Prepstellar
Get the app

PL-900 · Platform Value and App Types

20 cards

AI and Agent Business Value

Swipe, scroll or use ← →

  1. Copilot in Power Apps

    Generative AI features reduce the effort needed to turn a business description into a starting solution. The value is measured in the distance between an idea and something you can open.

    In Power Apps, Copilot can turn plain-language descriptions into working apps and data models. Describe an equipment-booking process in a couple of sentences and what comes back is not a sketch: it is an app with tables behind it.

    1 / 20

  2. Copilot in Power Apps

    The first result is a starting point, not a verdict. A maker can continue the conversation to refine the generated data model and application instead of treating the first result as final.

    What the maker supplies What comes back What happens next
    A plain-language description A working app and a data model Conversation refines both

    That loop is where the value sits. A generated app that could not be adjusted would only move the work somewhere else; one that can be talked into shape removes it.

    2 / 20

  3. Quick check

    A maker describes a booking process to Copilot in Power Apps and dislikes part of the generated data model. What is the position?

    1. AThe generated model is fixed, so the maker must abandon it and rebuild the application from an empty screen

      Nothing is fixed: refinement through continued conversation is part of what the feature offers.

    2. BThe result is a starting point that further conversation can refine

      Right. The maker can continue the conversation to refine the generated data model and application rather than treating the first result as final.

    3. CThe description has to be rewritten as a database script before any part of the generated model can change

      The whole point is that plain language, not database administration, drives both the first result and its refinement.

    3 / 20

  4. Plans for a whole business problem

    One app is rarely the whole answer. Plans can generate a multi-component Power Platform solution from a natural-language description and supporting visuals.

    A plan outlines user roles and requirements and can include Dataverse tables, canvas or model-driven apps, cloud flows, Power Pages sites, Power BI, and Copilot Studio agents.

    4 / 20

  5. Plans for a whole business problem

    Scope of the requirement The generative starting point
    One app and the data behind it Copilot in Power Apps
    A problem crossing roles, data, apps, automation and agents Plans

    This makes Plans useful when the requirement spans several Power Platform components rather than one isolated app. Employee onboarding is the classic case: roles, records, an app, an approval, and an agent to answer questions — five things that all have to agree with each other.

    5 / 20

  6. Quick check

    An onboarding requirement involves employee roles, Dataverse data, an app, an approval automation and an agent. Which generative starting point fits?

    1. APlans, which outlines the roles and requirements and proposes the components together

      Right. Plans generates a multi-component solution from a natural-language description, outlining user roles and requirements alongside tables, apps, flows, sites, reports and agents.

    2. BApp Copilot, which produces one app and lets the wider roles and requirements be dropped from the picture

      Copilot in Power Apps turns a description into one app and its data model, which is narrower than this requirement.

    3. CA conversational agent, replacing the apps, data and automation

      An agent is one component of the solution, not a substitute for its data, applications and automation.

    6 / 20

  7. Copilot in Power Automate

    Power Automate applies the same conversational value to automation. Copilot can interpret a natural-language description and propose a cloud flow with triggers, actions, and connections.

    Follow-up prompts can modify, extend, or simplify the proposed flow. The maker reviews the proposal before accepting it into the flow designer, so conversation accelerates the starting structure while review remains part of the authoring experience. What arrives is a flow's skeleton — what starts it, what it does, what it connects to — rather than a finished automation nobody has looked at.

    7 / 20

  8. Quick check

    What does Copilot in Power Automate return from a description such as "notify the account manager whenever a high-value order arrives"?

    1. AA static screen layout with forms, galleries and labels

      Arranging screen elements is app design; this Copilot works on automation.

    2. BAn interface generated from the related Dataverse tables

      Generating an interface from related tables describes a model-driven app rather than a proposed flow.

    3. CA proposed cloud flow with its triggers, actions and connections, for the maker to review

      Right. Copilot interprets the natural-language description and proposes a cloud flow with triggers, actions and connections, which the maker reviews before accepting it.

    8 / 20

  9. Keep your progress in the app

    That’s 3 of 8 quick checks. In the app they stay answered, and every lesson remembers where you left off.

    Get the app Keep reading here

  10. Refinement is part of the value

    Across all three surfaces the pattern repeats: generate, then converse. The app, the data model, the plan and the flow all arrive as proposals.

    Surface What is generated How it is improved
    Power Apps An app and a data model Continued conversation
    Plans Roles, requirements and components Iteration on individual requirements
    Power Automate A flow with triggers, actions and connections Follow-up prompts that modify, extend or simplify it

    Refinement is not damage control after a poor first draft. It is how a described intention becomes specific enough to run.

    9 / 20

  11. Quick check

    Why is the ability to refine a generated result central to the business value of these features?

    1. ABecause the first app, plan or flow is a starting point that follow-up requests reshape

      Right. Generated apps, data models, plans and flows are starting points, and continued conversation or follow-up prompts adjust them.

    2. BBecause the first proposal has to be accepted unchanged

      The maker reviews before accepting, and follow-up prompts can modify, extend or simplify what was proposed.

    3. CBecause refinement means replacing the natural-language request with hands-on database administration work

      Refinement happens through the same conversational route as the original description, not through database administration.

    10 / 20

  12. What Copilot Studio is for

    Microsoft Copilot Studio is a low-code platform for building, deploying, and managing intelligent agents without requiring deep programming expertise.

    Its agents can answer questions from knowledge sources, take action through tools and connectors, automate multi-step business processes, and integrate with external systems.

    11 / 20

  13. What Copilot Studio is for

    The business value is therefore broader than question answering: an agent can connect information access with action.

    What an agent can do What that looks like
    Answer from knowledge sources Explain a policy to an employee
    Act through tools and connectors Open a ticket in the service system
    Automate a multi-step process Carry a case from intake to resolution
    Integrate with external systems Update a record held outside Power Platform

    An assistant that can only explain the expenses policy is half an agent. One that can also file the expense claim is the whole proposition.

    12 / 20

  14. Quick check

    How is the value of a Copilot Studio agent best described?

    1. AIt designs static websites, which cannot connect to any of the organization's business systems

      Agents integrate with external systems and act through connectors; they are not a static website builder.

    2. BIt stores relational business data, with no conversational experience on top of it

      Storing business data in tables is what Dataverse contributes; an agent supplies the conversational and acting layer.

    3. CIt answers from knowledge and acts through tools and connectors

      Right. Copilot Studio agents answer questions from knowledge sources and take action through tools and connectors, automating multi-step processes and integrating with external systems.

    13 / 20

  15. Conversational agents

    Conversational agents respond to a user's natural-language question or request. The person starts, the agent answers, and it can act on what was asked.

    They fit employee self-service, IT support, customer service, sales support, and knowledge-access scenarios in which a person initiates the interaction. An employee asking how many holiday days are left, and then requesting to book three of them, is the shape of the whole category.

    14 / 20

  16. Quick check

    Employees need to ask benefits questions in their own words and get transactional help with what they ask. Which agent fits?

    1. AAn autonomous agent started by a fixed schedule

      A schedule starting the work is the autonomous pattern; here a person initiates each interaction.

    2. BA conversational agent, which responds when the employee starts the request

      Right. Conversational agents respond to a user's natural-language question or request, which covers employee self-service and transactional support.

    3. CA plan, which produces components but no interaction

      A plan proposes the components of a solution; it is not the runtime that talks to employees.

    15 / 20

  17. Autonomous agents

    Autonomous agents act proactively on a schedule, after a trigger event, or within a multi-step process without requiring a person to initiate every action.

    They fit repetitive or data-intensive work such as lead qualification, contract processing, invoice processing, supply-chain monitoring, and compliance checks. Nobody opens a conversation to start any of those; the work arrives on its own.

    16 / 20

  18. Autonomous agents

    The selection rule is based on the initiating pattern and expected outcome. Choose a conversational agent when a user needs dialogue, guidance, or transactional support. Choose an autonomous agent when the work should begin proactively and continue through a defined business process.

    Question to ask Conversational Autonomous
    Who starts it? A person, in their own words A schedule, an event, or a process step
    What does it look like? Self-service, IT support, customer service, sales support, knowledge access Lead qualification, contracts, invoices, supply chain, compliance
    How often per case? Once per conversation Repeatedly, without a person for each case
    17 / 20

  19. Quick check

    Invoices arrive continuously and must be inspected, validated and pushed into the accounting system without anyone opening each case. What fits?

    1. AA conversational agent, waiting for someone to ask a question about an invoice

      A conversational agent needs a person to start each interaction, which is what this requirement rules out.

    2. BA canvas app, in which a person positions the controls needed for each invoice that arrives

      An app is a screen a person operates, so it reintroduces exactly the per-case human step being removed.

    3. CAn autonomous agent, beginning from the event and running the multi-step process

      Right. Autonomous agents act after a trigger event and continue through a multi-step process without a person initiating every action, and invoice processing is one of their listed scenarios.

    18 / 20

  20. Key takeaways

    • Generative AI can convert natural-language business intent into starting apps, data models, flows, or multi-component plans that makers can refine.
    • Copilot Studio agents can combine answers from knowledge with actions through tools and connectors.
    • Conversational agents respond to users, while autonomous agents initiate work from schedules, events, or processes.
    • Match the scope to the tool: one app to Copilot in Power Apps, an automation to Copilot in Power Automate, a cross-component business problem to Plans.
    19 / 20

  21. Quick check

    Which summary of generative AI and agents in Power Platform is accurate?

    1. AGenerated apps, plans and flows are final, and every agent has to wait for a user before doing anything at all

      Generated results are starting points that makers refine, and autonomous agents act without a user starting each action.

    2. BGeneration gives a refinable start; agents respond or act on their own

      Right. Generative AI turns business intent into a starting point that can be refined, and agents are either conversational or autonomous depending on what initiates the work.

    3. CGeneration replaces the business description entirely, and autonomous agents are the ones that answer user questions

      The description is the input rather than something replaced, and answering user questions is the conversational pattern.

    20 / 20

  22. 8 quick checks · then the test

    In the app, finishing the quick checks opens this lesson’s 10-question test, and the ones you miss come back exactly when you’re about to forget them.

    Take this lesson’s test → Get the app
← Previous Keep your progress · Get the app Next card →

The whole course, on your phone

Lessons you can read, audio you can listen to on the way to work, and practice that remembers what you got wrong.

Get the app