AI for Crypto-Native Venture Fund

Leading the Crypto Venture Industry with AI – Abdul Al Ali, Venture Investor, at Decasonic

Introduction

At Decasonic, we have long believed that being AI-native is a baseline expectation for how venture funds strengthen their ability to surface and execute on investment-alpha. As investor operators investing at the intersection of Web3 x AI, we are constantly implementing AI to enhance judgement, deepen due diligence, and strengthen our collaboration with founders. We are proud we have developed industry leading AI to better scale productivity for our small venture team.  

In recent months, a growing question has been asked: How does an AI-native venture firm build AI? Today, we are introducing one of the AI applications we have built internally, AI Due Diligence. 

One of the many AI applications we have built at Decasonic, our multi-agent AI Due Diligence application enhances our ability to evaluate liquid tokens with speed, depth and scale. This team is designed to co-partner with our human investment-team, allowing us to make more informed investment decisions faster and at scale. 

An increasingly difficult question most of the businesses are tackling is the measurable impact of AI in their core operations. Traditional business metrics often assess the impact of AI on organizational efficiency by means of ROI, headcount, and revenue growth. 

Most recent publicly available figures reference 40-45% improvements in efficiency, alongside 20% ROI from organizations fully adopting AI in their core operations. The long-term potential impact from AI-driven optimizations in businesses is yet to be measured, but it’s an area that is often tracked as CapEx spending on AI continues to grow. 

In venture capital, and especially in funds focused on early-stage opportunities that could be pre-revenue, pre-product market fit, AI integration is often overlooked. This is because of a potential shared misconception on AI usage that stems from delegation versus enhancement. 

This area of focus is subjective, and AI is mainly an objective partner that is capable of analyzing through vast streams of data to provide often overlooked insights. Being investor-operators with an active, AI-native team means we experiment and build with the wide range of the latest AI tools available to us.

Investing in AI demands a deep understanding of emerging technologies. These AI builds help us become better investors, recognizing areas where the market is evolving, and opportunities that are at the forefront of emerging innovation. Our ability to provide value-add insights to portfolio company founders is amplified. 

Background: How AI might be Deployed in Venture Capital 

Beyond internal automation, AI adoption across venture capital signals a broader shift from intuition-led investing to intelligence-augmented decision-making. 

Broadly, according to Bain & Company, 87% of companies were already developing, piloting or deploying generative AI by early 2024, yet only around 35% had a clearly defined vision for how it would deliver business value. Based on our many conversations with other venture capital firm partners, many are using AI at the individual level, but have yet to embed across collaboration to drive differentiated capabilities. This is a core unlock, where AI amplifies the human-collaboration through multiplayer like workflows and interactions. 

We believe this gap lies in operational design, embedding AI into the investment workflow itself. Some of the existing designs for AI-implementation within companies relies on creating new workflows vs enhancing and amplifying existing ones through AI-integration. This is the gap we aim to address in our own internal builds, and one that helped amplify our internal collective, team impact. Decasonic’s AI Due Diligence (AI DD) represents this next evolution: venture operations as continuously learning, AI knowledge-driven firms.

Some of the areas where AI is often referenced as being accretive to VC funds include:

1. Deal Sourcing: AI automates the screening of large startup datasets and private company repositories, surfacing on-thesis opportunities based on structured parameters, including opportunities defined as ‘on-thesis’. 

   1. For us at Decasonic this is not about replacing human judgment, it’s about scaling it. It shifts the dynamic of work for the investment team away from actively surfacing opportunities and assessing inbound to human-judgement on time prioritization of the pipeline of opportunities available to us. AI surfaces high-probability leads; our venture partners then refine, prioritize, and engage with founders directly. This “AI-judge in the loop” approach ensures we remain both data-informed and founder-connected. It also accounts for potential edge-cases where AI can be misdirectional. 

2. Due Diligence: At Decasonic, our underwriting centers around three key underwriting pillars: Product-Market Fit, Narrative, and Execution. AI enhances this process by structuring raw data into decision-ready insights. Over time, we’ve evolved from single-model interfaces (custom GPTs)  to an orchestrated system of agents, each focused on a core component of our diligence. 

3. Portfolio Management: Amplifies our ability to provide value-add insights to our portfolio company founders. Internal builds that have measurable increase on productivity and internal KPIs allows us to provide insights and drive towards impact in portfolio companies. 

We internally leverage our criteria for “investment alpha” in order to find areas where we can drive enhanced returns to our LPs. AI DD is one of many applications we build in pursuit of investment alpha. 

Introduction to AI Due Diligence (AI DD) 

AI Due Diligence (‘AI DD’) is our internal team of 35 AI agents (with 1 central orchestrator, 3-sub orchestrators) designed to scale the scope and depth of our liquid token diligence. Built into our native AI-OS, AI DD functions as a precision layer: Gathering, structuring, and interpreting data at speed and scale.

Screenshot: AI Due Diligence Application

At the highest level, AI DD operates as a coordinated system of specialized agents, orchestrated by a central controller we call Ingrid. Ingrid routes each task to one of three focused sub-orchestrators, Product-Market Fit (PMF), Narrative, and Execution, mirroring our human underwriting workflows. Each sub-team runs in parallel, leveraging curated data sources, contextual reasoning, and internal knowledge repository to produce structured, verifiable-backed diligence.

Our ecosystem of tools includes both proprietary and public integrations, human and agentic knowledge repository,  specific LLMs for real-time web data, Browser Base for product evaluations, and crypto analytics platforms like CoinGecko, Messari, and Nansen for token and on-chain intelligence. 

Together, these systems allow AI DD to distinguish signal from noise and output human-parity (human-capable investment token memos) across 20+ liquid tokens in under six hours, a 20x+ improvement than non-AI workflows. 

Each memo is reviewed and verified by our investment team. We designed AI DD to prioritize verifiable information gathering over outsourcing investment-decisions, with built-in citation retrieval and cross-verification between independent data points. Human oversight remains central, AI handles structure and speed, while our investment partners are responsible for assessing the quality and providing investment-judgement. 

One issue we faced in the early design of AI DD was information-verification. It is likely the biggest question any AI-builder relying on external data-sources will face. How do you trust the quality of the output? Increasingly, models are becoming much more efficient in providing citations of the data being used. You can now actively leverage API calls with a ‘return citations’ field, where the sources are presented. 

Source: Link

Team Breakdown

Our system’s central orchestrator is Ingrid. Ingrid serves as the coordination layer for AI Due Diligence, managing the flow of information, task distribution, and synthesis across three specialized sub-orchestrators focused on Product-Market Fit (PMF), Narrative, and Execution.

Each sub-orchestrator mirrors a distinct lens of our investment underwriting framework. Ingrid first plans and coordinates structured data gathering across relevant domains: Developer activity, protocol metrics, and open-source repositories for Execution; communications strategy, market positioning, and project/token sentiment for Narrative; and traction signals, user growth, and product differentiation for PMF. This ensures every diligence cycle begins with a comprehensive, evidence-based context. The goal is to allow for AI-driven source optimizations. We provide Ingrid with extensive tools in the form of APIs and MCPs. Ingrid has context on each of the functionality of those tools, and can determine when to leverage them as needed, providing sub-orchestrators with enough context for deterministic interactions. 

Once the raw data insights are collected, Ingrid deploys fine-tuned, internal Decasonic LLMs to structure the sourced information. She synthesizes the sub-teams’ findings into a unified token investment memo, aligning the structure and tone to human-parity standards. 

In venture investing, differentiation often comes from speed, accuracy, and network effects. AI Due Diligence’s orchestration system gives Decasonic a structural advantage in all three. It compresses the information asymmetry gap, allowing our team to discover patterns across on-chain and off-chain data faster than peers. This is where proprietary AI systems strengthen our ability to generate alpha, deepen conviction, and compound founder outcomes.

Amplifying Human-AI Alignment 

The AI DD team exists to deliver fast, rigorous, and repeatable due diligence on crypto liquid tokens. Its goal is to compress the research cycle from days to minutes (per individual token), without compromising on analytical depth, verifiable info, and information rigor, maintaining a comprehensive due diligence memo. 

Usability is a core principle. A key to ensure your AI application and build lacks adoption from core team members is a significant, sudden change in habitual workflows. This disrupts a company's operations, but is more manageable for leaner teams. 

In the case of AI DD, a human investment team member guides our AI DD team on the projects to analyze and run due diligence for. The output as previously mentioned is two fold: Summary and a Comprehensive Token Memo. The investment team member exercises judgement by reading through the summary, and assessing respective time allocation towards enhancing the comprehensive memo published by the AI DD team. This maintains the core principle of Human-AI Alignment, with delegation of tasks between the AI Investment Team Members and the Human Investment Team Members. 

A key principle in designing our internal AI applications is human-AI alignment. The goal in this design is to amplify the respective ‘super powers’ of humans and AI in order to achieve enhanced outcomes. In our case, AI is responsible for information retrieval, normalization, de-duplication, and factual summarization. Humans focus on judgment calls, synthesis across context, and decision making. The two roles compliment each other, allowing for the scaling of human judgement and the ability to make investment decisions, faster. 

Technical Diagram

Below is a representation of the technical architecture that powers our AI Due Diligence (AI DD) team. This high-level overview illustrates how our orchestrated agent system converts fragmented, unstructured information into decision-ready diligence outputs. We hope the below diagram is helpful to partner venture investors, and future portfolio company founders. 

At the core of the architecture sits our Central Orchestrator (“Ingrid”), which routes tasks, manages dependencies, and synthesizes outputs across our three specialized sub-teams: PMF, Narrative, and Execution. Each team operates autonomously and in coordination. 

• Input Layer - User & Backend API

  The process begins with a user prompt or CSV input submitted via our internal backend API. This defines the scope of the analysis and the input. 

• Orchestration Layer - Ingrid:

  Ingrid decomposes each diligence request into structured tasks distributed across the three sub-teams. She ensures data provenance, parallel execution, and quality control throughout each stage.

• Sub-Teams - PMF, Narrative, Execution:

  Each team sources information from tailored data domains:

  • PMF focuses on product documentation, and traction signals.

  • Narrative aggregates from web, media, and social data with a focus on social signals.

  • Execution analyzes protocol metrics, developer activity, code repositories, and founder-background. 

• Data Persistence Layer - Vector Databases:

  Each team’s factual extractions are stored as dated vector entries, enabling longitudinal analysis and continuous re-underwriting over time. 

• Output Layer - Report Builder & Distribution:

  Ingrid synthesizes the results into structured memos, accessible via internal databases. 

Screenshot: AI Due Diligence (AI DD) Technical Architecture Overview

AI x VC Convergence

We’re entering an era where venture capital itself becomes an AI-native product. The convergence of large language models, on-chain data, and knowledge graphs means that due diligence, portfolio support, and risk modeling will increasingly be machine-assisted. For venture capital, this represents differentiation. Those who operationalize AI as infrastructure will define the next generation of alpha creation.

Conclusion 

AI DD is a reflection of our core belief that AI doesn’t replace human-judgement, it amplifies and scales it. It represents how we operationalize our belief across our investing practice, enabling us to deliver sharper insights, faster execution, and enhanced value-add to our portfolio companies in the age of AI-driven acceleration.

At its heart, AI DD is a story of collaboration between human judgment and AI precision. Every AI-generated insight loops back to a human investor’s conviction, experience, and empathy. In that sense, AI amplifies the depth of human understanding. For Decasonic, that balance is the ultimate measure of innovation where technology scales both output and insight.

If you are building at the intersection of Web3 x AI, and can use the expertise of our investor-operator investment team, reach out to us. 

The content of these blog posts is strictly for informational and educational purposes and is not intended as investment advice, or as a recommendation or solicitation to buy or sell any asset. Nothing herein should be considered legal or tax advice. You should consult your own professional advisor before making any financial decision. Decasonic makes no warranties regarding the accuracy, completeness, or reliability of the content in these blog posts. The opinions expressed are those of the authors and do not necessarily reflect the views of Decasonic. Decasonic disclaims liability for any errors or omissions in these blog posts and for any actions taken based on the information provided.