# What is the agentic economy?

A market definition of assistants, automation, agents and delegated work—and the boundary where software begins carrying a job forward.

Source: https://aecon.ai/guides/what-is-the-agentic-economy
Author: Agentic Economy
Published: 2026-09-13
Updated: 2026-09-14

**Chapter 1 of The Hitchhiker's Guide to the Agentic Economy**

[Report hub](/guides/hitchhikers-guide-agentic-economy-2026) · In this chapter: [category tiers](#section-a-practical-vocabulary) · [delegated-work model](#section-agentic-economys-delegated-work-model) · [reference job](#section-reference-job-ae-job-001-a-competitor-briefing) · [bounded job](#section-the-bounded-job-is-the-markets-useful-unit)

## Our view

The shift becomes agentic when software can choose among permitted steps and carry a job forward for a person or organisation. The action may be small: call a tool, update a system, ask another service for a quote, place an order or route an exception to a person. What matters is not independence. It is a real decision or action inside a stated boundary, with enough evidence for the result to be checked.

An **AI agent** is a software system that uses a model or learned policy to select and carry out permitted steps towards an objective. **Delegated work** is the bounded job that a principal authorises that system to advance.

That boundary keeps ordinary automation, model-assisted workflows and agents comparable without pretending they are the same. The broader agentic economy includes all of them when they support production or exchange around delegated software work. The business questions remain concrete: what job moved, which choices were delegated, what changed and who remains accountable?

> **Key Takeaways**
>
> - Work becomes agentic when software selects among permitted actions and changes a tool, service or business process.
> - The practical unit is `objective + context + authority + action + evidence + accountability`.
> - Autonomy is a property of a particular task and boundary, not a permanent label for a product or company.
> - A fixed workflow and a model-directed agent can both create value. The right design is the least complex system that can perform the job dependably.
> - An accepted outcome is a useful operational endpoint; value and margin still require a baseline, costs and repeat demand.

![A controlled launch apron with a clear path, representing bounded software action](/media/editorial-library/launch-apron-full.webp)

### Sources and scope

Agentic Economy developed the definitions and operating model in this chapter by comparing primary technical guidance with observable business states. Sources are linked at the claim they support. The model, job card and worked offer are Agentic Economy analysis dated 13 September 2026, not an external standard. [About Agentic Economy](/about) explains the publication. Readers can submit corrections through the [contact page](/contact).

## The category is forming around delegated work

The market is moving beyond assistants that produce answers towards systems that advance work. Model providers are adding tool use and computer control. Software platforms are embedding agents inside existing applications. Framework companies are building orchestration and memory layers. Service providers are beginning to package completed tasks rather than access to software alone.

OpenAI states the category in operational terms: agents are “systems that independently accomplish tasks on behalf of users” ([OpenAI, "New tools for building agents"](https://openai.com/index/new-tools-for-building-agents/)). The important word is *tasks*. The market is moving from generated material towards completed work, even though providers draw the autonomy boundary in different places.

We see the common unit as delegated work: a defined objective, the context and authority available to the system, the actions it takes, the evidence it leaves and the person or organisation that remains accountable. Products implement that unit differently, but it gives the market a clearer boundary than the word *agent* by itself.

## From an answer to an action

Software has acted without a person at the keyboard for decades. Scheduled jobs, rules engines, webhooks and robotic process automation can all update systems automatically. The newer distinction is not action versus passivity. It is whether a model or learned policy can select among permitted actions while pursuing an objective.

Generative AI made the interface conversational. An assistant can summarise a contract, draft a campaign or suggest code while a person chooses the consequential next step.

Agentic systems move that boundary. The system may choose a next step, select a tool, inspect the result and continue until it reaches a stopping condition. An accounts-receivable agent might identify an overdue invoice, check the account policy, draft an appropriate message, record the contact and schedule a follow-up. A purchasing agent might compare approved suppliers, request a quote and prepare an order for human approval. A research agent might commission a bounded data service, validate the returned file and add it to a briefing.

These systems differ in complexity, but each advances a real process. The business is delegating some combination of selection, sequencing and execution.

Anthropic distinguishes workflows, where code predetermines the path, from agents, where a model dynamically directs its own process and tool use. Its engineering guidance also recommends using simple, composable patterns when they are sufficient ([Anthropic, "Building effective agents"](https://www.anthropic.com/engineering/building-effective-agents)). That distinction is useful for design without turning the market into a contest over which product deserves the word agent. A fixed workflow with a clear control surface may be the stronger commercial system.

<!-- ORIGINAL ANALYSIS: Agentic Economy developed the six-field Delegated Work Model and dated it 13 September 2026. -->

## Agentic Economy's Delegated Work Model

Agentic Economy's working model follows one unit of delegated work. Version 1.0 of the model, dated 13 September 2026, has six fields:

```text
objective + context + authority + action + evidence + accountability
```

### Objective

The objective states the job and its stopping condition. "Help with sales" is too broad. "Prepare a sourced briefing on this prospect before tomorrow's meeting" gives the system a deliverable, subject and deadline.

### Context

Context supplies the information the system needs: the customer's request, internal records, current policies, product data, prior actions and relevant external sources. Context needs a freshness and provenance boundary. An agent acting on an obsolete price list or an untrusted instruction can execute the wrong decision perfectly.

### Authority

Authority defines what the system may see, decide, change and spend. It links the action to a principal, such as a person or organisation, and constrains the agent by role, resource, amount, supplier, time or approval condition. Authentication proves which credential is present. Authority answers the harder question: is this credential allowed to take this action now?

### Action

The action changes something outside the model. It may call an API, use a browser, send a message, create a record, run code, request another service or initiate a transaction. Each action creates side effects and therefore needs controls for retries, duplication and failure.

### Evidence

Evidence records what the system received, decided, called and produced. Useful evidence is more than a chat transcript. It connects the objective, source data, policy, tool response, delivered result, approval and exception so the business can reconstruct the job.

### Accountability

Accountability establishes who owns the outcome and how a problem is corrected. A person, team, supplier or institution still carries the obligation to monitor, accept, reject, refund, repair or explain the result. Delegation changes how work is performed. It does not make responsibility disappear.

For consequential work, name roles rather than writing “the business is accountable”:

| Role | Decision owned |
| --- | --- |
| **Business owner** | Job purpose, benefit, customer promise and operating result |
| **Control owner** | Authority, data, payment, evaluation or recovery control |
| **Independent challenger** | Evidence review outside the delivery team where consequence requires it |
| **Risk-acceptance authority** | Residual risk, exception duration and conditions for continued use |
| **Incident owner** | Containment, notification, correction, restoration and customer remedy |

One person may hold several roles in a small firm. The decisions and evidence remain explicit.

The model joins technology to economics. A model provider supplies capability. A runtime manages state. A tool or service performs an action. An identity or policy layer limits authority. A payment system transfers value. An enterprise system records the commercial event. A human or organisation remains responsible for the customer outcome.

## A practical vocabulary

The market uses overlapping terms. The following distinctions make products and claims comparable.

| Term | Working meaning in this report | The decision boundary |
| --- | --- | --- |
| **AI assistant** | Produces information or a recommendation for a person | A person chooses and performs the next consequential action. |
| **Automation** | Executes predetermined rules or steps | Code fixes most of the path in advance. |
| **AI workflow** | Combines models and deterministic steps in a defined process | The workflow may branch, but the system's range is designed explicitly. |
| **AI agent** | Chooses and carries out some steps towards an objective using permitted tools and information | The model can select among actions inside a boundary. |
| **Multi-agent system** | Uses several specialised agents or agent roles to divide and coordinate work | Coordination and hand-off state become part of the system. |
| **Agentic commerce** | An agentic system crosses a buyer-supplier or payment boundary during discovery, ordering, delivery or support | The report names which commercial states and authority the system actually handles. |
| **Agentic economy** | The broader system of production, coordination and exchange that forms as software takes on bounded work | Includes infrastructure, firms and institutions around the action, not just the agent product. |

None of these forms is automatically more mature. A deterministic automation can outperform a flexible agent on a stable, high-volume task. A human-approved assistant can be the appropriate design where judgement, regulation or consequence requires review. The relevant questions are whether the system performs the intended job, how often it fails, what intervention it needs and whether the economics improve.

## Autonomy belongs to the task

Calling a product autonomous hides the design choices that matter. The same system may work unattended for one task, require approval for another and have no authority at all in a third.

Autonomy can vary across at least five dimensions:

| Dimension | Lower delegation | Higher delegation |
| --- | --- | --- |
| **Planning** | Person supplies each step | System selects and revises the path |
| **Tool choice** | Fixed tool and parameters | System selects tools and parameters |
| **External action** | Draft or recommendation only | System changes records or communicates externally |
| **Spending** | No payment authority | Bounded or policy-controlled spend |
| **Review** | Every action approved | Exceptions or samples reviewed |

Anthropic's research on autonomy in deployed systems warns that observed behaviour depends on product design, task structure and human oversight, and that product telemetry may not see downstream review ([Anthropic, "Measuring AI agent autonomy in practice"](https://www.anthropic.com/research/measuring-agent-autonomy)). The implication is operational: measure the task, permitted action set and review condition. Do not infer a whole organisation's autonomy from a session length or product label.

Longer operation also changes risk. A system that can make one low-value request has a small failure surface. A system that can run for hours, call several tools, modify records and spend money accumulates state and consequence. Reliability at each step compounds across the path. The organisation needs checkpoints, budgets, timeouts, idempotency, escalation and a way to resume or reverse work.

## The agent is a system, not a model

A capable model is necessary for many agentic tasks, but the operating system around it determines whether the work survives contact with a business.

| Layer | Job in the system | Typical failure if it is missing |
| --- | --- | --- |
| **Model and inference** | Interpret objectives, reason and choose steps | The system cannot handle ambiguity or adapt its plan. |
| **Context, memory and data** | Supply current information and retain relevant state | The system acts on missing, stale or unauthorised information. |
| **Runtime and orchestration** | Manage execution, state, retries, hand-offs and long-running work | The process stalls, repeats or loses state. |
| **Tools and service access** | Perform actions in software and the world | The agent can recommend but cannot complete the job. |
| **Identity, authority and policy** | Establish the principal and constrain permitted action | The system overreaches or cannot prove permission. |
| **Evaluation and observability** | Test quality and record what happened | Failures remain invisible or cannot be explained. |
| **Security and recovery** | Contain hostile input, data loss and operational failure | A small error becomes a business incident. |

An **agentic AI harness** is the wider software environment that gives a model tools, memory, orchestration and controls. The Australian Signals Directorate uses this term in guidance focused on risks around the model, including tool permissions, untrusted content, data access and monitoring ([ASD, "Agentic AI harnesses"](https://www.cyber.gov.au/business-government/secure-design/artificial-intelligence/agentic-ai-harnesses)). This system view is also the right commercial view. A customer buys a dependable result, not a model in isolation.

## Where the economic change appears

Delegated action alters four familiar parts of a business.

### The interface becomes an objective

People used to learn the software's menus and forms. An agent can receive an objective and operate several systems on the person's behalf. The interface shifts from "which button do I press?" to "what outcome do I want, and what authority should the system receive?"

### The product can become a completed task

Software businesses commonly charge for access, seats or consumption. When software performs more of the work, suppliers can also package an outcome: a reconciled account, a qualified lead list, a checked document or a resolved support request. The commercial unit can move closer to a job, although outcome pricing only works when the outcome and responsibility are clear.

### Distribution can become machine-mediated

Agents may discover tools, products and services through registries, protocols, assistants, marketplaces or direct APIs. A supplier's offer must be legible to software: who it is for, what inputs it accepts, what it returns, how much it costs and what happens when delivery fails. Reachability does not guarantee consideration. Ranking, trust and evidence still determine whether the supplier enters the shortlist.

### Management becomes authority design

Managers already decide who can approve a purchase, contact a customer or change a record. Agentic systems make those decisions executable in policy and software. The work includes defining budgets, permissions, review thresholds, evidence requirements and escalation owners. This is closer to operating-model design than prompt writing.

## Reference job AE-JOB-001: a competitor briefing

Consider an illustrative Perth advisory firm that repeatedly prepares competitor briefings for clients. This synthetic reference job is `AE-JOB-001`. It connects the report's chapters, but it is not evidence of an actual deployment or market price. The firm wants to make its expertise available to a customer's software while retaining professional review.

| Element | Design |
| --- | --- |
| **Objective** | Produce a sourced briefing on a named company and market by a stated deadline. |
| **Context** | Client request, approved source classes, geography, prior briefings and confidentiality rules. |
| **Authority** | Access only to the submitted client material; spend capped for external data; publication prohibited. |
| **Action** | Search approved sources, buy a bounded data input if needed, draft the briefing and route it to an analyst. |
| **Evidence** | Source links, retrieval dates, data-service receipt, change log and analyst decision. |
| **Accountability** | The advisory firm owns accuracy, delivery and correction; the analyst accepts or rejects the draft. |

The firm turns that design into a buyer-facing offer:

| Offer field | Illustrative term |
| --- | --- |
| **Job** | One sourced competitor briefing using up to ten approved public sources |
| **Price** | A$750 including GST |
| **Delivery** | One business day after a complete brief is accepted |
| **Review** | Analyst checks sources, unsupported claims and the final recommendation |
| **Variable spend** | External research purchases capped at A$25 unless the buyer approves more |
| **Acceptance** | Required sections present; every material factual claim sourced; no prohibited client data used |
| **Remedy** | Rejected work corrected within two business days or the fee refunded |
| **Data boundary** | Client material retained only for the agreed delivery and correction period |

These are example terms, not a market-price benchmark or tax advice. Their purpose is to expose the operating decisions hidden by the phrase *AI service*. If delivery consumes A$18 of model and data usage plus 45 minutes of analyst time, the firm can calculate contribution margin. If analyst review expands to three hours, it has found a workflow or pricing problem even when the model produces a polished draft.

The agentic part is not the presence of a chatbot. It is the controlled movement of the job through research, service access, drafting and review. The commercial offer remains familiar: a competitor briefing with a delivery promise. The software makes the offer easier to order and more repeatable to deliver.

The example also shows why the full market matters. The firm may use a model provider, search and data services, an orchestration framework, a payment credential, an accounting system and an observability product. Each participant supplies one layer. The customer experiences one job.

NIST describes the leading agent pattern as a general-purpose model embedded in software scaffolding that lets it manipulate tools beyond text output. Its 2025 work also argues that tool capabilities and constraints should be communicated across the supply chain ([NIST, "Tool Use in Agent Systems"](https://www.nist.gov/news-events/news/2025/08/lessons-learned-consortium-tool-use-agent-systems)). This supports evaluating the whole operating system, not only the underlying model.

## What counts as completion

<!-- UNIQUE INSIGHT: This chapter defines accepted work as the operational bridge between technical execution and commercial value. -->

The weakest measure of agentic work is that the model produced a response. A stronger sequence follows the job:

1. The system received a valid objective.
2. It used permitted context and authority.
3. The required action executed without an unresolved duplicate or error.
4. The promised output or state change was delivered.
5. The customer or responsible system accepted the result.
6. Commercial and operational records reconciled.
7. The customer returned or expanded use.

This sequence matters because the market's easiest events to count are intermediate. A tool invocation records execution. A payment records value movement. A delivery response records receipt. Acceptance records that the responsible buyer or system judged the promised work complete. Margin and durable value still require costs, incidents, rework and repeat use.

## The bounded job is the market's useful unit

Agentic Economy describes a delegated job through eleven fields. Together they separate a working commercial system from a model demonstration:

| Field | What it establishes |
| --- | --- |
| Delegated job | The exact result the system is expected to produce |
| Economic actors | Who requests, pays for and accepts the work |
| Discretion | Which choices the software can make |
| Reach | Which data, systems and spend it can access |
| Completion evidence | What proves delivery and acceptance |
| Exception ownership | Who reviews and remedies failure |
| Cost base | Which costs determine whether the job is viable |
| Baseline | Current volume, labour time, error rate and full cost |
| Investment | Implementation, security and change cost to recover |
| Return boundary | Benefit, maximum loss and payback rule |
| Decision rights | Who owns funding, expansion, redesign and termination |

Different interface markets address different parts of that boundary:

| Need | Starting point |
| --- | --- |
| Call one defined operation | API |
| Connect an AI application to tools or context | MCP |
| Delegate a task to another independently operated agent | A2A or a product-specific task interface |
| Express product, checkout or order state | A commerce protocol or merchant API |
| Initiate or settle payment | The rail appropriate to value, frequency, geography and recourse |
| Preserve judgement as part of the product | A human-reviewed service workflow |

The market will increasingly be judged against the prior process rather than a model benchmark. Accepted-job rate, human effort, full cost and incident severity determine whether a delegated job expands. Rework, intervention and hard data or authority failures determine where it contracts.

The card is not a business case until the financial rows are filled. For `AE-JOB-001`, A$750 is an illustrative offer price, not evidence that the job clears a return threshold. A fundable case still needs observed demand, delivery volume, implementation cost, fully loaded labour, accepted-job cost, incident exposure and a payback rule.

Material controls have the same recurring anatomy: an owner, implementation point, test, expected result, threshold, review cadence, retained evidence, exception authority and escalation path. A design record describes intent. A passing test establishes performance in a stated environment.

The [control and capability analysis](/blog/control-and-capability) places approval and authority around a bounded job.

## Where we see the market going

The word *agent* will become less useful as the market separates into clearer product categories: assistants that prepare work, orchestration systems that coordinate it, operator platforms that execute it and services that sell a completed result. The durable category boundary will be the delegated job, not the model interface.

This shifts competition towards the systems around the model. Providers that can state the work, authority, evidence, acceptance and remedy will be able to sell business capability. Providers that expose only a conversational surface will remain easier to compare as software features.

## About the author and editorial record

Agentic Economy is the accountable publisher of this report. Joel Chan founded the publication in Perth to research the infrastructure, companies and operating choices shaping the agentic economy in Australia. This chapter provides source-led analysis and an original operating model. It is not legal, financial or procurement advice.

Next: [Demand, adoption and the measurement problem](/guides/ai-agent-adoption-measurement).