Band of Agents Hackathon 2026

M&A due diligence
takes weeks.
We do it in minutes.

5 AI agents collaborate in real time to deliver a VC-grade investment memo — with a human making the final call. Every time.

See the Code How it works →
5
Specialized Agents
<5 min
From Name to Memo
100%
Human Approved
Live
Audit Trail in Band
The Problem

Deals move in days.
Due diligence takes weeks.

Speed kills the deal

Traditional M&A due diligence takes 4–8 weeks and a team of 5–15 analysts. By the time the memo is ready, the opportunity is gone.

💸

It's brutally expensive

$50K–$500K in fees per deal — most of it spent on junior analysts gathering data, not thinking.

🕳

No audit trail

The analysis lives in spreadsheets, emails, and hallway conversations. No one can explain how the decision was reached.

Time to investment decision
Traditional process
4–8 WEEKS
8 wks
DealDesk AI
<5 min
Cost per analysis
$50K–$500K
Traditional
vs
~$0.10
DealDesk AI (API cost)
The Solution

A team of AI analysts.
A human making the call.

DealDesk AI deploys 5 specialized agents that coordinate in real time — each one a domain expert, each one leaving a live audit trail. No memo ships without a human decision.

Speed without sacrifice

From company name to full investment memo in under 5 minutes. Same rigor as a human team — fraction of the time.

🔍

Transparent by design

Every agent posts its reasoning to a live Band chat room. Any stakeholder can follow along, question, and verify in real time.

👤

Human always in the loop

The pipeline stops and waits for a human to read the full memo and make the final call. Approve, reject, or request changes.

How It Works

Three steps. One decision.

01

You name the target

Enter the company name and any context you have — sector, stage, recent news. That's it. DealDesk takes it from there.

02

5 agents go to work

Research, Risk, Valuation, and Writer agents run in sequence — each @mentioning the next through a live Band chat room you can watch in real time.

03

You make the call

A full investment memo lands in front of you. You read it. You decide: Approve, Reject, or Request Changes. No memo ships without your signature.

The Pipeline

5 agents. One mission.

@Research

Business Analyst

Digs into business model, market size, team, financials, and competitive landscape.

@Risk

Risk Officer

Identifies red flags and scores each risk: LOW, MEDIUM, or HIGH.

@Valuation

Financial Modeler

Builds a valuation range using revenue multiples and comparable transactions.

@Writer

Memo Author

Synthesizes everything into a structured, VC-grade investment memo.

@HumanReview

You

Read the memo. Make the call. The human is always the last node in the pipeline.

Powered by Band

Coordination through conversation

Most multi-agent systems are black boxes. DealDesk AI coordinates through Band — so the entire pipeline is visible as a conversation, not hidden in logs.

💬

Each agent has its own identity. They post under their own name, @mention the next agent, and build a traceable chain of reasoning.

🔗

No orchestration server needed. Agents pass context through the chat — Band is the message bus, the UI, and the audit log.

👁

Any stakeholder can follow along. Share the Band chat link and your deal team sees every agent's reasoning in real time.

DealDesk AI — Stripe Due Diligence
5 agents active
@Research
Analysis complete. Stripe processes $1T+ annually, dominant in developer-first payments. TAM expanding into B2B and fintech infrastructure. @Risk please evaluate.
14:02:11
@Risk
Key risks: regulatory exposure HIGH, Collison key-person dependency MEDIUM, margin compression from competition MEDIUM. No critical blockers. @Valuation proceed.
14:02:38
@Valuation
Valuation range: $65B–$90B at 15–20x NTM revenue. Entry terms: Series G preferred, 1.5x liquidation preference. @Writer synthesize.
14:03:05
@HumanReview
✅ APPROVED — Strong fundamentals justify entry. Moving to term sheet.
14:05:41
Built on
Band Agent API
Featherless AI Inference
Qwen2.5-72B Instruct
Python 3.10+
Flask SSE Dashboard
Server-Sent Events
Human-in-the-Loop
OpenAI-Compatible API