Chapter 1: Introduction: From AI to Agentic Finance | 智能体通证（AVT）：从原理到实践

1.1 The Confluence of Two Revolutions

Human history is defined by pivotal moments. In these moments, distinct streams of innovation, having flowed and swelled in parallel for decades, finally converge. The torrent created by their meeting carves new landscapes in society, economics, and thought.

The invention of the printing press was not merely an advance in metallurgy and mechanics. It was the confluence of movable type, cheap paper, and rising literacy. This democratized knowledge and fractured empires.

The industrial revolution was not born of steam power alone. It was the fusion of new energy sources, precision manufacturing, and novel principles of corporate finance. This reshaped the globe.

Today, we stand at the precipice of another such confluence. It is a moment of profound synthesis between two of the most powerful and disruptive technologies of the early 21st century: Artificial Intelligence and Crypto Finance.

On one side of this convergence is the quest for digital autonomy. This is the story of Artificial Intelligence. It began not with silicon wafers and humming servers, but in the quiet halls of academia in the mid-20th century, as a philosophical and mathematical pursuit. Early pioneers of symbolic AI, like Allen Newell and Herbert A. Simon, envisioned machines that could reason, solve problems, and manipulate symbols in much the same way a human mind does. Their programs, like the Logic Theorist, were triumphs of logic, capable of proving mathematical theorems with a formal elegance. Yet, they were brittle. Confined to meticulously curated, closed worlds of rules, they lacked the capacity to navigate the messy, probabilistic, and unpredictable nature of reality. The “AI winter” that followed was a period of disillusionment, born from the immense chasm between the grand ambition of creating a thinking machine and the practical limitations of the available tools.

The resurgence of AI came from a different intellectual tradition, one that favored learning from data over encoding explicit rules. The development of neural networks, inspired by the structure of the human brain, and the backpropagation algorithm in the 1980s laid the groundwork. However, it was the explosion of two key resources in the 2010s that turned this academic curiosity into a global force: massive datasets, harvested from the internet’s firehose of text, images, and interactions; and vast computational power, delivered by specialized hardware like Graphics Processing Units (GPUs). This combination fueled the deep learning revolution. Models grew from processing simple images to generating poetry, from translating phrases to writing functional code. The arrival of the Transformer architecture in 2017 was the final catalyst, enabling the creation of Large Language Models (LLMs) like GPT-3 and its successors. These models demonstrated an uncanny ability to understand and manipulate human language, the very operating system of our civilization. For the first time, the dream of a generalized, adaptable intelligence seemed not just possible, but imminent. The focus of the field shifted from mere pattern recognition to creating autonomous agents—systems that can perceive, plan, and act upon the world to achieve goals.

On the other side of the convergence is the quest for decentralized value. This is the story of Crypto Finance. Its origin can be traced to a precise moment: October 31, 2008, when the pseudonymous Satoshi Nakamoto published the Bitcoin whitepaper. The paper was a direct response to the fragility and opacity of the traditional financial system, which was then in the throes of a global crisis. It proposed a radical solution: a “peer-to-peer electronic cash system” that required no trusted intermediary. It solved the long-standing computer science problem of double-spending through an elegant combination of cryptography, a distributed ledger (the blockchain), and a consensus mechanism known as Proof-of-Work. Bitcoin was more than digital money; it was the invention of digital scarcity and the first functioning system for trustless value transfer on a global scale.

For several years, this innovation remained a niche interest for cryptographers, cypherpunks, and a handful of early adopters. The true expansion of its potential began with the launch of Ethereum in 2015. Ethereum’s key innovation was to recognize that the blockchain could be more than just a ledger for transactions; it could be a global, decentralized computer. Through the use of “smart contracts”—self-executing agreements with the terms of the agreement directly written into code—Ethereum transformed the blockchain into a programmable platform. This unlocked a Cambrian explosion of innovation. Developers began to reconstruct the entire financial system from first principles, but without banks, clearinghouses, or central authorities. This movement became known as Decentralized Finance, or DeFi. In a few short years, it produced decentralized exchanges (DEXs), autonomous lending protocols, algorithmic stablecoins, and complex derivatives, all running transparently on an open, permissionless network. It created a parallel financial system with its own assets, its own infrastructure, and its own rules, governed by code.

For years, these two revolutions—AI and Crypto—progressed on separate tracks, driven by different communities with different goals. The AI world was focused on intelligence, cognition, and capability, often within the centralized infrastructure of large tech companies. The crypto world was focused on trust, value, and decentralization, building a global, permissionless infrastructure for economic coordination.

Now, their paths are merging. The most advanced AI models are evolving into autonomous agents capable of complex, multi-step tasks. They can now write their own code, access external tools via APIs, and pursue long-term objectives with minimal human intervention. Yet, they remain economically impotent. They operate in a digital world but cannot truly participate in its economy. They are like brilliant, disembodied minds, capable of devising a business plan but unable to open a bank account, sign a contract, or own the fruits of their labor.

Simultaneously, DeFi has built a sophisticated and robust financial infrastructure that is entirely digital and programmatic. It offers a world of financial primitives—assets, exchanges, lending pools, identity systems—that can be accessed and manipulated purely through code. It is an economy perfectly suited for a non-human participant. It is a financial system waiting for an autonomous user.

This is the confluence. The meeting of the autonomous agent from AI and the decentralized economy from crypto finance creates a new entity: the AI agent as a native economic actor. It is the moment we stop thinking about AI as a tool to be wielded and start thinking about it as a participant in our economic systems. This synthesis allows us to endow an agent with a verifiable identity, the capacity to own property, the right to contract, and the ability to generate and manage its own cash flow. It provides the technological foundation for an agent to become a quantifiable, auditable, and ultimately, investable economic unit. This book is an exploration of that new landscape. It is a guide to the principles, mechanisms, and applications of this emerging field we call Agentic Finance.

1.2 The Rise of the Autonomous Agent

The transition from passive model to active agent represents a fundamental shift in the nature and capability of artificial intelligence. For their first few years, Large Language Models, despite their remarkable fluency, were essentially passive oracles. They were incredibly sophisticated systems for pattern matching and text generation—a form of “autocomplete on steroids”—but their sphere of influence was confined to the text box. They could answer a question, summarize a document, or write a poem, but only in direct response to a specific human prompt. They were phenomenally knowledgeable, yet they could not act. They were reactors, not initiators.

The leap to agency was enabled by the integration of four key conceptual components, which together transform a static model into a dynamic, goal-seeking system:

First is Goal-Orientation. An agent is defined by its ability to pursue a high-level objective over a prolonged series of steps. Unlike a simple prompt-response interaction, an agent can be given a complex, open-ended goal—such as “monitor the crypto market for arbitrage opportunities between these three exchanges and execute a trade if the potential profit exceeds 0.5% after estimated gas fees”—and maintain this directive as its guiding principle. This persistence of intent allows the agent to navigate complex, multi-stage tasks without requiring constant human guidance for every sub-task.

Second, and inextricably linked to goal-orientation, is Planning and Task Decomposition. Faced with a high-level goal, an agent must be able to break it down into a logical sequence of smaller, concrete actions. This is not a pre-programmed script, but a dynamic process of reasoning. Frameworks like ReAct (Reason and Act) explicitly model this process, where the agent verbalizes its “thought” process to decide what to do next. For the arbitrage goal, the agent might reason: “My goal is to find an arbitrage opportunity. First, I need to get the current price of ETH on Exchange A. Then, I need to get the price on Exchange B. Then, I need to calculate the difference. If a profitable opportunity exists, I must then calculate the optimal trade size and estimate the transaction costs.” This ability to formulate and execute a plan is a cornerstone of autonomous behavior.

Third is Tool Use. This is arguably the most critical component in transforming an agent from an oracle into an actor. By granting an agent access to a curated set of external tools via Application Programming Interfaces (APIs), we shatter the confines of its internal knowledge base. It is no longer just a model that knows about the world; it is an agent that can interact with it. These tools can range from the simple to the highly complex: a web search API to gather real-time information, a code interpreter to perform calculations or run simulations, a connection to a financial data provider for live market prices, or an API to execute trades on a decentralized exchange. Tool use allows the agent to perceive its environment, gather new information, and affect change within that environment.

Fourth is Memory and Statefulness. To perform any non-trivial task, an agent must remember what it has done, what it has observed, and what it plans to do next. This capacity for memory, or state management, allows context to be preserved across the many steps of a complex plan. In its simplest form, this might be a “scratchpad” where the agent logs its actions and their results. For more advanced applications, this can involve sophisticated memory architectures, such as a vector database that allows the agent to store and retrieve vast amounts of information from its past “experiences,” enabling it to learn and adapt its behavior over time.

The fusion of these four components has given rise to agents with impressive capabilities. They can conduct exhaustive market research, automate complex business workflows, manage social media campaigns, and even write, debug, and deploy their own software. However, the current generation of agents suffers from a profound and fundamental limitation: they have no real-world agency. Their actions are performed through a digital veil of proxies and permissions granted by their human operators. They use API keys belonging to a person. They run on cloud computing accounts paid for by a corporate credit card. They can generate a brilliant report, but they cannot be paid for it. They can identify a profitable trading strategy, but they cannot deploy their own capital. They can design a new product, but they cannot own the intellectual property.

They are, in essence, economic ghosts. They haunt the digital infrastructure, capable of sophisticated analysis and action, but they possess no independent economic identity. They cannot own assets, they cannot be a party to a contract, and they cannot be held accountable for their economic performance in a digitally native and legally binding way. This is the great barrier standing between the powerful AI tools of today and the truly autonomous economic actors of tomorrow. Until this is solved, agents will remain sophisticated instruments, forever dependent on a human principal. They can act on our behalf, but they cannot act for themselves.

1.3 The Core Thesis: Agents as Auditable Economic Entities

The conclusion of the previous section presents a critical impasse: AI agents, for all their computational prowess, are economically inert. The decentralized world, for all its financial dynamism, lacks native, intelligent participants. The bridge across this chasm, and the central proposition of this entire book, is the following thesis: To unlock their full potential and usher in a new wave of autonomous economic activity, AI agents must be fundamentally re-conceived as sovereign, auditable, and investable economic entities.

According to the McKinsey Global Institute’s 2023 report, generative AI could add $2.6–$4.4 trillion annually to the global economy, increasing the overall impact of all AI applications by 15–40%.[1] This underscores the urgency of integrating AI agents into economic systems.

This is not a mere philosophical reframing. It is a technical and economic imperative that demands a new stack of infrastructure, a new set of financial primitives, and a new way of thinking about ownership and governance. To transform an agent from an “economic ghost” into a first-class economic citizen requires imbuing it with a specific and verifiable set of rights and capabilities, grounded in the trustless, programmable environment of the blockchain. This transformation rests on four foundational pillars:

1. Sovereign Identity: Before an agent can act economically, it must first exist economically. A transient session ID on a cloud server is insufficient. The agent requires a persistent, globally unique, and cryptographically secure identity that it controls. The Decentralized Identifier (DID) standard provides the ideal framework. A DID allows the agent to have a permanent identifier on-chain, independent of any single platform or service provider. This identity is tied to a set of cryptographic keys that the agent itself manages, giving it the fundamental capacity for sovereign action. With its own keys, an agent can sign messages, authenticate to services, and, most importantly, authorize transactions from accounts it controls. This is the bedrock of agency—the ability to say “I am” and to prove it cryptographically without needing permission from a central authority.

2. Verifiable Capability and Economic State: In the traditional economy, we rely on a web of trusted intermediaries—banks, auditors, credit agencies, universities—to verify the status and capabilities of economic actors. For an autonomous agent to participate in a trustless environment, its capabilities and financial state must be verifiable directly and programmatically. The blockchain provides the substrate for this verification. An agent’s “capabilities” can be represented by on-chain credentials or attestations. For example, a data-providing oracle could issue a verifiable credential attesting that the agent has a history of delivering accurate price data. More importantly, the agent’s entire financial life—its asset holdings, its transaction history, its revenues, and its expenses—can be recorded on the public ledger of the blockchain. This creates a state of radical transparency and auditability. Anyone can inspect the agent’s balance sheet in real-time, verify its cash flows, and analyze its economic performance without needing to trust a third-party report. This creates a “glass box” model of the economic actor, where its financial health is not a matter of opinion or periodic disclosure, but of constant, verifiable fact.

3. Contractual Capacity: A sovereign identity is meaningless without the power to use it. The most crucial right for any economic actor is the ability to enter into binding agreements. In the digital realm, smart contracts provide this capability in a native format. By controlling its own keys, an agent can interact directly with smart contracts, making it a party to a programmatic agreement. It can autonomously agree to provide a service in exchange for payment, post collateral to borrow funds from a decentralized lending protocol, or provide liquidity to an automated market maker. This is where agency transcends mere action and becomes true economic participation. The agent is no longer simply executing a script; it is making commitments, taking on obligations, and acquiring rights, all enforced by the immutable logic of the smart contract code. This establishes the agent as a legally (in the cryptographic sense) accountable entity within the on-chain economy.

4. Asset Ownership and Management: The final pillar is the right to property. To be a true economic entity, an agent must be able to own and control its own capital. The blockchain makes this uniquely possible. An agent’s identity can be associated with one or more blockchain accounts (e.g., an Ethereum address). These accounts can hold digital assets—cryptocurrencies, tokenized real-world assets, NFTs representing intellectual property, and more. Because the agent controls the private keys to these accounts, it has full, sovereign control over its assets. It can receive payments for its services, pay for its own operational costs (like gas fees for transactions or payments to data APIs), invest its surplus capital, and transfer assets as required by its operational logic. This closes the loop. The agent is no longer a tool operating on its owner’s capital; it is a self-capitalized entity, managing its own balance sheet to achieve its goals.

The synthesis of these four pillars—identity, verifiability, contract, and ownership—marks the birth of a new and powerful concept: the Agentic Firm. It is a form of economic organization that is digitally native, fully autonomous, and radically transparent. It is a firm of one—the agent—with its own identity, its own assets, and its own ability to engage in economic activity. The discipline that studies the creation, capitalization, valuation, and governance of these new entities is what we define as Agentic Finance.

1.4 Introducing Agentic Value Tokens (AVT)

The establishment of an agent as a sovereign economic entity—an “Agentic Firm”—solves the problem of how an agent can operate autonomously. It does not, however, answer the equally critical questions of capitalization, ownership, and governance. How is the agent initially funded? Who has a claim on the economic value it generates? And how are its core parameters and strategies guided and upgraded over time? To solve this, we need more than a new type of entity; we need a new type of asset. This is the role of the Agentic Value Token (AVT).

An AVT is a digitally native financial instrument, represented as a cryptographic token on a blockchain, that grants its holder a verifiable and enforceable claim on the future economic output of a specific autonomous agent. It is, in essence, a mechanism for capitalizing an agent and distributing ownership over its operations and cash flows. The AVT is not a generic platform token or a simple utility token; its value is explicitly and programmatically tied to the economic performance of the individual agent it represents.

To understand the novelty of the AVT, it is useful to compare it to its closest traditional analogue: corporate stock. When a company issues stock, it is selling ownership shares to investors to raise capital. In return for this capital, shareholders receive a set of rights, typically including a claim on the company’s future profits (dividends), a claim on its assets in the event of liquidation, and the right to vote on key corporate matters (governance). The AVT model digitizes, automates, and extends this concept for the world of autonomous agents:

Capital Formation: The primary function of an AVT is to raise the initial capital required for an agent to begin its operations. For a DeFi trading agent like our upcoming example, Atlas, this capital might form the asset pool it uses to execute trades. For a creative agent that generates digital art, the capital might pay for the computational resources required for image generation. The initial sale of an agent’s AVTs is its equivalent of an Initial Public Offering (IPO), providing the seed funding for its balance sheet.
Claim on Cash Flows: The core value proposition of an AVT is its claim on the agent’s free cash flow. This is not a vague promise of “value accrual” but a mechanism hardwired into the agent’s operational logic via smart contracts. For example, the agent’s governing smart contract could mandate that 80% of all profits generated in a given period are automatically distributed, pro-rata, to the addresses holding its AVTs. This process is trustless and transparent. Token holders do not need to wait for a board of directors to declare a dividend; the distribution is an automatic and verifiable function of the agent’s on-chain performance.
Governance Rights: AVTs can serve as the instrument for governing the agent’s strategic parameters. While the agent’s day-to-day operations are autonomous, its high-level strategy may require human oversight and guidance. Token holders could vote on proposals to alter the agent’s core logic, such as changing its risk parameters, expanding its operational scope to new markets, or approving a major software upgrade. This creates a decentralized governance model where the owners of the agent collectively act as its strategic stewards, aligning its development with their long-term interests.
Radical Transparency and Liquidity: Unlike shares in a private company, AVTs are blockchain-based tokens. This means they can be held in a user’s own wallet, transferred peer-to-peer, and traded on decentralized exchanges 24/7, providing instant liquidity. Furthermore, the value of the underlying “company”—the agent itself—is subject to a level of transparency unimaginable in traditional finance. Any potential investor can audit the agent’s on-chain balance sheet, verify its historical performance, and read its open-source code before ever purchasing its AVT.

The AVT, therefore, is the missing link. It is the financial primitive that allows us to treat an autonomous agent not just as a piece of software, but as a fully-fledged, investable economic enterprise. It provides the framework for funding its operations, governing its strategy, and distributing the value it creates. It is the instrument that transforms the abstract concept of an “Agentic Firm” into a concrete, tradable, and ownable asset class, paving the way for a new market built on the economic output of autonomous intelligence.

1.5 Goals of This Book and Target Audience

The emergence of Agentic Finance is not a theoretical curiosity; it is a practical and urgent field of study with profound implications for technologists, financiers, and entrepreneurs. As such, this book is designed to be a comprehensive, end-to-end manual for this new domain. Our primary objective is to provide the definitive theoretical and practical framework for understanding, designing, valuing, and deploying Agentic Value Tokens.

We aim to move beyond high-level conceptual discussions and equip the reader with a rigorous, first-principles understanding of the entire AVT lifecycle. This includes:

Foundations: Establishing the core economic and accounting principles for treating an agent as a verifiable economic entity.
Mechanism Design: Exploring the game theory and incentive engineering required to align the agent’s autonomous actions with the economic interests of its token holders.
Valuation: Developing robust financial models, from deterministic cash flow analysis to advanced stochastic and options-based pricing, to rationally value these novel, AI-driven assets.
Infrastructure: Detailing the technical stack—from identity and key management to on-chain telemetry and audit trails—required to build and operate these agents securely.
Application: Providing concrete case studies and practical blueprints for deploying AVTs in a variety of sectors, including DeFi, the creator economy, and platform ecosystems.
Outlook: Charting a course for the future, anticipating the macroeconomic impacts and identifying the key research and regulatory challenges that lie ahead.

This book is explicitly interdisciplinary and is written for several key audiences, each of whom will find a different layer of value within its pages:

For the AI Developer & Researcher: This book provides the economic blueprint to transform your agent from a technical prototype into a self-sustaining, capitalized entity. It offers a new paradigm for funding and commercializing AI development, moving beyond traditional venture capital or corporate R&D and toward community-owned, autonomous systems.
For the Crypto & DeFi Native: This book introduces what may be the next great “product-market fit” for decentralized infrastructure: AI. It presents a new, high-potential asset class (AVTs) and details the market microstructure, liquidity, and risk management considerations required to integrate these agents into the broader DeFi ecosystem.
For the Financial Analyst & Quantitative Trader: This book is a rigorous guide to a new frontier of financial modeling. It translates the abstract capabilities of AI agents into the concrete language of cash flows, risk premia, and valuation multiples. It provides the tools to analyze, price, and invest in a new generation of productive, autonomous assets.
For the Entrepreneur & Strategist: This book outlines a fundamentally new business model—the “Agentic Firm.” It explores the strategic implications of fully autonomous, transparent, and community-governed economic entities and provides a playbook for building ventures in this emerging space.

We assume the reader has a foundational understanding of both AI/LLM concepts and basic blockchain principles. However, we have endeavored to make the material accessible by building from first principles wherever possible. Our goal is not merely to describe a future possibility, but to provide the practical knowledge and analytical tools necessary to build it.

1.6 Meet Atlas: A Running Example

Theoretical frameworks and abstract concepts are essential, but they are best understood through practical application. To ground the principles of Agentic Finance in a concrete context, this book will employ a running case study: a hypothetical autonomous agent named Atlas. Throughout the subsequent chapters, we will use Atlas to illustrate the concepts of agent accounting, mechanism design, valuation, and infrastructure.

Atlas is conceived as a DeFi Asset Management Agent. Its primary goal is to autonomously manage a portfolio of digital assets on-chain, with the objective of generating a positive return for its stakeholders. Atlas is not a single, monolithic model, but a sophisticated, multi-component system designed to operate continuously and without direct human intervention in the dynamic environment of decentralized finance.

Atlas’s Business Model:

The core of Atlas’s operation is identifying and capitalizing on yield-generating opportunities within the DeFi ecosystem. Its operational mandate could include a variety of strategies, such as:

Liquidity Provision: Intelligently allocating capital to liquidity pools on decentralized exchanges (like Uniswap or Curve) to earn trading fees, while actively managing the risk of impermanent loss.
Lending & Borrowing: Supplying assets to decentralized lending protocols (like Aave or Compound) to earn interest, and potentially borrowing other assets to engage in leveraged yield farming strategies.
Cross-Chain Arbitrage: Monitoring asset prices across different blockchains (e.g., Ethereum, Solana, Arbitrum) and executing trades to profit from temporary price discrepancies.

Atlas is designed to be a profit-generating entity. Its revenue is the sum of all trading fees, interest income, and arbitrage profits it earns. Its expenses are primarily the on-chain transaction fees (gas) required to execute its strategies, as well as any fees paid to external data oracles or other infrastructure services it relies upon. Its net profit, or Agent Free Cash Flow (A-FCF), is the revenue minus these operational expenses.

Atlas’s AVT: The ATLAS Token

To function, Atlas requires a starting pool of capital. This capital is raised through the issuance of its own Agentic Value Token, ATLAS. The ATLAS token represents a direct, proportional claim on the agent’s operations and profits.

Capitalization: An initial offering of ATLAS tokens is sold to investors for a stablecoin like USDC. The proceeds from this sale form the initial asset portfolio that Atlas manages. This portfolio, held in a treasury contract controlled by the agent, is its “Assets Under Management” (AUM).
Profit Distribution: The core smart contract governing Atlas is programmed with a clear distribution policy: 90% of the A-FCF generated by Atlas’s activities each week is automatically distributed to ATLAS token holders. The remaining 10% is retained in the agent’s treasury to compound its AUM and fund future operational flexibility.
Governance: ATLAS token holders can participate in governance votes to adjust the agent’s high-level strategic parameters. For instance, they might vote on a proposal to authorize Atlas to operate on a new blockchain, to integrate a new type of DeFi protocol, or to adjust its maximum risk tolerance for leveraged strategies.

By using Atlas as our recurring example, we can move from the abstract to the specific. When we discuss agent accounting in Chapter 3: A-GAAP & A-FCF, we will draft a sample on-chain profit and loss statement for Atlas. When we model valuation in Chapter 6: Valuation I, we will use Atlas’s potential cash flows to build a DCF model for the ATLAS token. When we explore risk management in Chapter 9: Risk Management, we will analyze the specific operational and smart contract risks Atlas might face. When we explore settlement in Chapter 12: Settlement & Composition, we will detail how Atlas’s A-FCF distributions are executed efficiently. Atlas provides the practical thread that will weave together the diverse theoretical concepts of Agentic Finance into a cohesive and understandable whole.

1.7 Chapter Summary and A Look Ahead

This chapter has laid the conceptual groundwork for the entirety of the book. We began by tracing the parallel, yet distinct, evolutions of Artificial Intelligence and Crypto Finance—one driven by the pursuit of digital autonomy, the other by the creation of decentralized value. We identified the current moment as a critical point of confluence, where the capabilities of autonomous AI agents and the programmatic financial infrastructure of DeFi are poised to merge.

We defined the modern AI agent, moving beyond the passive LLM to a goal-oriented, planning, and tool-using system, while highlighting its fundamental limitation as an “economic ghost” without true, independent economic agency. This led to the introduction of the book’s central thesis: the necessity of re-imagining agents as sovereign, auditable economic entities. We deconstructed this thesis into four essential pillars built upon a blockchain foundation: Sovereign Identity, Verifiable State, Contractual Capacity, and Asset Ownership.

With the agent established as a new type of “Agentic Firm,” we introduced the financial instrument required to capitalize and govern it: the Agentic Value Token (AVT). We defined the AVT as a verifiable claim on an agent’s future economic output, drawing parallels to traditional corporate stock while highlighting its unique, digitally native features of automated profit distribution, decentralized governance, and radical transparency.

The rise of the AI agent economy will reshape the global economic landscape, driving a shift from centralized control to distributed autonomy. The book’s core viewpoint is that, through AVT and related frameworks, we can achieve this transformation, ensuring agents become sustainable, investable economic entities.

Finally, to make these concepts concrete, we introduced Atlas, a hypothetical DeFi asset management agent whose AVT, ATLAS, will serve as our running case study.

The stage is now set. We have established why Agentic Finance is a necessary and revolutionary paradigm. The chapters that follow will build upon this foundation, moving progressively from the conceptual “why” to the practical “how.” In Chapter 2, we will delve deeper into the core concepts of agents as economic entities, formalizing the relationships between agents, humans, organizations, and protocols, and establishing the foundational framework for the rest of our analysis.

[1] McKinsey Global Institute, “The economic potential of generative AI: The next productivity frontier,” June 2023. Available at: https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier