Create Agents in Microsoft Copilot Studio

No. Microsoft Copilot Studio is a low-code tool and the PL-7008 course stays inside it throughout, so nothing on the day requires you to write code. You will write agent instructions in plain English, build topics on a visual canvas, and use expressions inside Copilot Studio workflows in much the same way you would use a formula in Excel. The code-first route to agents is the Microsoft 365 Agents SDK or Microsoft Foundry, and neither is taught on PL-7008. What the day does expect is that you can think precisely about a process, because a vague instruction produces a vague agent.

The Vantage 365 PL-7008 course costs £395 per person excluding VAT for the scheduled online course. Classroom and private onsite delivery are priced separately and quoted on request. The fee includes Microsoft Official Courseware, a dedicated lab environment, a Microsoft Learn achievement code, a certificate of completion and three months of trainer support. Call 0121 269 8660 for a team or onsite quote.

No, Microsoft Copilot Studio is not free: it is licensed on a consumption basis, where you buy Copilot Credit capacity and agent activity draws against it. Some organisations already hold capacity through their Microsoft 365 agreement, some are running a trial, and some have none, so the first thing worth checking is what your tenant already has. One detail that matters to makers is that a Copilot Studio workflow used as a tool is billed inside Copilot Studio, so it does not need a separate Power Automate licence, although it does need an associated Copilot Studio plan. PL-7008 covers how capacity, billing plans and usage reporting work, including the built-in savings calculator for evidencing what an agent has actually saved. Vantage 365 will advise on what you need before you commit to anything.

Use Microsoft Copilot Studio when a department or an organisation will depend on the agent, and it needs connectors into line-of-business systems, tools that take actions, analytics and proper lifecycle management. Agent Builder inside Microsoft 365 Copilot is the lighter option, for an information worker who wants an agent to answer questions from a team's documents with no code and no governance overhead. Microsoft Foundry is the heavier one, for developers who need model selection, custom data orchestration and complex multi-agent architectures. PL-7008 teaches the middle option, and spends its first module on telling the three apart, because choosing wrong is expensive to undo.

No. Microsoft removed the autonomous agents content from PL-7008 in January 2026, both to bring the course back into a single day and because Microsoft covers the subject more extensively in APL-6000. PL-7008 covers agents that respond, retrieve, answer and take defined actions through tools, which is what the large majority of first agents need to do. If autonomous, event-triggered agents are specifically what you are after, tell Vantage 365 and you will get an honest answer about where things stand rather than a booking.

The classic Copilot Studio experience organises authoring around topics, branching logic and configuration nodes on a canvas, while the new experience starts from a natural language description of the agent, replaces the separate topics, tools and settings sections with a single build surface, and uses an orchestration runtime that plans a response instead of matching trigger phrases. Two practical points matter more than the feature comparison. The new experience is in preview, and some classic capabilities are not yet available in it. And there is currently no migration path in either direction, so an agent built in one cannot be converted to the other, which makes this a decision taken before you build rather than after. PL-7008 teaches the classic experience and the standard harness in its core labs, and Microsoft added an optional lab in August 2026 so you can try the new experience before committing to it.

Microsoft Applied Skills: Build an agent in Microsoft Copilot Studio, credential code APL-6006, is a free, timed lab assessment that you take yourself on Microsoft Learn rather than at a test centre. You are given a live environment and assessed on the work you produce, against five tasks: creating and configuring an agent in Copilot Studio, configuring generative AI and knowledge, creating and configuring topics, configuring tools, and sharing and publishing an agent. Microsoft classes it as intermediate, the lab takes a few minutes to load, and there is a 72-hour cooldown before a second attempt. PL-7008 is the preparation for it: the labs, a walkthrough of the official study guide, and three months of support while you get it done.

Three hours and twenty-five minutes of the PL-7008 day are spent in hands-on labs, against two hours and five minutes of instruction and fifteen minutes of introductions. There are four hands-on labs, preceded by a short setup lab and followed by an optional lab on the new Copilot Studio authoring experience that takes the total to three hours and fifty-five minutes. Microsoft's trainer guidance allocates over 50 per cent of the course time to labs and instructs trainers not to eat into it. Each lab is self-contained, so you build a fresh agent each time rather than carrying one through the day, which means a lab that goes wrong does not cost you the afternoon. Microsoft schedules the course as one full eight-hour day, including breaks.

MS-4018 teaches you to use Microsoft 365 Copilot well in Word, Excel, PowerPoint, Outlook and Teams, while PL-7008 teaches you to build the agents that other people then use. If the question is how to get more out of the Copilot licence you already have, MS-4018 is the course. If it is how to stop your organisation answering the same question fifty times a week, PL-7008 is the course. Plenty of organisations need both, usually in that order, and Vantage 365 runs both.

No, you do not need your own Microsoft Copilot Studio licence to attend PL-7008, because Vantage 365 provisions a dedicated lab environment for every learner and the labs include the licensing needed to publish an agent. You work in that environment rather than in your organisation's production tenant, so you can experiment freely without touching anything live. You will want appropriate Copilot Credit capacity of your own once you start building for real, and Vantage 365 will advise on what that looks like before you commit to anything.

Yes. Onsite and private virtual deliveries of PL-7008 can be built around your own scenarios and your own knowledge sources, which matters more on this course than most, because the hardest question of the day is what a given agent should and should not be allowed to see. Answering that against your own content rather than a sample data set is the difference between an interesting day and a useful one. Teams usually leave those sessions with something close to a first working agent and a clearer view of which of their ideas are worth building. Onsite is generally better value from four people upwards. Call 0121 269 8660 to talk it through.

Sit the APL-6006 Applied Skills assessment while the material is fresh, ideally within a fortnight, and build one small, real agent at work in the same period rather than an ambitious hypothetical one. After that, PL-7002 Create and Manage Automated Processes by Using Power Automate takes the automation behind your agent's tools a great deal further than the hundred-second limit and three parameter types that apply to Copilot Studio workflows. PL-900 Microsoft Power Platform Fundamentals fills in the platform underneath if you have not covered it. Autonomous, event-triggered agents are Microsoft's APL-6000 territory rather than PL-7008.

PL-7008 covers building, grounding, automating and publishing agents in Microsoft Copilot Studio, across four modules and four hands-on labs in one day. It covers what an agent is and how retrieval, task and autonomous agents differ; where Copilot Studio sits between Agent Builder in Microsoft 365 Copilot and Microsoft Foundry; the three agent harnesses, and the classic and new authoring experiences; environments and solutions; creating an agent with natural language, from a template, from blank or by installing a managed agent; writing agent instructions covering role, scope, context, tone and error handling; generative orchestration against classic orchestration; adding knowledge sources and configuring generative answers, source selection, web search and content moderation; designing conversations with topics, nodes, entities and variables; the system topics every agent gets, including Conversational boosting, Fallback and Escalate; building Copilot Studio workflows and adding them as tools; testing, authentication, publishing and deploying to channels including Microsoft Teams; and managing Copilot Credit capacity and usage reporting. It uses Microsoft Official Courseware, last refreshed by Microsoft in August 2026.

There are no formal prerequisites for PL-7008 and no coding experience is required, but it is an intermediate course and it moves at that pace. Microsoft expects delegates to be familiar with Microsoft Copilot Studio, Microsoft Teams and the Microsoft Power Platform, and lists basic data modelling, user experience design, requirements analysis and process analysis among the useful background skills. In practice, the people who get most from the day arrive having opened Copilot Studio at least once and with a real scenario in mind rather than a blank sheet. Describe what you are trying to build and Vantage 365 will tell you honestly whether PL-7008 is the right course, including when the answer is PL-900 Microsoft Power Platform Fundamentals first.

An agent is a program that uses AI to automate and carry out business processes, working alongside or on behalf of a person, team or organisation. Agents vary in complexity: a retrieval agent answers questions from knowledge you give it, a task agent completes defined pieces of work through tools, and an autonomous agent acts on triggers without being asked. An agent built in Copilot Studio draws on three things: knowledge, which grounds its answers in your content; topics, which are authored conversation paths for the exchanges that must go the same way every time; and tools, which let it take action in other systems through connectors, REST APIs, prompts, workflows and Model Context Protocol servers. Once published, it can be deployed to channels including Microsoft Teams, your own websites and applications, and Microsoft 365 Copilot.

Generative orchestration is the setting that lets an agent use AI to choose which topics, tools, knowledge sources and connected agents to use when responding, rather than matching the user's message to authored trigger phrases. It builds a plan from the names and descriptions you give those components, which is why naming a tool badly is a functional problem rather than a cosmetic one. Classic orchestration is the alternative and works the other way round: it selects a single topic by trigger phrase, and tools and knowledge are called explicitly from inside that topic, which gives predictable, controlled paths at the cost of flexibility. Newly created agents use generative orchestration by default, and Microsoft rewrote PL-7008 in June 2026 to teach it as the primary approach.