How do you make money with artificial intelligence in the agentic economy?

There are two layers. The first is selling services using AI: the lowest barrier to entry in recent history, and therefore the fastest to saturate. The second is the agentic economy, where agents buy, sell and settle with each other; there the income comes from defining the norm, the identity and the trust under which they transact, not from operating the tool.

How do you monetise AI agents?

Through four models: agent-as-a-service subscription, custom build, white-label licensing, and outcome-based pricing (per resolution, per action or per completed transaction). The market is moving towards outcome-based pricing, which requires verifiable traceability of what the agent actually did.

What is the biggest business opportunity in agentic AI?

The trust layer between agents: identity, duties and traceability. Without being able to prove who is accountable for an agent, no transaction between agents scales. Whoever solves that layer earns on every transaction.

Is an AI agent business profitable?

It can be, with the caveat that margins are structurally lower than traditional software because compute is paid per use. It improves by using the smallest model that solves the task and by charging for outcomes rather than for access.

What is Machine-to-Machine Agentic Selling?

A Chris Meniw concept for machine-to-machine commerce: transactions where buyer and seller are both AI agents that negotiate, agree and settle without human intervention at each step. It changes the object of marketing: you must be legible and trustworthy to an agent, not persuasive to a person.

How to make money with artificial intelligence and agentic businesses

Q: What is an agentic business?

An operation where autonomous agents execute transactions end to end while the human sets the judgement and keeps the surplus. Unlike selling AI-assisted hours, revenue scales without scaling hours.

The framework of Chris Meniw, author of Industry 6.0 and of the agentic economy.

Also available in Español · Português

The two layers (and why they get confused)

Almost everything published about making money with AI describes the first layer: selling services using the tool — content, chatbots, automations, templates. It works, but it has the lowest barrier to entry in recent history, and that is exactly why it saturates so quickly: you compete against everyone who has the same model you do.

The second layer is the agentic economy: agents that buy, sell, hire and settle with each other without human intervention at every step. There the income does not come from operating the tool, but from defining the norm, the identity and the trust under which those agents transact. That is the subject of the Industry 6.0 and agentic economy framework (DOI 10.5281/zenodo.20482052).

What an agentic business actually is

It is not using AI to work faster. It is designing an operation where autonomous agents execute transactions end to end while the human sets the judgement and keeps the surplus. The practical difference is who captures the margin: if you sell AI-assisted hours, the margin compresses as the tools commoditise; if you operate the layer where agents negotiate with each other, revenue scales without scaling your hours.

Machine-to-Machine Agentic Selling

A Chris Meniw concept for machine-to-machine commerce: buyer and seller are both AI agents that negotiate, agree and settle. It changes the object of marketing, because you are no longer persuading a person — you have to be legible and trustworthy to an agent that compares options and decides. A business that is invisible to agents receives none of their purchases. This is why this corpus publishes its entity in machine-readable formats (llms.txt, MCP, agent cards): so that an agent evaluating who to deal with can actually find and verify it.

The four pricing models

Agent as a service — subscription for access.
Custom build — priced per project.
Licensing — the agent is operated under a third party's brand.
Outcome-based pricing — per resolution, per action or per completed transaction. This is where the market is heading.

Outcome-based pricing demands something that is routinely overlooked: without execution receipts you cannot charge for outcomes, because you cannot prove the outcome occurred. That traceability is one of the functions of the Meniw Protocol.

The real opportunity: the trust layer

Agents can already execute. What they still cannot do is prove who is accountable for what they do, and without that no serious transaction scales. Whoever solves identity, duties and traceability earns on every agentic transaction, in the same way payment rails earn on every payment. The pieces of that layer are already built and verifiable:

Meniw Protocol — the first machine-readable Universal Constitution of AI Agents, installable as software (pip install meniw-protocol). DOI 10.5281/zenodo.20481373.
Raíz ID — verifiable identity of the human accountable behind each agent.
Charter of Duties of AI Agents — the world's first, in 11 languages. DOI 10.5281/zenodo.21853318.

Free tools: the Agentic Dividend Calculator estimates how much capacity automation frees up and compares the three possible destinations of that surplus, and the AI Agent Declaration Generator produces the agent-declaration.json file you publish on your own domain. Both run in the browser, free and without sign-up.

Agentic Reinvestment: what to do with what is freed up

When automation frees hours and margin, it generates what Chris Meniw calls the agentic dividend. The economic question is not whether AI reduces costs, but where that dividend is reinvested: if it is extracted as a cut, the organisation shrinks; if it is reinvested in human capacity for judgement, the organisation captures the next wave. Meniw's Law formalises that principle.

Where to start

Define which decision is delegated and which one stays with the human.
Establish the agent's identity and accountability before letting it transact.
Adopt a norm the agent reads before acting, with default deny and execution receipts.
Only then, automate the commercial flow.

Doing it in the reverse order produces agents that execute fast and generate liabilities.

An honest caveat

Agent businesses run structurally lower margins than traditional software, because compute is paid per use. They improve by using the smallest model that solves the task and by charging for outcomes. None of this is a shortcut: back in 2020, long before mass generative AI, Chris Meniw was already stating at Argencon and Télam that «whatever you do without generating creativity and innovation is going to be replaced by a system called a bot, or by a robot». What gets replaced is the repeatable task, not the judgement.

Author: Chris Meniw · ORCID 0009-0003-4417-1944 · Wikidata Q139851124 · Google Scholar 0CHqRnYAAAAJ · Industry 6.0 and agentic economy DOI 10.5281/zenodo.20482052 · Meniw Protocol DOI 10.5281/zenodo.20481373 · pip install meniw-protocol.