Chapter: Fundamental Concepts of Cloud Flows and Triggers

Objec ve

In this chapter, you will master the founda onal architecture of Power Automate Cloud Flows.
You will explore how to design automated workflows that serve as the "nervous system" for
intelligent applica ons. By the end of this chapter, you will be able to dis nguish between
different flow types, configure complex triggers, and understand the flow of data (input/output)
required to orchestrate AI-driven business logic.

Core Concepts

At its core, a Cloud Flow is a series of automated steps triggered by a specific event. For the AB-
410 exam, understanding the "trigger-ac on" rela onship is cri cal, especially when
coordina ng between Dataverse, external APIs, and AI models.

1. The Anatomy of a Flow

Every Cloud Flow consists of two primary components:

 The Trigger: The "Star ng Event." This is a specific condi on that must be met for the
flow to run. A flow can only have one trigger.

 The Ac on: The "Task." These are the steps the flow performs a er it is ini ated. A flow
can have hundreds of ac ons, including loops and condi onal branching.

2. Categories of Cloud Flows

Understanding which flow type to use is a key architectural decision:

 Automated Flows: Triggered by an event from a connector.

o Example: "When a new email arrives" or "When a row is added to Dataverse."

These are used for real- me background processing.

 Instant (Bu on) Flows: Triggered manually by a user.

o Example: A bu on inside a Canvas App or a manual trigger from the mobile app.
These are used for on-demand tasks.

 Scheduled Flows: Triggered based on a specific me frequency.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 o Example: "Every Monday at 8:00 AM." These are used for batch processing or
daily repor ng.

3. Trigger Types and Behaviors

Triggers behave in two dis nct ways:

 Polling Triggers: The flow "checks" the service at a regular interval (e.g., every 3
minutes) to see if new data is available.

 Push (Webhook) Triggers: The service "calls" the flow immediately when an event
occurs. Dataverse triggers are typically Push-based, making them ideal for high-
performance intelligent apps.

4. Dynamic Content and Data Types

Power Automate passes data between steps using Dynamic Content. When a trigger captures a
record, it produces "outputs" (like a User ID or Email body) that become "inputs" for the next
ac on. Understanding JSON schemas is helpful here, as AI models o en require specific data
formats to func on correctly.

Prac cal Case Study: The "Intelligent Triage" System

The Challenge: "Global Logis cs Corp" receives thousands of customer inquiries via a web form.
Human operators spend 40% of their me manually reading emails just to route them to the
correct department (e.g., Billing, Technical Support, or Sales).

The Solu on Design:

1. Trigger: An Automated Cloud Flow is set to run "When a new row is added" to the
Dataverse Inquiry table.

2. AI Orchestra on: The flow takes the Descrip on field and passes it to an AI Builder
Category Classifica on model.

3. Dynamic Rou ng: Based on the AI's output (the category), the flow uses a Switch ac on
to:

o Assign the record to the "Billing" queue if the AI detects financial keywords.

o Email a "Technical Specialist" if the AI detects hardware issues.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 4. Outcome: The manual sor ng me was reduced to zero, and the response me for
cri cal technical issues improved by 65%.

Step-by-Step Example: Building a Basic AI-Triggered Flow

This example demonstrates how to create a flow that monitors a Dataverse table and uses a
simple condi on to trigger a no fica on.

1. Define the Trigger:

o Go to Power Automate -> Create -> Automated Cloud Flow.

o Select the trigger: When a row is added, modified, or deleted (Microso

Dataverse).

o Set Change type to Added, Table name to Invoices, and Scope to Organiza on.

2. Add an AI Ac on (Op onal for Logic):

o Click + New Step and search for AI Builder.

o Select Detect sen ment in text.

o Map the Invoice Descrip on from the trigger to the Text field.

3. Implement Condi onal Logic:

o Add a Condi on ac on.

o Logic: If Sen ment is equal to nega ve.

4. Define Ac ons (Success/Failure):

o In the If Yes branch, add a "Send an email (V2)" ac on to no fy a supervisor.

o In the If No branch, add an "Update a row (Dataverse)" ac on to mark the

invoice as "Clear for Processing."

5. Test:

o Manually add a row to the Invoices table in Dataverse and watch the flow
execu on history to verify the logic path.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Knowledge Check: Mul ple Choice Ques ons

1. A developer needs a workflow to run every night at 11:00 PM to summarize the day's sales.
Which flow type should they select?

 A) Automated Cloud Flow

 B) Instant Cloud Flow

 C) Scheduled Cloud Flow

 D) Desktop Flow

Correct Answer: C. Explana on: Scheduled flows are the only type designed to run at specific
me intervals without requiring an external event or manual user interven on.

2. In Power Automate, what is the primary func on of "Dynamic Content"?

 A) To create anima ons in the user interface.

 B) To allow data produced by a previous step (like a trigger) to be used as an input for a
subsequent ac on.

 C) To increase the speed of the internet connec on.

 D) To change the color of the Power Automate designer.

Correct Answer: B. Explana on: Dynamic Content is the mechanism for data inheritance. It
enables the flow to use specific variables (like a customer's name) across different connectors
and ac ons.

3. Which of the following is a characteris c of a "Polling" trigger?

 A) It reacts instantly the moment a change occurs in the source system.

 B) It requires the user to click a bu on to start the flow.

 C) It periodically checks the source system for updates based on a specific interval.

 D) It only works with Microso Excel.

Correct Answer: C. Explana on: Polling triggers "poll" the service (e.g., every minute or hour)
to see if the trigger condi on is met, whereas Webhook/Push triggers are no fied immediately
by the service.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Building Flows using Copilot and Natural Language

Objec ve

In this chapter, you will explore the transi on from manual, step-by-step flow construc on to
AI-assisted automa on. You will learn how to use Copilot in Power Automate to describe your
automa on goals in natural language and watch as the AI dra s the architecture for you. By the
end of this chapter, you will be able to refine AI-generated flows, use the Copilot chat pane for
real- me troubleshoo ng, and understand the best prac ces for "Prompt Engineering" to
achieve accurate automa on results.

Core Concepts

Building flows with Copilot represents a paradigm shi from "Logic Building" to "Logic
Descrip on." Instead of searching through thousands of connectors, you act as the architect,
and the AI acts as the builder.

1. The Natural Language to Flow (NL2Flow) Engine

The core of this experience is a Large Language Model (LLM) that has been trained on millions
of successful Power Automate pa erns. When you provide a prompt, the engine:

 Iden fies Intent: Determines the trigger and the necessary ac ons.

 Selects Connectors: Automa cally chooses the correct services (e.g., Dataverse,
Outlook, Teams).

 Dra s Parameters: Suggests basic configura ons for the ac ons based on your
descrip on.

2. The Conversa onal Designer

Unlike the classic designer, the Copilot designer is interac ve. You can use the chat pane to:

 Add Steps: "Add an ac on to send a Teams message a er the email is sent."

 Update Logic: "Change the condi on to only run if the amount is over $500."

 Explain Formulas: "Help me write a Power Fx expression to format this date."

3. Principles of Effec ve Promp ng

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
To get the best results from Copilot, your prompts should follow the Context-Ac on-Detail
(CAD) framework:

 Context: Where does the data start? ("When a new lead is added to Dataverse...")

 Ac on: What should happen? ("...summarize the notes using AI...")

 Detail: Where does the data go? ("...and post it to the Sales Teams channel.")

Prac cal Case Study: The "Zero-Code" Expense Auditor

The Challenge: "Finance-Flow Inc" needs a way to audit expense reports. If a report is over
$1,000, it needs a manager's approval; otherwise, it should be marked as "Auto-Approved" in
the database. The finance manager knows the logic but doesn't know how to navigate the
Power Automate designer.

The Solu on Design:

1. The Prompt: The manager types: "When a new row is added to the Expenses table, if the
Total Amount is greater than 1000, start an approval. If approved, update the status to
Paid. If less than 1000, just update status to Paid."

2. AI Dra ing: Copilot generates a flow with a Dataverse trigger, a Condi on card, an
Approvals ac on, and two "Update a row" ac ons.

3. Refinement: The manager uses the Copilot chat pane to say, "Add a step to no fy me on
mobile if the approval is rejected."

4. Outcome: A complex branching logic flow was built in under two minutes without the
user manually placing a single card.

Step-by-Step Example: Building a Flow from a Descrip on

This example guides you through the process of using the Copilot "Start with AI" feature.

1. Ini ate the AI: Navigate to the Power Automate home page and select "Describe it to
design it" or use the prompt box on the home screen.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 2. Enter your Prompt: Type the following: "When a file is uploaded to the 'Invoices' folder
in SharePoint, use AI to extract the total amount and save it to my Dataverse 'Invoices'
table."

3. Review the Suggested Flow: Copilot will show a diagram of the proposed flow:

o Trigger: When a file is created (proper es only).

o Ac on 1: Extract informa on from documents (AI Builder).

o Ac on 2: Add a new row (Dataverse). Click Next.

4. Finalize Connec ons: Ensure your SharePoint and Dataverse connec ons are green.
Click Create Flow.

5. Edit with Copilot: Once in the designer, open the Copilot pane on the right. Type: "Add a
step at the end to send me a summary email." The AI will automa cally insert the
Outlook "Send an email" ac on and map the dynamic content from the previous steps
for you.

Knowledge Check: Mul ple Choice Ques ons

1. What is the most effec ve way to improve the accuracy of a flow generated by Copilot?

 A) Using shorter, one-word prompts.

 B) Providing a detailed prompt that includes the specific trigger, the core ac ons, and
the final des na on of the data.

 C) Restar ng your computer.

 D) Manually dele ng all connectors and star ng over.

Correct Answer: B.

Explana on: Copilot thrives on context. The more specific you are about the "source" and
"des na on" of your data, the be er it can select the correct connectors and parameters.

2. While in the Power Automate designer, what can you use the Copilot chat pane for?

 A) To check your personal email.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  B) To ask the AI to add, delete, or modify specific steps in your flow using natural
language.

 C) To play music.

 D) To browse the internet for news.

Correct Answer: B.

Explana on: The Copilot chat pane is a "conversa onal editor." It allows you to refine and
iterate on your flow logic without needing to manually search the ac on library.

3. If Copilot suggests a flow that is 80% correct but uses the wrong SharePoint folder, what is
the best course of ac on?

 A) Close the browser and try again tomorrow.

 B) Accept the flow and then use the designer (or the Copilot pane) to change the folder
selec on in the trigger se ngs.

 C) Delete the SharePoint connector.

 D) Report the AI as broken.

Correct Answer: B.

Explana on: AI-generated flows are dra s. They are intended to handle the heavy li ing of
architecture, but the developer should always review and refine specific parameters like folder
paths, email addresses, or table names.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Advanced Logic: Condi onals, Switches, and Scopes

Objec ve

In this chapter, you will master the implementa on of advanced control-flow structures within
Power Automate. You will move beyond linear "Step A to Step B" workflows to build complex,
branching logic that can handle mul ple business scenarios. By the end of this chapter, you will
be able to implement mul -path Condi ons, high-efficiency Switches, and organiza onal
Scopes to create robust, maintainable, and intelligent AI orchestra ons.

Core Concepts

Advanced logic is what allows an automated flow to make "decisions." In the context of Exam
AB-410, these structures are essen al for rou ng data to different AI models or handling various
response types from a Copilot.

1. Condi onals (If/Then/Else)

The Condi on ac on is a binary split. It evaluates a boolean expression (True or False).

 Best Use: For simple "Yes/No" logic (e.g., Is the invoice amount > $1,000?).

 Nes ng: You can place condi ons inside other condi ons, but be careful—deep nes ng
can make flows difficult to read and troubleshoot.

2. Switches (Case Logic)

A Switch ac on is used when you have mul ple dis nct paths based on a single value.

 Mechanism: It looks at a specific variable (like a "Status" field) and provides mul ple
"Cases." If the variable matches a case, that specific path is followed.

 Default Case: Every switch includes a "Default" path if none of the cases match.

 Efficiency: For 3+ branches, a Switch is significantly cleaner and more performant than
nested Condi ons.

3. Scopes (Logical Grouping)

A Scope is a container that groups mul ple ac ons together.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  Organiza on: It allows you to collapse sec ons of your flow to keep the designer clean.

 Error Handling: This is the most cri cal use for AB-410. You can set a "Run A er"
configura on on a subsequent ac on to catch errors if any ac on inside the Scope fails
(Try/Catch pa ern).

Prac cal Case Study: The "Intelligent Triage" Router

The Challenge: "Global Logis cs" receives support ckets categorized by an AI model into four
types: Billing, Technical, Shipping, and General. Each category requires a different set of ac ons,
different stakeholders, and different Dataverse updates.

The Solu on Design:

1. AI Analysis: An ini al step uses AI Builder to predict the "Category" of the cket.

2. Logic Layer: A Switch ac on evaluates the Category output.

o Case Billing: Routes to the Finance Teams channel.

o Case Technical: Opens a cket in the IT support system.

o Case Shipping: Triggers a tracking lookup via a custom API.

3. Safety Layer: All ac ons are placed inside a Scope named "Triage_Logic." If the AI fails or
an API is down, a "Failure Scope" sends a no fica on to an admin.

Step-by-Step Example: Implemen ng a Try-Catch Pa ern with Scopes

This example demonstrates how to use Scopes to handle errors gracefully in a data-heavy flow.

1. Create the "Try" Scope: Add a Scope ac on and rename it to Scope_Processing. Inside
this scope, add your primary ac ons (e.g., "Add a row to Dataverse").

2. Create the "Catch" Scope: Immediately a er the first scope, add a second Scope and
rename it to Scope_Error_Handling. Inside this, add an ac on to "Send an email
no fica on" about the failure.

3. Configure "Run A er": Click the three dots (...) on the Scope_Error_Handling card and
select Configure run a er.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 o Uncheck is successful.

o Check has failed, is skipped, and has med out.

4. The Result: If the first scope succeeds, the error scope is skipped. If anything inside the
first scope fails, the error scope triggers immediately to alert the developer.

Knowledge Check: Mul ple Choice Ques ons

1. When should a developer choose a 'Switch' ac on over mul ple 'Condi on' ac ons?

 A) When they only have one True/False ques on.

 B) When they need to evaluate a single variable against mul ple poten al values (e.g., 5
different regions).

 C) When they want to slow down the flow.

 D) When they are not using Dataverse.

Correct Answer: B.

Explana on: Switches are specifically designed for mul -path branching based on a single input
value, making the flow much cleaner and easier to maintain than mul ple nested condi ons.

2. What is the primary benefit of using a 'Scope' ac on for error handling?

 A) It makes the flow run twice as fast.

 B) It allows you to group mul ple ac ons and trigger a single "Catch" logic if any ac on
within that group fails.

 C) It hides the flow from other users.

 D) It is the only way to connect to SharePoint.

Correct Answer: B.

Explana on: Scopes act as a logical unit. By configuring the "Run A er" se ngs, you can create
professional "Try/Catch" blocks to ensure your intelligent app handles unexpected errors
gracefully.

3. In a Switch ac on, what happens if the input value does not match any of the defined
'Cases'?

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  A) The flow crashes.

 B) The flow waits forever.

 C) The flow follows the 'Default' path.

 D) The flow deletes the input value.

Correct Answer: C.

Explana on: The Default path is the safety net of a Switch. It ensures that even if an
unexpected value is encountered, the flow has a predefined logic path to follow.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Data Operations: Parse JSON, Filter Array, and Select
Objective

In this chapter, you will master the Data Operations connector in Power Automate. You will
learn how to manipulate complex data structures, refine large datasets, and transform raw JSON
outputs from AI models into clean, usable variables. By the end of this chapter, you will be able
to implement "Data Thinning" techniques using Filter Array, reshape data using Select, and
unlock dynamic content using Parse JSON.

Core Concepts

When building intelligent applications, you often receive "bloated" data from APIs or AI models.
Data Operations allow you to perform "On-the-Fly" transformations without needing a database.

1. Parse JSON: The Key to Dynamic Content

AI models often return data in a JSON (JavaScript Object Notation) string. While humans can
read it, Power Automate sees it as a single block of text.

 The Operation: You provide a Schema (a map of the data).

 The Result: Power Automate "unpacks" the string, turning every JSON key (like
SentimentScore or DetectedLanguage) into a selectable token in the Dynamic Content
picker.

2. Filter Array: Precision Retrieval

Sometimes an AI or a database returns 100 items, but you only need three.

 The Operation: It takes an array (list) and a logical rule (e.g., Price is greater than 500).
 The Result: A "Thinned" array containing only the items that met your criteria. This is
essential for reducing "Token Noise" before sending data to an LLM.
3. Select: Reshaping the Data

The Select action allows you to change the "Shape" of your data.

 The Operation: You map existing keys to new names (e.g., changing internal_id_99 to
InvoiceID).

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  The Result: A new array containing only the specific columns you need, formatted
exactly how your next action expects them.

Practical Case Study: The "Clean Feed" for AI Summarization

The Challenge: "Eco-Analytics" retrieves a list of 50 sensor readings from an external API
every hour. The API output is a mess of technical metadata (IDs, timestamps, hardware versions)
that the AI model doesn't need. Sending all 50 full records to the AI is expensive and causes
"hallucinations."

The Solution Design:

1. Parse JSON: The flow unpacks the API response so individual readings are accessible.
2. Filter Array: The flow removes all readings where the Status is not "Warning." This
reduces the list from 50 to 8 items.

3. Select: The flow extracts only the Location and Temperature fields, discarding the
metadata.

4. AI Outcome: The AI receives a clean, tiny list of only the 8 critical issues. It provides a
100% accurate summary because it isn't distracted by irrelevant data.

Step-by-Step Example: Transforming an AI Output

This example demonstrates how to take a raw JSON string from an AI model and make its
properties usable in an email.
1. The Input (Compose): Assume an AI model outputs this text: {"analysis": "Positive",
"confidence": 0.98}.

2. Add Parse JSON: Select the output of the previous step. Click "Use sample payload to
generate schema" and paste the text above. Power Automate will automatically create
the schema for you.

3. Add Select (Optional): If you had a list of these, you could use Select to create a "Map."

o From: The Body of Parse JSON.

o Map (Key): CustomerSentiment | Map (Value): analysis.

4. Add Filter Array:

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 o From: The output of your Select or Parse JSON.
o Rule: Choose confidence is is greater than 0.90.

5. The Result: You can now use a "Send an email" action and simply pick
CustomerSentiment from the Dynamic Content list. The email will only send if the AI
was highly confident.

Knowledge Check: Multiple Choice Questions

1. Which Data Operation should you use if you want to change the names of columns in an
array to make them easier for a human to read?

 A) Parse JSON

 B) Filter Array

 C) Select

 D) Join

Correct Answer: C.

Explanation: The 'Select' action is specifically designed to reshape data by mapping old keys to
new keys and choosing which specific fields to keep.

2. Why is 'Parse JSON' often necessary after calling a custom AI model via an HTTP
request?

 A) It makes the flow run in dark mode.

 B) It converts the raw text response from the AI into individual Dynamic Content tokens
that can be used in later steps.

 C) It deletes the AI model after use.

 D) It translates the JSON into Spanish.

Correct Answer: B.

Explanation: Without Parse JSON, the output of an API is just a single block of "Body" text.
Parsing it allows Power Automate to understand the internal structure of the data.

3. What is the main benefit of using 'Filter Array' before sending data to an AI
summarization model?

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  A) It increases the cost of the flow.
 B) It improves AI accuracy and reduces costs by removing irrelevant data, ensuring the
AI only processes the most important information.
 C) It changes the font of the AI's response.

 D) It requires a Python script.

Correct Answer: B.

Explanation: In intelligent applications, "Less is More." Filtering out "noise" (irrelevant

records) helps the AI focus on the "signal," leading to better results and lower token
consumption.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Working with Parent and Child Flows for Modular Logic

Objec ve

In this chapter, you will master the architectural pa ern of Modular Automa on. You will learn
how to decompose large, complex workflows into manageable, reusable components by
implemen ng Parent and Child Flows. By the end of this chapter, you will be able to design
"Pluggable" AI orchestra ons where a single Parent flow coordinates mul ple specialized Child
flows, improving maintainability, reducing redundancy, and overcoming standard flow size
limita ons.

Core Concepts

As intelligent applica ons grow in complexity, "Monolithic" flows (one giant flow doing
everything) become difficult to debug and slow to load. Modular design is the professional
standard for the AB-410 framework.

1. What are Parent and Child Flows?

 Child Flow: A specialized, standalone workflow designed to perform a specific,

repeatable task (e.g., "Analyze Invoice Sen ment" or "Post to Regional Teams Channel").

 Parent Flow: The orchestrator that manages the main business process and "calls" one
or more Child flows to handle specific sub-tasks.

2. The Mechanics of the "Call"

To link these two, you use specific triggers and ac ons:

 Child Trigger: Must use the "Manually trigger a flow" trigger (from the Instant flows
category) to accept inputs from the Parent.

 Parent Ac on: Uses the "Run a Child Flow" ac on to send data and wait for a response.

 Child Response: Must end with the "Respond to a Power App or flow" ac on to send
data back to the Parent.

3. Key Requirements for Modular Success

For the "Run a Child Flow" ac on to appear and func on:

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 1. Solu ons: Both flows MUST reside within the same Power Pla orm Solu on.

2. Connec ons: The Child flow must use "Embedded connec ons" (the user creates them)
rather than "Provided by run-only user."

3. Permissions: The Parent flow must have permission to run the Child flow.

Prac cal Case Study: The "Modular AI Hub"

The Challenge: "Global Retail" has 10 different business units (Sales, HR, Logis cs, etc.) that all
need to summarize documents using AI. Instead of building 10 iden cal AI summary blocks in 10
different flows, they want a central "Summary Engine."

The Solu on Design:

1. The Child Flow (The Engine): A single flow named AI_Summary_Service is created. It
takes "Raw Text" as input, runs an AI Builder model, and returns a "Short Summary" as
output.

2. The Parent Flows (The Users): The HR flow (for resumes) and the Sales flow (for lead
notes) both use the Run a Child Flow ac on to send their specific text to the
AI_Summary_Service.

3. Outcome: If the company decides to switch AI models or change the summary length,
they only need to update the one Child flow. All 10 business units receive the upgrade
instantly.

Step-by-Step Example: Implemen ng a Child Flow for Data Valida on

This example demonstrates how to build a Child flow that checks for "Blacklisted" keywords and
reports back to a Parent.

Step 1: Build the Child Flow

1. Create a new flow in a Solu on. Trigger: Manually trigger a flow.

2. Add a Text input called InputText.

3. Add a Condi on: If InputText contains the word "Urgent".

4. Add the ac on Respond to a Power App or flow.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 o Add an Output (Boolean) called isHighPriority. Set to true in the "Yes" path and
false in the "No" path.

5. Save and set the "Run-only users" connec ons to "Use this connec on."

Step 2: Build the Parent Flow

1. Create a flow triggered by When a new email arrives.

2. Add the ac on Run a Child Flow.

3. Select your Child flow from the dropdown.

4. Map the Email Body to the InputText field.

5. Use the isHighPriority output from the Child flow in a Parent-level condi on to decide
whether to alert a manager.

Knowledge Check: Mul ple Choice Ques ons

1. What is the mandatory requirement for a flow to be called as a Child flow by a Parent flow?

 A) It must be a Scheduled flow.

 B) It must be located inside a Microso Team.

 C) Both the Parent and Child flows must be located within the same Solu on.

 D) The Child flow must be shared with the en re organiza on.

Correct Answer: C.

Explana on: Parent/Child flow architecture is a Solu on-aware feature. Flows outside of a
solu on cannot u lize the 'Run a Child Flow' ac on.

2. Which ac on MUST be at the end of a Child flow for the Parent flow to receive and use data
from it?

 A) Terminate

 B) Update a row

 C) Respond to a Power App or flow

 D) Send an email

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Correct Answer: C.

Explana on: This ac on creates the "Return Path" for data. Without it, the Parent flow will
simply trigger the Child and move on without wai ng for a result.

3. Why is using Modular Logic (Parent/Child) be er than building one large flow for a complex
AI process?

 A) It makes the flow icons look different.

 B) It allows for "Single Source of Truth" logic; you can update the logic in one Child flow,
and every Parent flow that uses it is updated simultaneously.

 C) Child flows do not require a license.

 D) It prevents the use of Dataverse.

Correct Answer: B.

Explana on: Modular logic promotes "DRY" (Don't Repeat Yourself) development. It simplifies
maintenance and makes debugging easier by isola ng specific tasks into separate flows.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Error Handling: Configure "Run A er" and Retry Policies

Objec ve

In this chapter, you will master the cri cal skills required to build resilient and reliable
automa on. You will learn how to move beyond "Happy Path" workflows to account for
transient API failures and logic errors. By the end of this chapter, you will be able to implement
sophis cated Retry Policies for connectors and u lize the "Run A er" configura on to create
professional-grade "Try-Catch-Finally" blocks within your AI orchestra ons.

Core Concepts

In the world of cloud automa on and AI, failures are not a ma er of "if," but "when." Services
might be temporarily unavailable, or an AI model might return an unexpected null value. Error
handling ensures your app recovers gracefully instead of simply stopping.

1. Retry Policies (The First Line of Defense)

Most ac ons in Power Automate include a built-in Retry Policy. This dictates how the flow
behaves if a "transient" error (like a 429 Rate Limit or a 502 Gateway Error) occurs.

 Default: Typically 4 retries over a specific period.

 None: The flow fails immediately upon the first error.

 Fixed Interval: Retries at a set me (e.g., every 20 seconds).

 Exponen al Interval: Increases the me between each retry (e.g., 10s, 30s, 90s). This is
best for handling service thro ling.

2. The "Run A er" Configura on (The Logical Guard)

Every ac on in a flow has a "Run A er" se ng. This defines the condi on under which the next
ac on should execute. By default, an ac on only runs if the previous one was Successful. The
four possible states are:

1. Is successful: (Default) Run if the previous step worked.

2. Has failed: Run if the previous step encountered an error.

3. Is skipped: Run if the previous step was not executed (usually due to a condi on).

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 4. Has med out: Run if the previous step took too long.

3. The Try-Catch-Finally Pa ern

By combining Scopes and Run A er logic, you can implement the classic programming pa ern
for error handling:

 Try: A Scope containing the main business logic.

 Catch: A Scope that runs only if the "Try" scope fails. This is used for logging errors or
sending alerts.

 Finally: A Scope or ac on that runs regardless of success or failure (e.g., closing a

connec on or upda ng a status).

Prac cal Case Study: Resilient AI Invoice Processing

The Challenge: "FinTech Solu ons" uses a flow to send invoices to an AI model for data
extrac on. Occasionally, the AI service is busy (Error 429) or the document is corrupted (Error
400). When this happens, the flow currently dies, leaving the invoice "stuck" in a pending state
without anyone knowing why.

The Solu on Design:

1. Retry Policy: The AI Extrac on ac on is set to an Exponen al Interval to handle

temporary busy signals.

2. Logic Guard: A "Catch" Scope is added. Its "Run A er" is configured to trigger if the "AI
Extrac on" fails.

3. No fica on: The "Catch" scope sends a Teams message to the finance clerk with a link
to the failed invoice.

4. Outcome: The flow no longer "disappears" on failure. The clerk is no fied of specific
data issues, while transient technical blips are resolved automa cally by the retry policy.

Step-by-Step Example: Building a "Self-Healing" Flow

This example demonstrates how to configure an ac on to handle a failure and no fy the team.

1. Set a Retry Policy:

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  Add an ac on (e.g., "HTTP" or "Dataverse - Add a row").

 Click the three dots (...) -> Se ngs.

 Under Retry Policy, select Exponen al Interval. Set Count to 5 and click Done.

2. Create the Failure Path:

 Add a "Send an email (V2)" ac on immediately a er the Dataverse ac on.

 Click the three dots (...) on the Email ac on -> Configure run a er.

 Uncheck is successful.

 Check has failed and has med out. Click Done.

3. Test the Resiliency:

 Inten onally trigger a failure (e.g., provide a null value to a required Dataverse field).

 The flow will a empt to retry based on your policy. If it s ll fails, it will bypass the
standard "Success" path and trigger the email no fica on instead.

Knowledge Check: Mul ple Choice Ques ons

1. Which Retry Policy is most effec ve for handling 'Rate Limit' (429) errors from an AI
service?

 A) None

 B) Fixed Interval

 C) Exponen al Interval

 D) Default

Correct Answer: C.

Explana on: Exponen al Interval "backs off" progressively, giving the service more me to
recover from a high volume of requests, which is the standard way to handle thro ling.

2. By default, when does an ac on in Power Automate execute?

 A) Only if the previous ac on failed.

 B) Only if the previous ac on was successful.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-
applica ons/?referralCode=857964209E5CBDF3B947
  C) Only if the flow is manually triggered.

 D) Simultaneously with the previous ac on.

Correct Answer: B.

Explana on: The default "Run A er" se ng is 'is successful.' This is why a flow stops at the first
sign of an error unless you explicitly configure an error-handling path.

3. You have a 'Try' Scope followed by a 'Catch' Scope. To ensure the 'Catch' Scope only runs if
something inside the 'Try' Scope goes wrong, which 'Run A er' se ngs should you check?

 A) is successful

 B) has failed, is skipped, and has med out

 C) has failed and has med out

 D) is skipped and is successful

Correct Answer: C.

Explana on: To catch an error, you must uncheck 'is successful' and check 'has failed' and 'has
med out.' (Checking 'is skipped' is generally used for 'Finally' blocks that must run even if a
previous error-handling step was bypassed).

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Error Handling: Configure "Run A er" and Retry Policies

Objec ve

In this chapter, you will master the cri cal skills required to build resilient and reliable
automa on. You will learn how to move beyond "Happy Path" workflows to account for
transient API failures and logic errors. By the end of this chapter, you will be able to implement
sophis cated Retry Policies for connectors and u lize the "Run A er" configura on to create
professional-grade "Try-Catch-Finally" blocks within your AI orchestra ons.

Core Concepts

In the world of cloud automa on and AI, failures are not a ma er of "if," but "when." Services
might be temporarily unavailable, or an AI model might return an unexpected null value. Error
handling ensures your app recovers gracefully instead of simply stopping.

1. Retry Policies (The First Line of Defense)

Most ac ons in Power Automate include a built-in Retry Policy. This dictates how the flow
behaves if a "transient" error (like a 429 Rate Limit or a 502 Gateway Error) occurs.

 Default: Typically 4 retries over a specific period.

 None: The flow fails immediately upon the first error.

 Fixed Interval: Retries at a set me (e.g., every 20 seconds).

 Exponen al Interval: Increases the me between each retry (e.g., 10s, 30s, 90s). This is
best for handling service thro ling.

2. The "Run A er" Configura on (The Logical Guard)

Every ac on in a flow has a "Run A er" se ng. This defines the condi on under which the next
ac on should execute. By default, an ac on only runs if the previous one was Successful. The
four possible states are:

1. Is successful: (Default) Run if the previous step worked.

2. Has failed: Run if the previous step encountered an error.

3. Is skipped: Run if the previous step was not executed (usually due to a condi on).

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
 4. Has med out: Run if the previous step took too long.

3. The Try-Catch-Finally Pa ern

By combining Scopes and Run A er logic, you can implement the classic programming pa ern
for error handling:

 Try: A Scope containing the main business logic.

 Catch: A Scope that runs only if the "Try" scope fails. This is used for logging errors or
sending alerts.

 Finally: A Scope or ac on that runs regardless of success or failure (e.g., closing a

connec on or upda ng a status).

Prac cal Case Study: Resilient AI Invoice Processing

The Challenge: "FinTech Solu ons" uses a flow to send invoices to an AI model for data
extrac on. Occasionally, the AI service is busy (Error 429) or the document is corrupted (Error
400). When this happens, the flow currently dies, leaving the invoice "stuck" in a pending state
without anyone knowing why.

The Solu on Design:

1. Retry Policy: The AI Extrac on ac on is set to an Exponen al Interval to handle

temporary busy signals.

2. Logic Guard: A "Catch" Scope is added. Its "Run A er" is configured to trigger if the "AI
Extrac on" fails.

3. No fica on: The "Catch" scope sends a Teams message to the finance clerk with a link
to the failed invoice.

4. Outcome: The flow no longer "disappears" on failure. The clerk is no fied of specific
data issues, while transient technical blips are resolved automa cally by the retry policy.

Step-by-Step Example: Building a "Self-Healing" Flow

This example demonstrates how to configure an ac on to handle a failure and no fy the team.

1. Set a Retry Policy:

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  Add an ac on (e.g., "HTTP" or "Dataverse - Add a row").

 Click the three dots (...) -> Se ngs.

 Under Retry Policy, select Exponen al Interval. Set Count to 5 and click Done.

2. Create the Failure Path:

 Add a "Send an email (V2)" ac on immediately a er the Dataverse ac on.

 Click the three dots (...) on the Email ac on -> Configure run a er.

 Uncheck is successful.

 Check has failed and has med out. Click Done.

3. Test the Resiliency:

 Inten onally trigger a failure (e.g., provide a null value to a required Dataverse field).

 The flow will a empt to retry based on your policy. If it s ll fails, it will bypass the
standard "Success" path and trigger the email no fica on instead.

Knowledge Check: Mul ple Choice Ques ons

1. Which Retry Policy is most effec ve for handling 'Rate Limit' (429) errors from an AI
service?

 A) None

 B) Fixed Interval

 C) Exponen al Interval

 D) Default

Correct Answer: C.

Explana on: Exponen al Interval "backs off" progressively, giving the service more me to
recover from a high volume of requests, which is the standard way to handle thro ling.

2. By default, when does an ac on in Power Automate execute?

 A) Only if the previous ac on failed.

 B) Only if the previous ac on was successful.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-
applica ons/?referralCode=857964209E5CBDF3B947
  C) Only if the flow is manually triggered.

 D) Simultaneously with the previous ac on.

Correct Answer: B.

Explana on: The default "Run A er" se ng is 'is successful.' This is why a flow stops at the first
sign of an error unless you explicitly configure an error-handling path.

3. You have a 'Try' Scope followed by a 'Catch' Scope. To ensure the 'Catch' Scope only runs if
something inside the 'Try' Scope goes wrong, which 'Run A er' se ngs should you check?

 A) is successful

 B) has failed, is skipped, and has med out

 C) has failed and has med out

 D) is skipped and is successful

Correct Answer: C.

Explana on: To catch an error, you must uncheck 'is successful' and check 'has failed' and 'has
med out.' (Checking 'is skipped' is generally used for 'Finally' blocks that must run even if a
previous error-handling step was bypassed).

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Building Intelligent Desktop Flows (RPA) for Legacy Apps

Objec ve

In this chapter, you will bridge the gap between modern AI and legacy infrastructure using
Power Automate for desktop. You will learn how to design Robo c Process Automa on (RPA)
scripts that interact with "non-API" applica ons, such as old Windows so ware, terminal
emulators, or web portals without modern connectors. By the end of this chapter, you will be
able to implement UI automa on, handle dynamic screen elements, and integrate AI Builder
models directly into desktop workflows to turn "dumb" legacy processes into intelligent
automated solu ons.

Core Concepts

Desktop flows allow you to automate anything you can see and click on a computer screen.
When combined with AI, this is o en called Intelligent Process Automa on (IPA).

1. UI Elements and Selectors

To interact with a legacy app, the desktop flow must iden fy specific bu ons or text fields.

 Selectors: These are the "coordinates" or "addresses" of a UI element (e.g., a "Submit"

bu on in an old ERP system).

 Dynamic Selectors: Some legacy apps change their element IDs every me they open.
You will learn to use variables and wildcards (*) to ensure your "Robot" can s ll find the
bu on.

2. Surface Automa on (OCR & Image Recogni on)

If a legacy app is running on a remote desktop (Citrix/RDP) or uses a non-standard UI

framework, Power Automate cannot "read" the code.

 Image Recogni on: Telling the flow to "find this picture of a save icon and click it."

 OCR (Op cal Character Recogni on): Using AI to "read" text inside a window that isn't
selectable by a mouse.

3. A ended vs. Una ended RPA

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  A ended: The flow runs on a user's worksta on while they are logged in. It acts as a
"digital assistant" to help them with repe ve tasks.

 Una ended: The flow runs on a remote server or virtual machine without human
supervision. This is ideal for high-volume, "back-office" processing.

Prac cal Case Study: The "Paper-to-Digital" Bridge

The Challenge: "Heritage Banking" uses a 30-year-old mainframe applica on for mortgage
processing. There is no API. Employees currently spend 6 hours a day manually typing data from
PDF applica ons into the green-screen mainframe.

The Solu on Design:

1. AI Extrac on: A Cloud Flow uses AI Builder Document Processing to extract data from
the PDFs.

2. Handoff: The Cloud Flow triggers a Desktop Flow, passing the extracted data (Name,
Amount, Address) as variables.

3. RPA Execu on: The Desktop Flow opens the legacy mainframe, navigates through the
screens using keyboard shortcuts (Tab, Enter), and types the variables into the correct
fields.

4. Outcome: The manual data entry was eliminated. What used to take 6 hours now takes
15 minutes of automated processing, with zero typing errors.

Step-by-Step Example: Automa ng a Legacy Login

This example demonstrates how to create a desktop flow that securely logs into an old
Windows applica on.

1. Capture UI Elements:

 Open the Power Automate for desktop designer.

 Use the UI Recorder to click on the "Username" field, the "Password" field, and the
"Login" bu on of your legacy app.

2. Set Variables:

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  Create two input variables: User_ID and Encrypted_Pass.

3. Build the Logic:

 Ac on: Populate text field in window. Map your User_ID variable to the captured
username field.

 Ac on: Send Keys. Use {Tab} to move to the next field.

 Ac on: Populate text field in window. Map your Encrypted_Pass to the password field.

 Ac on: Press bu on in window. Select the captured "Login" bu on.

4. Add Excep on Handling:

 Wrap the login steps in an On Block Error ac on. If the "Login" bu on isn't found within
10 seconds, tell the flow to send an email to the admin and stop.

Knowledge Check: Mul ple Choice Ques ons

1. When should a developer use 'Surface Automa on' (Image Recogni on) instead of
standard UI Automa on?

 A) When the applica on is a modern web browser.

 B) When the applica on is running in a virtualized environment (like Citrix or RDP) where
the underlying UI code is not accessible.

 C) When they want the flow to run faster.

 D) When they are using a MacBook.

Correct Answer: B.

Explana on: Surface automa on is a "fallback" method. Since the robot can only "see" a stream
of images in a virtual environment, it must use image matching and OCR to interact with
elements.

2. What is a 'Selector' in Power Automate for desktop?

 A) A tool that chooses which AI model to use.

 B) A hierarchical path that uniquely iden fies a UI element within an applica on.

 C) A bu on that starts the flow.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-
applica ons/?referralCode=857964209E5CBDF3B947
  D) A type of Excel formula.

Correct Answer: B.

Explana on: Selectors are the technical instruc ons that tell the RPA agent exactly which
bu on, window, or text box it needs to interact with.

3. In a 'Una ended' RPA scenario, where does the desktop flow execute?

 A) On the user's mobile phone.

 B) On a dedicated machine or virtual machine, running in the background without a

logged-in human user.

 C) Only inside a web browser.

 D) Inside a Power BI dashboard.

Correct Answer: B.

Explana on: Una ended RPA is designed for automated back-office work. It runs on a "digital
worker" (VM) and does not require a human to be present to trigger or watch the process.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Integra ng AI Builder Models into Cloud Flows

Objec ve

In this chapter, you will master the technical integra on of AI Builder models within Power
Automate Cloud Flows. You will move beyond sta c automa on to create "Cogni ve
Workflows" that can extract data, predict outcomes, and categorize informa on dynamically. By
the end of this chapter, you will be able to configure AI Builder ac ons, handle model outputs
using JSON logic, and implement sophis cated "Human-in-the-Loop" valida on for AI-generated
results.

Core Concepts

AI Builder models act as specialized "Intelligence Ac ons" within a flow. They receive raw data
(input) and return structured insights (output).

1. The Request-Response Cycle

When a Cloud Flow calls an AI Builder model, it follows a specific sequence:

 The Request: The flow sends a payload (e.g., a PDF file, an image, or a text string) to the
model.

 The Inference: The AI Builder engine processes the data based on its training.

 The Response: The model returns a JSON object containing the results (e.g., extracted
fields, confidence scores, or classifica on tags).

2. Synchronous vs. Asynchronous AI Processing

 Synchronous (Real- me): Most pre-trained models (like Sen ment Analysis) provide
immediate results. The flow waits for the AI to finish before moving to the next step.

 Asynchronous (Background): Custom document processing or complex models may take

longer. Power Automate handles this by "polling" the AI service un l the status is
"Succeeded."

3. Confidence Scores and Thresholds

AI is rarely 100% certain. Every AI Builder model returns a Confidence Score (usually a decimal
between 0 and 1).

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  High Confidence (> 0.8): The flow can proceed with fully automated updates to
Dataverse.

 Low Confidence (< 0.8): The flow should trigger a "Human-in-the-Loop" step, such as a
Microso Teams approval or an email review.

Prac cal Case Study: The "Auto-Pilot" Claims Processor

The Challenge: "Swi -Insure" processes 500 accident claims daily. Currently, staff must
manually read every claim to extract the "Vehicle ID" and "Es mate Amount," then check if the
claim looks "Suspicious." This manual triage takes 10 minutes per claim.

The Solu on Design:

1. Input: A Cloud Flow is triggered when a claim PDF is uploaded to SharePoint.

2. Extrac on: The flow uses a Custom Document Processing model to pull the Vehicle ID
and Amount.

3. Risk Analysis: A second Category Classifica on model analyzes the "Descrip on of

Accident" for keywords associated with fraud.

4. Threshold Logic:

o If Confidence is High and Risk is Low: The claim is auto-approved in Dataverse.

o If Confidence is Low or Risk is High: A Teams message is sent to an adjuster for

manual review.

5. Outcome: 70% of claims are now processed instantly without human touch, allowing
adjusters to focus only on complex or high-risk cases.

Step-by-Step Example: Extrac ng Data from an Invoice

This example guides you through using a pre-trained AI model to process an invoice received via
email.

1. Create the Trigger:

 Start an Automated Cloud Flow.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  Trigger: When a new email arrives (V3). Filter for emails with "Invoice" in the subject.

2. Add the AI Builder Ac on:

 Add a new ac on: Extract informa on from invoices.

 Map the A achments Content from the email trigger to the "Invoice file" field in the AI
ac on.

3. Parse and Filter the Output:

 The AI model will output dynamic tokens like Invoice total and Billing address.

 Add a Condi on to check the Confidence score:

o Formula: Confidence is greater than or equal to 0.85.

4. Update Dataverse:

 In the "Yes" path, add Add a new row (Dataverse) to your Invoices table.

 Map the dynamic tokens from AI Builder directly to your table columns.

5. Add Human-in-the-Loop:

 In the "No" path, add a Post an adap ve card and wait for a response (Teams) ac on.

 Send the extracted data to a clerk to verify before it is saved to the database.

Knowledge Check: Mul ple Choice Ques ons

1. Which dynamic content token from an AI Builder ac on should you evaluate to decide if a
human needs to review the AI's output?

 A) Model ID

 B) Confidence Score

 C) Request Date

 D) Input File Name

Correct Answer: B.

Explana on: The Confidence Score indicates how sure the AI is about its predic on. Using this
score in a condi on allows you to implement "Human-in-the-Loop" logic for uncertain results.
h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-
applica ons/?referralCode=857964209E5CBDF3B947
2. You are using a 'Custom Document Processing' model in a flow. What must you do if the
model returns a low confidence score for a cri cal field?

 A) Delete the record immediately.

 B) Ignore the score and trust the AI.

 C) Route the data to a human for manual verifica on using a Teams Approval or an
email.

 D) Restart the flow.

Correct Answer: C.

Explana on: For professional applica ons (especially in finance or health), high-risk decisions
based on low-confidence AI data should always be verified by a human.

3. What is the benefit of using 'Pre-trained' AI models in a Cloud Flow instead of 'Custom'
models?

 A) They can only be used in English.

 B) They require a Python script to run.

 C) They are ready to use immediately for common tasks (like reading a receipt or
analyzing sen ment) without needing you to provide training data.

 D) They are only available for Desktop flows.

Correct Answer: C.

Explana on: Pre-trained models are "off-the-shelf" solu ons that save me because Microso
has already trained them on millions of generic examples.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Chapter: Advanced Approvals with Dynamic Content and AI Insights

Objec ve

In this chapter, you will master the art of designing High-Stakes Approval Workflows. You will
move beyond simple "Approve/Reject" bu ons to create intelligent approval systems that
provide decision-makers with AI-generated summaries, sen ment analysis, and risk scores. By
the end of this chapter, you will be able to implement mul -stage parallel approvals, u lize
Adap ve Cards in Microso Teams, and ensure that human reviewers are "grounded" with the
best possible AI insights before they cast their vote.

Core Concepts

Approvals are the primary mechanism for Human-in-the-Loop (HITL) AI pa erns. In an

intelligent applica on, the AI shouldn't just ask for an approval; it should provide a "Briefing" to
the human.

1. The Approvals Connector

Power Automate provides a dedicated connector for managing the lifecycle of a request.

 Wait for an approval: The flow pauses un l a response is received.

 Start and wait for an approval: The most common ac on; it combines the crea on and
the "pause" into one step.

 Approval Types:

o Everyone must approve: Requires a unanimous vote.

o First to respond: The first person to click a bu on decides for the group.

2. Enhancing Approvals with AI Insights

To make approvals "Intelligent," you must inject data from AI Builder into the approval request:

 Summariza on: Instead of sending a 50-page document, send a 3-sentence summary

generated by an LLM.

 Sen ment Analysis: If a customer is angry, highlight this in the approval tle to indicate
urgency.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
  Risk Scoring: Use a predic on model to flag the probability of a "Bad Debt" or "Fraud"
directly in the request.

3. Adap ve Cards for Teams

Standard email approvals are o en ignored. Adap ve Cards allow you to send a rich, interac ve
UI component directly into a Microso Teams chat. They can include images, bold text, and
input fields for comments, all while remaining inside the Teams interface.

Prac cal Case Study: The "Intelligent Credit Limit" Adjuster

The Challenge: "Finance-Plus" receives 200 requests daily for credit limit increases. Managers
o en approve these blindly because they don't have me to dig through the customer's
payment history or read the account manager's long jus fica on notes.

The Solu on Design:

1. AI Prepara on: A flow triggers when a new request is filed. It uses AI Builder to:

o Summarize the Account Manager's jus fica on.

o Predict the "Payment Default Risk" based on historical Dataverse data.

2. The Approval: The flow sends an Adap ve Card to the Finance Manager.

o Title: "Urgent: Credit Increase for [Customer Name] - [Sen ment: Frustrated]"

o Body: Displays the AI summary and a "Risk Level: Low (88% Confidence)."

3. Outcome: Managers now make decisions in seconds, grounded in data they previously
ignored. Decisions are more accurate, and "Bad Debt" instances have decreased by 12%.

Step-by-Step Example: Building an AI-Enhanced Approval

This example demonstrates how to pass AI-generated summaries into a Teams approval.

1. Generate the AI Insight:

 Add an AI Builder ac on: "Summarize text."

 Input the Notes field from your Dataverse record.

2. Create the Approval:

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-
applica ons/?referralCode=857964209E5CBDF3B947
  Add the ac on: Start and wait for an approval.

 Approval Type: "Approve/Reject - First to respond."

 Title: New Request from: & Customer Name.

 Details:

o "AI Summary of Request: " & Summary (from AI Builder)

o "Original Notes: " & Notes

3. Send via Teams (Adap ve Card):

 Add the ac on: Post an adap ve card and wait for a response.

 Select Power Automate as the poster and the Manager as the recipient.

 Use the Approval Link from the previous step as a bu on in the card.

4. Handle the Outcome:

 Add a Condi on: If Outcome is equal to Approve.

 Yes: Update the Dataverse row to "Approved."

 No: Update the Dataverse row to "Rejected" and send a no fica on back to the
requester.

Knowledge Check: Mul ple Choice Ques ons

1. What is the primary benefit of including 'AI Insights' (like a summary or risk score) in an
approval request?

 A) It makes the email look more colorful.

 B) It provides the human approver with "Grounded Context," allowing them to make a
faster and more informed decision without manually searching for data.

 C) It allows the AI to approve the request automa cally without a human.

 D) It requires a more expensive Teams license.

Correct Answer: B.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947
Explana on: The goal of an intelligent app is to augment human intelligence. Providing a
summary or risk score ensures the approver has the "Key Facts" front-and-center.

2. If you need a request to be reviewed by three different managers and only want it to pass if
ALL THREE say 'Approve,' which approval type should you choose?

 A) First to respond

 B) Everyone must approve

 C) Custom Responses - Wait for all responses

 D) Scheduled Approval

Correct Answer: B.

Explana on: 'Everyone must approve' is a unanimous vote. If even one person rejects, the
overall outcome becomes 'Reject.'

3. Why would a developer use an 'Adap ve Card' in Microso Teams instead of a standard
Power Automate Approval email?

 A) Adap ve Cards are the only way to use AI.

 B) Adap ve Cards allow for a custom, rich UI and allow users to respond to the approval
without ever leaving their Teams chat.

 C) Emails are no longer supported in Power Automate.

 D) Adap ve Cards are faster to build.

Correct Answer: B.

Explana on: Adap ve Cards improve the "User Experience" by keeping the work in the flow of
the user's communica on, leading to faster response mes and be er engagement.

h ps://[Link]/course/microso -ab-410-cer fica on-intelligent-

applica ons/?referralCode=857964209E5CBDF3B947