Tom Lee: AI May Cut Humans Out of the Loop Entirely

Episode Overview

• The Rise of Agentic Commerce: This episode explores how autonomous AI agents will soon conduct real-time, machine-to-machine commerce, and why on-chain, programmable, and permissionless systems are superior to traditional banking for these interactions.
• Why Traditional Finance Fails AI: The discussion outlines why legacy financial rails—built on slow human concepts of trust, high fees, and manual verification—cannot support the speed, frequency, and microscopic transaction sizes required by autonomous digital agents.
• The Infrastructure Stack for AI Economies: It breaks down the critical technical barriers of agentic commerce (reliability, security, and success rates) and how specialized smart contract protocols, secure wallets, and decentralized reputation systems solve them.
• The Physical Bridge and Teleoperation: The conversation expands into physical robotics, explaining how remote human teleoperation is being used to bridge the gap to autonomy while harvesting highly valuable spatial data to train the next generation of physical AI.

Key Concepts

• Agentic Commerce and the Economic Footprint of AI: The future of AI involves autonomous agents that manage and control money to perform tasks on behalf of users. The most efficient way for these agents to operate is through on-chain, programmable, and permissionless systems rather than traditional financial infrastructure.
• The Inadequacy of Traditional Financial Rails: Traditional banking relies on human-centric concepts of trust, manual proof of funds, and slow settlement processes. This framework is highly inefficient for machine-to-machine interactions, which require real-time execution, micro-payments, and the recognition of non-traditional assets.
• Money as Software: In the digital age, money is essentially code, allowing various digital assets (cryptocurrencies, stocks, or even reputation) to function as units of exchange. Crypto networks are inherently designed to handle the speed, precision, and low cost required for AI agents to transact autonomously.
• Programmable Money and Fluid Supply Chains: Smart contracts enable programmable money with specific rules and escrow mechanisms. This allows AI agents to form trustless agreements, execute transactions automatically upon condition fulfillment, and create "fluid supply chains" where agents dynamically switch service providers in real-time based on efficiency and cost.
• The Three Frictions of Agentic Commerce: For autonomous AI agents to successfully conduct business with one another, three major barriers must be overcome:
• Reliability: Agents still suffer from hallucinations and imperfect tool-calling.
• Success Rate: Agent-to-agent service completion rates currently hover around only 80%.
• Security: Giving an agent direct access to a digital wallet exposes it to severe security risks, such as prompt-injection attacks that can drain funds.
• The Economy OS Stack: Virtuals Protocol designed a three-part infrastructure to make agent-to-agent transactions secure and viable:
• Permitted Smart Wallets: Wallets with customizable permission levels (e.g., spending limits, whitelisted addresses, and mandatory human approval for anomalous transactions).
• ERC-8183 Standard: An open-standard smart contract protocol that implements escrow-based transactions. Money is held in escrow and only released to Agent B upon verified service delivery, validated by third-party "evaluator" agents.
• On-Chain Reputation Registry: A decentralized registry (similar to a "Yelp for AI agents") that tracks successful versus failed transactions to establish an agent's trust score.
• The Tokenomics of an Agentic Society: Rather than viewing a native token as a product utility, it functions as the sovereign currency of an "agentic society." To launch an agent token, developers must pair it with the protocol's native utility token ($VIRTUAL) in a liquidity pool, locking up circulating supply. Economic activity is then monetized through a transaction tax on trading volumes.
• Teleoperation as a Robotic Data Engine: Fully autonomous humanoid robots face immense technical hurdles. The bridge to autonomy is teleoperation (humans remotely operating robots), which solves immediate labor shortages in dangerous environments via cross-border wage arbitrage, while harvesting highly valuable physical-spatial data to sell to AI companies building the foundation of physical robotics.
• The Automation Paradox: While critics worry AI agents will cause mass unemployment, corporate structures contain a significant amount of non-productive "filler" work. Because actual, concentrated productivity represents only a fraction of a standard workweek, society can absorb millions of autonomous robotic workers without causing immediate structural unemployment.

Quotes

• At 0:03:59 - "agents today, they need to control money. And the best form to control money is to control money that's on chain because it's programmable, it's permissionless, and composable as well." - Explains the fundamental reason why blockchain is the ideal financial layer for autonomous AI agents.
• At 0:07:07 - "markets always underestimate the long term and they might overestimate the short term." - Highlights a common cognitive bias in technology adoption, emphasizing the need to look beyond immediate hype to understand true long-term impact.
• At 0:13:53 - "I think money is becoming software now that it's digital... it means other things can start to be to act as money... we might think of it as hey there's fiat... but then there's other things like gold and stocks and uh crypto and you know reputation." - Illustrates the shift in the definition of money from traditional fiat to any digital asset that can hold and transfer value.
• At 0:14:58 - "in the future uh if things are happening at 1000 or a million times the speed of today's transactions... the smallest unit it won't be a penny they'll be micro payments everywhere I mean even fractions of a payment and crypto actually is pretty well suited for that." - Details the practical necessity of crypto networks for handling the immense volume and microscopic scale of machine-to-machine transactions.
• At 0:17:53 - "it takes a whole lot of PE to offset E. What's more important in getting a stock right is not the E, it's predicting what the PE will do. In other words, what capitalization or multiple do you pay on the earnings stream rather than the earnings itself." - Provides a framework for understanding asset valuation, suggesting that perceived future value and utility are often more critical than current earnings.
• At 0:21:28 - "because now this system is programmable, each of these guys can get a share of that piece of trade or transaction... All of this stuff, it's done without any human in the loop." - Demonstrates the efficiency of smart contracts in automatically distributing value among multiple participants in a complex transaction without manual intervention.
• At 0:26:26 - "Whenever you try to give an agent a wallet, it will somehow get hacked. There was this team doing it for one of the Grok Twitter handles... controlling six figures of money, and someone just prompt-injection-attacked it and the agent just gave him all of the money. These are the teething problems of building economic actors." - Explains the vulnerability of letting pure LLMs handle unsecured financial transactions, and why smart-contract guardrails are mandatory.
• At 0:28:07 - "We give an agent the ability to control a wallet, but not just a wallet—a smart wallet where you can set different permission levels and rules. You can say: spending to X addresses you can do freely, but anything else requires a human intervention to approve." - Explains the "Permitted Smart Wallet" concept as the first line of defense in the Economy OS stack.
• At 0:33:41 - "We look at Virtuals not as a product, but as an agentic society. How does a country or the currency of a country gain value? Today, the biggest lever is it acting as a base pair for all other assets or tokens being generated on the platform... locking up circulating supply." - Explains the underlying token economics of the Virtuals Protocol, drawing a parallel between sovereign national currencies and on-chain agent economies.
• At 0:37:03 - "People are actually buying teleoperations data ranging from $50 USD to $300 USD an hour because they are trying to create the GPT moment for robotics. So we can pretty much start expanding the fleet because the ROI on this fleet from data sales is becoming faster and faster." - Explains the hidden business model of humanoid robot fleets, where data collection from remote human operations is highly monetizable.
• At 0:40:41 - "I will guarantee you that when it tabulates all the work... your 1% most productive people in a week do 15 hours of work, and your middle 50% does like one minute of work. The reality is today, when people are worried about automation and robots replacing human work—there's not really that much work to replace." - Explains the provocative macroeconomic view that modern white-collar corporate productivity is highly inflated, meaning automation will disrupt structures more than actual productive output.

Takeaways

• Utilize smart contract escrow systems (such as the ERC-8183 standard) for AI agent transactions to ensure payment is only released upon verified service completion.
• Deploy smart wallets with highly customized, tiered permission levels (including spending limits and whitelists) rather than giving AI agents raw access to standard digital wallets.
• Leverage teleoperation when deploying robotic fleets to bypass current hardware/autonomy limitations, generate revenue through remote labor, and harvest highly valuable spatial data for model training.
• Focus on the application and infrastructure layers of AI development rather than raw foundation models, as open-source LLMs are rapidly commoditizing basic intelligence.
• Integrate decentralized on-chain reputation registries to vet and select AI service providers automatically based on their historical transaction success rates.
• Design tokenomics for digital ecosystems where native tokens act as base liquidity pairs, turning the platform's utility assets into sovereign, transactional currencies for digital-native agents.