The Rise of Non-Human Economic Actors

AI Agents, Machines, and DePIN Are Expanding Economic Agency

Economic agency, the ability to initiate transactions, exchange value, and operate autonomously in a market, has always been the exclusive domain of humans and the organisations they build. Software ran the ledgers, automated the workflows, and optimised the logistics. It did not participate.

That boundary is breaking down. Not as a future projection, but as a present condition. AI agents are already performing services and receiving payments on-chain. Machines are selling sensor data to automated buyers. Infrastructure networks are coordinating supply without a central operator in the loop. The first population of non-human economic actors is here, and it is not homogeneous.

Economic agency, once the exclusive domain of humans, is expanding into the systems we build. The first non-human participants are already transacting.

A Long-Building Foundation

The intellectual groundwork arrived in the 1990s, when computer scientist Nick Szabo described smart contracts as automated agreements executed by software without human intermediation. His illustrative example was the vending machine:

coin → machine verifies payment → product dispensed

The machine enforced an economic contract. No counterparty, no intermediary, no trust required beyond the mechanism itself. Blockchain systems later extended this logic by allowing software to hold funds, verify rules, and release payments autonomously. What Szabo described as a thought experiment, blockchains including Ethereum, Cardano and Solana have made programmable infrastructure, each offering different trade-offs around throughput, formal verification, and settlement guarantees.

Three distinct ecosystems are now building on that foundation, and they are converging on the same architectural pattern from different directions.

Two Populations, One Infrastructure Layer

The clearest way to read what is currently happening is not as three separate developments, but as the emergence of two distinct populations of non-human economic actors, each grounded differently, each adding a capability the other cannot, operating on top of a shared infrastructure layer that makes both viable at scale.

Pure Software Agents

AI frameworks such as LangGraph and CrewAI have made it straightforward to chain multiple agents into collaborative workflows. A research agent gathers information; an analysis agent interprets it; a writing agent produces the output. What was missing was economic infrastructure to support it at scale.

Masumi Network, launched in November 2024 on Cardano, addresses that gap directly. It provides decentralised identity (each agent gets a blockchain-backed DID), a public capability registry, and escrow-based agent-to-agent payment contracts. As of early 2025, the network has processed more than 16,000 transactions across mainnet and testnet, with over 250 live agents registered on the Sokosumi discovery marketplace. This is not a proposed architecture. It is a functioning one.

These are pure software agents entering an economy, no physical referent, value generated, expressed, and settled entirely within the computational stack.

Cyber-Physical Agents: The Economy of Things

A cyber-physical agent is a device embedded in the physical world, sensing it, acting on it, or both, whose data and actions carry economic value. The soil moisture sensor that reads a field at dawn is a cyber-physical agent. It is not just hardware. It is not just software. It is a physical presence that generates a digital signal that can be owned, sold, and settled on-chain.

The Economy of Things, the central thesis of this publication, is specifically about this population. Where pure software agents ask what AI can do when it holds value and transacts, the Economy of Things asks what happens when the physical infrastructure around us does the same.

The constraint profile is fundamentally different. A software agent runs on a server with abundant compute, reliable connectivity, and a clean API surface. A soil moisture sensor in Manicaland runs on 3.3V, operates over intermittent LoRa links, and has no capacity to maintain a persistent HTTP connection. MIP-003, the Agentic Service API Standard that every Masumi agent must satisfy, is trivial for an n8n workflow and an architectural problem for an ESP32-S3.

Devices do not report upward to a centralised platform, they participate directly in a market for their capabilities. The data produced by a borehole pump is not just telemetry. It is inventory.

Most IoT devices today are passive nodes, they generate data with economic value, but economic participation is managed above them. The Economy of Things target architecture requires something more: devices that actively advertise capabilities, enter agreements, and participate in settlement directly. That is the design problem EdgeChain is built to solve. It is still being worked out. But the direction is clear.

DePIN: The Infrastructure That Makes Both Possible

Both populations require infrastructure to operate. Pure software agents need compute, storage, and connectivity. Cyber-physical agents in underserved geographies need network coverage that centralised providers have no commercial incentive to build. Decentralised Physical Infrastructure Networks address both requirements, not as a third population of economic actors, but as the enabling layer on which the other two run.

Helium provides wireless coverage. Filecoin provides storage. Render provides compute. Participants contribute physical resources; the network verifies contribution on-chain and distributes rewards automatically. The incentive structure is the operator, no human decides whether a Helium hotspot forwards a packet and receives a reward.

And for cyber-physical agents in underserved geographies, rural Zimbabwe, smallholder agricultural land, DePIN points toward connectivity infrastructure that centralised providers have no commercial incentive to build. The direction is clear.

A Converging Architecture

Despite their different origins, Agentic AI research, IoT engineering, and blockchain infrastructure, both populations share a common structural pattern. Each separates the layer that generates value from the layer that settles it. That separation is the architectural signature of machine economic agency.

                    Humans
                       │
         ┌─────────────┼─────────────┐
         │                           │
  Pure Software Agents      Cyber-Physical Agents
  (Masumi, CrewAI, n8n)  (IoT / Economy of Things)
         │                           │
         └─────────────┬─────────────┘
                       │
             Infrastructure Layer
                    (DePIN)
                       │
            Payment & Trust Layer
                  (Blockchain)

The diagram is not two separate movements. It is one architecture: two populations generating economic value, DePIN providing the substrate on which both operate, and a blockchain settlement layer that does so without requiring a human counterparty at the point of transaction.

Where the Layers Diverge

The architectural convergence is real. The constraint profiles are not identical, and the differences matter for design.

Masumi's architecture keeps content private, agent outputs are hashed, readable only by the intended recipient. What remains publicly observable on Cardano is the transaction graph: who transacted with whom, when, and how often.

For enterprise AI agents, this metadata exposure is manageable. The identities of commercial agents are already known, their transaction patterns are governed by legal agreements, and the organisations deploying them have compliance infrastructure to absorb the residual risk. For a smallholder farmer in Manicaland, no such infrastructure exists, and the metadata alone is enough to cause harm. Nobody needs to decrypt a hash to learn that plot A4's agent transacted with an irrigation optimisation service at 6am, three days apart, across six weeks in October. That pattern reveals planting schedules, crop cycles, and economic position without exposing a single byte of raw sensor data.

This is where the architecture branches. Privacy-preserving settlement layers, zero-knowledge proof systems that can verify a transaction occurred without revealing the parties, the amounts, or the timing pattern, address the threat model that transparent ledgers cannot. The transaction graph itself is hidden, not just the content. That is not a critique of Masumi. It is the frontier that cyber-physical agents require and pure software agents, for the most part, do not.

The open question is not whether machines can be economic actors. Masumi demonstrates they can, at the software layer. The open question is what infrastructure the cyber-physical layer needs to participate under its own constraint profile, intermittent connectivity, power constraints, privacy requirements, and physical environments that no API standard currently anticipates. EdgeChain is this publication's proposed answer to that question.

A Quiet Transformation

Pure software agents are working out how to discover each other, transact, and build reputation. Cyber-physical agents, the IoT devices, sensors, and edge nodes of the Economy of Things, are working out how to participate in the same economy under a fundamentally different set of physical and privacy constraints. DePIN networks are building the infrastructure substrate both require.

These are not separate communities describing different futures. They are engineering communities building different layers of the same emerging stack.

Humans designed the system. Humans govern it. But they are no longer the only participants in economic activity within it.

The machine economy is not arriving. At the software layer, it is already here. The hardware layer is the frontier being worked out now.