# Five meanings of agent experience, and how to tell them apart. > Agent experience (AX) has five meanings: product AX, coding-agent AX, contact-centre AX, real-estate and insurance AX, and delegated-job AX. How to tell them apart. Published: 2026-10-11 Updated: 2026-10-11 Canonical: https://agentexperience.tech/insights/meanings-of-agent-experience/ Tags: agent experience, foundations, definition, disambiguation **Summary** "Agent experience" (AX) has five live meanings. Three are about AI agents: an AI agent as the user of a product (Mathias Biilmann and Netlify), coding agents choosing and correctly using your SDK, API or CLI (Microsoft's measurement sense), and one job an agent does for a person (this site's sense). Two are about people: contact-centre staff, and estate or insurance agents served by a brokerage or carrier. Read who the agent is and what the unit of work is, and the meaning is usually clear. 1. In AI, agent experience (AX) means the experience an AI agent has as the user of a product, a sense attributed to Mathias Biilmann of Netlify in January 2025. 2. For coding agents, Microsoft's 6 October 2026 definition asks two questions: does the agent choose your technology (propensity), and does it then use it correctly (efficacy)? 3. In contact centres, agent experience means the working day of human support staff, and that usage was in print by 2019. 4. In real estate and insurance, it means how a brokerage or an insurance carrier supports the human agents who sell for it. 5. This site uses a fifth, narrower sense: the quality of one job an AI agent does for a person, from finding the right page to handing the result back. 6. To tell them apart, ask who the agent is, whose interest the page serves, and what unit of work it judges. The same two words describe an AI agent using your API, a coding agent picking your SDK, a support worker's day, an estate agent's broker and an agent doing one job for a person. Each needs different work. **"Agent experience" (AX) has at least five live meanings, and three of them are about AI.** In AI, AX is the experience an AI agent has as the user of a product. For coding agents, it is whether the agent chooses your technology and then uses it correctly. In customer service, real estate and insurance, the same words describe the working life of human agents. > **Scope:** This page sorts the meanings. For the definition this site uses, where the term came from and what the evidence says, read [What is Agent Experience?](/insights/what-is-agent-experience/). All sources below were retrieved on 8 October 2026. ## The five meanings at a glance Each meaning answers a different question. The quickest way to tell them apart is to ask **who the agent is** and **what unit of work is being judged**. | Meaning | Who the agent is | Unit that is judged | Typical users of the term | |---|---|---|---| | **1. Product AX** | An AI agent using a product or platform | The whole product: API, docs, sign-in, errors | Mathias Biilmann and Netlify; API and developer-tool vendors | | **2. Coding-agent AX** | A coding agent in an editor or terminal | Your SDK, API or CLI, task by task | Microsoft's developer blog; developer-tool teams | | **3. Contact-centre AX** | A person who answers customers | One staff member's working day | Contact-centre vendors and analysts | | **4. Real-estate and insurance AX** | A licensed person who sells homes or policies | The support a brokerage or carrier gives them | Brokerages, insurance carriers and their software suppliers | | **5. Delegated-job AX** | An AI agent working for one person | One job, from the person's request to the hand-back | This site | ## 1. Product AX: an AI agent as the user of your product **Who uses it.** Mathias Biilmann, chief executive of Netlify, introduced this sense in early 2025 and defined it as "the holistic experience AI agents will have as the user of a product or platform" ([biilmann.blog](https://biilmann.blog/articles/introducing-ax/)). Netlify's own page states that "Netlify coined the term in 2025", which is a first-party claim ([netlify.com](https://www.netlify.com/agent-experience/)). API vendors, documentation platforms and developer-tool companies have used it since. **Scope.** The whole product as an agent meets it. Netlify's page lists three parts: how well an agent "can discover your API, call it, and recover when something breaks". It separates this from developer experience: "DX optimizes for the humans building on your platform. AX optimizes for the AI agents those humans increasingly send on their behalf." **Example.** A customer's agent tries to integrate with your API and fails without an error anyone sees. Netlify calls that "an AX problem, not a DX one". **How to tell.** The page talks about APIs, documentation, sign-in, MCP servers or `llms.txt`, and about agents as a new kind of user of the product. Netlify measures this sense with AXIS, its open-source scoring tool, which gives one number out of 100 ([axis.run](https://www.axis.run/)). ## 2. Coding-agent AX: does the agent choose you and use you correctly? **Who uses it.** Microsoft's developer blog published "What is Agent Experience (AX)?" by Waldek Mastykarz on 6 October 2026. It defines AX as "the experience AI agents have when discovering, choosing, and using your technology" ([developer.microsoft.com](https://developer.microsoft.com/blog/what-is-agent-experience-ax/)). The post says the idea applies to any AI agent, but its method and examples come from coding agents building with Microsoft's own frameworks. **Scope.** Your SDK, API, CLI, documentation and the extensions you ship to steer the agent, such as skills and MCP servers. The post reduces AX to two questions: - **Propensity:** given a task that does not name your product, does the agent find your technology and choose it? - **Efficacy:** given a task that does name it, does the agent use your current interface correctly, or one you have deprecated? Both are scored for quality and for cost. The post adds a small vocabulary: a **bare baseline** (the same tasks with no extensions), **lift and drag** (better or not better than that baseline), **task cost** and the **agent profile** (the operating system, harness, model and extensions a result was measured with). It also says that "best practices are hypotheses until you measure them". **Example.** Microsoft reports that when a coding agent upgraded projects built on one of its frameworks (five runs in GitHub Copilot Chat with Claude Sonnet 4.6), it passed 30 of 80 configuration checks. Fixing the release notes the agent read while it worked led it to use the right upgrade tool in every later run, without any extra instruction. That is Microsoft's own result, for its scenarios and agent profile. **A related inward sense.** DX, an engineering analytics company, uses "Agent Experience" for something else: at the end of each coding session, the agent scores how well it was set up, on requirements, steering and scope ([getdx.com](https://getdx.com/blog/introducing-agent-experience/), last updated 10 April 2026). There the agent belongs to your own engineer, not your customer, so the work sits closer to developer productivity. **How to tell.** The page names coding agents, an editor or a terminal, and talks about tasks, runs, baselines or token cost. If it asks whether the agent picks your product, it is the outward, Microsoft sense. If it asks how well your engineers' sessions went, it is the inward, DX sense. ## 3. Contact-centre AX: the working day of support staff **Who uses it.** Contact-centre software companies, outsourcers and the analysts who write for them. This sense came first. ICMI, which publishes for contact-centre leaders, wrote about "AX (agent experience)" on 1 September 2022 and pointed back to its 2019 report on the state of agent experience ([icmi.com](https://www.icmi.com/resources/2022/prioritize-agent-experience)). **Scope.** One person's job. Genesys defines it as "the quality of an agent's day-to-day work environment, including the tools, processes and support systems that shape how effectively they serve customers" ([genesys.com](https://www.genesys.com/definitions/what-is-agent-experience-ax)). Typical topics are switching between systems, coaching, schedules and staff turnover. **Example.** A support worker answers a billing call with four applications open: the customer record, a knowledge base, a ticket system and the phone. A contact-centre "agent experience platform" puts all four in one screen ([buwelo.com](https://buwelo.com/technology/agent-enablement/agent-experience-platform/)). **How to tell.** The page mentions calls, chats, handle time, shifts, coaching, attrition or customer satisfaction. AI often appears here too, but as a copilot that helps the human agent, not as the agent itself. ## 4. Real-estate and insurance AX: how a firm supports the people who sell for it **Who uses it.** Property brokerages and insurance carriers, which depend on licensed agents who are often independent. The brokerage Compass, for example, advertises many "Agent Experience Manager" roles within its customer operations team ([compass.com](https://www.compass.com/careers/departments/customer-operations)). In insurance, Glia published a "2024 State of the Insurance Agent Experience Report", a survey of more than 500 independent property and casualty agents about how carriers support them ([glia.com](https://www.glia.com/news/glia-releases-2024-state-of-the-insurance-agent-experience-report), 2 April 2024). **Scope.** The relationship between a firm and the people who sell for it: onboarding, support, tools, commission and the portal the agent uses to quote or file a claim. **Example.** An independent insurance agent wants to check a claim for a client and cannot finish the task in the carrier's portal, so she phones for help. Glia's survey reports, as its own finding, that not completing all tasks within the portal was the agents' top frustration. **How to tell.** The page mentions a brokerage, listings, a carrier, a producer, a portal, licensing or commission. The agent is always a person. ## 5. Delegated-job AX: one job an AI agent does for a person **Who uses it.** This site. It is a narrower version of the first meaning and does not compete with it. **Scope.** One job a person hands to an AI agent, in the person's own words, followed from start to finish: can the agent find the right place, read the facts, take the step safely, recover from a failure, hand the result back, and stay inside what the business allows? The [Open Agent-Readiness Rubric](/insights/agent-readiness-rubric/) checks exactly those six parts and gives a profile, not a total. **Example.** A person asks an assistant to book a bike service and pay a £20 deposit. The site's prices may be easy to read, while the deposit tool can be called with no sign of the person's approval. A product-wide score could average those two facts. A job-level profile reports them separately, with the approval problem first. **How to tell.** The page names a person who delegated the job, and talks about approval, hand-back, undo or a record the person can check. ## How the meanings relate ![Four questions asked in order from top to bottom; a yes gives the meaning, and a no moves to the next question. 1. Is the agent a support worker, on calls or chats? Then contact-centre AX, the staff member's working day. 2. Is the agent a person who sells homes or insurance? Then real-estate and insurance agent experience. 3. Is it about coding agents picking your SDK or API? Then coding-agent AX, Microsoft's sense. 4. Is the unit one job done for a person, up to the hand-back? Then delegated-job AX, this site's sense, highlighted. If every answer is no, product AX, the sense of Biilmann and Netlify.](https://agentexperience.tech/diagrams/meanings-of-agent-experience.svg) *Figure: How to tell which meaning of agent experience a page uses: ask in order, and the first yes is the meaning.* The two human meanings are separate fields with their own vendors and measures. They share nothing with the AI meanings except the word "agent", which in both cases means someone acting for someone else. The three AI meanings are one family seen at different zoom levels. Product AX looks at the whole product. Coding-agent AX looks at how one kind of agent chooses and uses one technology. Delegated-job AX looks at one job for one person. A team can use all three without contradiction, as long as it says which one a given decision rests on. ## Where the meanings conflict Four tensions are worth knowing before you borrow a method from one meaning for another. 1. **Whose interest comes first.** Propensity asks whether the agent picks *your* technology. A person who delegates a job wants the agent to pick the *right* option for them, which may not be yours. Both are fair questions, but they can point in opposite directions. 2. **One number or none.** AXIS gives a score out of 100. Microsoft keeps every result tied to the agent profile it was measured with. This site's rubric refuses a total, so a strong step cannot hide a weak one. These choices are deliberate, so do not compare a number from one approach with a profile from another. 3. **The same product name for opposite things.** In contact centres, an "agent experience platform" is a workspace for human staff. In AI search marketing, Scrunch uses the same phrase for a layer that serves AI agents a separate version of a website prepared for them ([helpcenter.scrunchai.com](https://helpcenter.scrunchai.com/en/articles/13656392-agent-experience-platform-axp)). A buyer who searches for one can easily land on the other. 4. **Outward or inward coding agents.** Microsoft's sense is about agents your customers' developers use. DX's sense is about agents your own engineers use. The fixes differ: better public docs and interfaces for the first, better prompts, repositories and task scoping for the second. ## A worked example **Ledgerline** is a small payroll software company, invented for this page. In one week, three people ask its product lead to "improve our agent experience". - **The head of support** wants fewer tabs on the support team's screens, because staff take longer on calls than they did last year. That is **contact-centre AX**. It needs a better staff workspace, not an API change. - **The developer relations lead** has noticed coding agents setting up Ledgerline's SDK with a method that was removed last spring. That is **coding-agent AX**, an efficacy problem. The first fix is the documentation and changelog the agents read, and the test is to run the same tasks before and after. - **A product manager** has heard from customers who ask an assistant to "run this month's payroll and send me the summary before anything is paid". That is **delegated-job AX**. It needs a clear approval step before money moves, and a summary the customer can check. It is worth reviewing as one job, track by track. The three requests share two words and nothing else. Writing the meaning next to each request turned one vague goal into three pieces of work with three different owners. ## How to tell which one a page means Use the five questions in the diagram above, or look for these clues. | If the page mentions | It probably means | |---|---| | Calls, chats, handle time, shifts, coaching, attrition | Contact-centre AX | | Brokerage, listings, carrier, producer, portal, commission | Real-estate or insurance AX | | Coding agents, SDK, CLI, runs, baseline, token cost, propensity | Coding-agent AX | | Your own engineers' sessions, prompts, repository readiness | DX's inward sense of coding-agent AX | | A person who delegated a task, approval, hand-back, undo | Delegated-job AX | | APIs, docs, sign-in, MCP, `llms.txt`, agents as users | Product AX | When a page uses more than one meaning, it usually says so in its first few paragraphs. If it does not, ask the author before you act on its advice. ## What to do next 1. **Name your meaning.** Write one sentence at the top of your next AX document saying whose experience you mean and what unit of work you are judging. 2. **Pick the matching method.** For coding-agent AX, run the same tasks before and after a change, as described in [how to measure agent experience](/insights/measure-agent-experience/). For delegated-job AX, check one job with [the agent-readiness rubric](/insights/agent-readiness-rubric/). 3. **Read scores with their object in mind.** Before you trust a readiness or AX score, check what it actually scored, using [the comparison of public agent-readiness scores](/insights/agent-readiness-scores-compared/). ## Frequently asked questions **What does agent experience mean?** It depends on who the agent is. In AI and software, agent experience (AX) usually means the experience an AI agent has as the user of a product, a definition attributed to Mathias Biilmann of Netlify. In customer service it means the working day of human contact-centre staff, and in real estate and insurance it means how a brokerage or carrier supports the people who sell for it. Check the words around the phrase before you act on it. **What is AX in AI?** In AI, AX stands for agent experience: how well an AI agent can discover, call and recover inside a product it uses, often on a person's behalf. Netlify defines it as the holistic experience AI agents have as users of a product or platform. Microsoft adds a measurement view built on two questions, propensity and efficacy. **What does AX mean for coding agents?** For coding agents, AX is the experience an agent has when it discovers, chooses and uses your technology, such as an SDK, an API or a CLI. Microsoft's developer blog, on 6 October 2026, splits it into propensity (does the agent pick your technology when the task does not name it?) and efficacy (once it uses your technology, does it use it correctly?), each scored for quality and cost. A separate, inward sense from DX scores how well your own engineers' coding-agent sessions were set up. **What is the agent experience platform?** There is no single one. The phrase names different products depending on the market. In contact centres it means a single workspace for human staff that brings together customer history, knowledge and tools. In AI search marketing, at least one vendor uses it for a layer that serves AI agents a separate version of a website prepared for them. Read the product page to see which kind of agent it is for. **What is AX in software?** In software, AX almost always means agent experience in the AI sense: how well software agents can find, understand and use a product, its documentation, its API and its tools. Some writers also use the letters for other things, such as AI experience design, so check the first use on the page. **What does AX stand for in business?** In customer-service businesses, AX usually stands for agent experience, meaning the tools, processes and support that shape a human support agent's work. In technology businesses it usually means agent experience for AI agents. The letters are also used in unrelated product names, so the surrounding words matter more than the acronym. Name your meaning on first use. A team that agrees on whose experience it is improving finds it much easier to agree on what to change.