized stages.

The frontend provides a business dashboard where users can start procurement or tender decisions, review evidence, compare suppliers, and inspect the final recommendation.

The system is designed around grounded evidence so that important business recommendations can be traced back to the information processed from the submitted documents.

Challenges we ran into

The biggest challenge was turning unstructured business documents into reliable, comparable information. Supplier quotations and tender documents can use different formats, terminology, pricing structures, and delivery conditions.

We also needed to make the final recommendation understandable rather than simply producing an AI-generated answer. We addressed this by separating extraction, analysis, risk assessment, and decision scoring into distinct stages and presenting the reasoning as an auditable decision trail.

Accomplishments that we're proud of

We are proud of building a working autonomous decision workflow rather than a simple chatbot.

DECIDE can process business information, compare alternatives, identify risks, calculate commercial impact, and produce a concrete recommendation with supporting evidence.

The dashboard also makes the multi-agent workflow visible, showing how the decision moves from document intake through analysis, risk evaluation, scoring, and action.

What we learned

We learned that useful business agents need more than a powerful language model. Reliable document grounding, structured extraction, deterministic calculations, risk checks, and transparent decision logic are equally important.

We also learned that the best agent experience is focused on completing a real workflow and producing an actionable result, rather than simply having a conversation.

What's next for DECIDE — Autonomous Business Decision Engine

Next, we plan to expand DECIDE into a broader business decision platform supporting more procurement categories, additional document formats, supplier intelligence, historical decision learning, automated approval workflows, and deeper enterprise integrations.

Our long-term goal is to make DECIDE a trusted autonomous decision layer for organizations where complex business information needs to become fast, transparent, evidence-backed action.

Built With

• gemini
• google-adk
• google-cloud
• typescript