day10 / our work.

DealPilot

DealPilot

DealPilot

DealPilot

AI gives every investor the same thinking tools. It does not give them the same memory.

Fintech

/01

Client

Cold Start is an innovative VC firm focused on building and investing in generation-defining companies. The firm operates beyond capital – embedding operators and resources alongside founders from early stage through growth.

Cold Start has helped create more than $2B in enterprise value, backed by funds including General Catalyst, Village Global, Founders Fund, and SoftBank.

As both investor and builder, their advantage depends on how well institutional judgment compounds over time

As both investor and builder, their advantage depends on how well institutional judgment compounds over time

/02

Challenge

01

Investment teams don't lack tools — they lack a system that thinks

Decks, transcripts, memos, CRM data, Slack threads: the information exists, but it lives in fragments

02

Existing solutions are deal-centric and workflow-driven — they store, they don't reason

What DealPilot needed wasn’t another layer on top of fragmented infrastructure and documents scattered across multiple systems.

The opportunity wasn’t to build another CRM, sourcing assistant, or data room. It was to build an AI-native investor operating system.

The opportunity wasn’t to build another CRM, sourcing assistant, or data room. It was to build an AI-native investor operating system.

/03

Solution

We designed DealPilot as an AI-native investor operating system where the primary interface isn't a dashboard or a form, it's a prompt.

We designed DealPilot as an AI-native investor operating system where the primary interface isn't a dashboard or a form, it's a prompt.

Investors can ask: "Give me a first glance on Company X", "Prep me for tomorrow's IC", "What risks are emerging across the portfolio?"

Research is the entry point. Structure appears only when needed.

But the bigger architectural decision was how the system handles memory

But the bigger architectural decision was how the system handles memory

Traditional tools assume structure precedes intelligence: you log a deal, fill in the fields, move on. DealPilot works the other way: intelligence generates structure.

Every memo, call transcript, passed deal, and internal debate feeds a persistent reasoning layer that compounds over time.

The result is institutional memory as infrastructure, not document retrieval.

The result is institutional memory as infrastructure, not document retrieval.

Past decisions inform current ones. And the firm's collective judgment doesn't walk out the door when a team member does. But perhaps the most underleveraged asset in venture is the relationship layer: who knows whom, strength of connection, shared history, reference calls.

DealPilot embeds this context directly into research, so relationship intelligence compounds alongside deal intelligence.

/04

Impact

We built an AI-first system that turned fragmented research into compounding institutional intelligence — the kind that survives team changes, scales across deals, and gets smarter over time. Less time on documents. More time on decisions

01

Before

After

01

Repeated research across deals

Repeated research across deals

Unified research across internal + external sources

Unified research across internal + external sources

02

Lost historical context

Lost historical context

Persistent institutional memory

Persistent institutional memory

03

Fragmented CRM + inbox knowledge

Fragmented CRM + inbox knowledge

Relationship graph embedded into research

Relationship graph embedded into research

04

No shared reasoning layer

No shared reasoning layer

Faster conviction cycles

Faster conviction cycles

05

Intellectual capital trapped in documents

Intellectual capital trapped in documents

Firm-owned intelligence layer

Firm-owned intelligence layer

/05

Team snapshot

AI-Augmented Senior Pod

2 FTE

+

Principal Product Architect

Principal Product Architect

0.5 FTE

0.5 FTE

Product Designer

Product Designer

0.5 FTE

0.5 FTE

AI Architect

AI Architect

1 FTE

1 FTE

Traditional build would require

7-8 FTE

We help teams build AI-first companies from day one and re-architect existing businesses to operate at AI speed.

131 Spring Street

New York, NY 10012

2026 All rights reserved.

We help teams build AI-first companies from day one and re-architect existing businesses to operate at AI speed.

131 Spring Street

New York, NY 10012

2026 All rights reserved.

We help teams build AI-first companies from day one and re-architect existing businesses to operate at AI speed.

131 Spring Street

New York, NY 10012

2026 All rights reserved.