Q1.An engineer is developing a chatbot that must remember previous exchanges throughout a single user session.
Which component is responsible for maintaining the conversational context?
Show answer
The correct answer is A.
Conversation state management retains the most recent exchanges and passes them on as context for the next response.
A large language model (LLM) does not automatically retain memory between requests, so it does not remember previous statements unless the context is provided explicitly.
An image recognition model, data compression, and a firewall are all unrelated to conversational memory.
Proper state management is essential for establishing a natural multi-turn conversation.
Handle conversation state with variables in Copilot Studio
Q2.You have set up a parent table and a child table and configured a one-to-many (1:N) relationship from the parent to the child.
When the owner of a parent record is changed, you want all child records beneath it to be automatically reassigned to the new owner as well.
How should you configure the behavior type of the relationship?
Show answer
The correct answer is D.
To automatically propagate a parent record’s owner change to all of its child records, use the Parental behavior type for the relationship.
In a parental relationship, operations such as assign, share, and delete cascade from the parent to the children.
Alternatively, using Configurable Cascading to set “Assign” to “Cascade All” produces the same result.
Referential and Restrict are inappropriate because an owner change does not cascade to the children with them.
Create and edit one-to-many relationships
Q3.In a model-driven app, you are configuring a mechanism to reliably ensure data quality.
The app is required to behave as follows.
When a certain condition is met, make input in a particular column mandatory.
Prevent the record from being saved until that column is filled in.
Implement this with no code, using no code or automation.
You need to configure a business rule that satisfies these requirements. What should you do?
Show answer
The correct answer is D.
When a no-code business rule sets the requirement level to “Business Required,” the column becomes mandatory the moment the condition becomes true, a red asterisk is displayed, and saving is blocked until it is filled in.
This satisfies all three requirements.
Showing a recommendation (A) does not prevent saving, and setting visibility (B) or a default value (C) does not serve to make a field mandatory.
Create a business rule
Q4.A property and casualty insurance company is building a customer service solution on Microsoft Dataverse to manage customer interaction records and inquiry resolution.
The data model includes tables for customer information, support cases, and agents.
You need to configure the relationships between the tables to meet the following requirements.
Associate each support case with a single customer.
Retain information about a customer’s dependents (such as children).
Record the history of agents assigned to a customer.
Select the appropriate relationship for each requirement.
Each correct answer is worth one point.
Drag each option to the matching requirement’s drop zone, or tap to place it. Click a filled zone to send it back. (The same option may be used more than once.)
Connection
One-to-many
Many-to-many
Show answer
The correct answers are Case → One-to-many / Dependent information → One-to-many / Agent history → Connection.
Since one customer can have multiple support cases and multiple children, while each case and each child belongs to a single customer, a one-to-many (1:N) relationship is appropriate for both “support cases” and “dependent information.”
A relationship that changes over time in a many-to-many fashion and that you want to carry additional information such as start and end dates or roles—like the history of agent assignments—is expressed with Connections.
A distinctive feature of connections is that they can retain both current and past relationships as history.
Create and edit table relationships
Q5.A service provider is running a Power Automate cloud flow in a production environment.
Reviewing the flow’s run history, you find numerous duplicate runs and 401 error codes occurring.
You need to resolve these errors.
Select the appropriate remedy for each error.
Each correct answer is worth one point.
Show answer
The correct answers are 401 error → Reauthenticate the connection / Duplicate runs → Enable concurrency control.
A 401 (Unauthorized) is caused by an expired authentication token or insufficient permissions, so it is resolved by reauthenticating the relevant connection.
Duplicate runs occur due to consecutive requests from upstream systems or double-firing of a trigger.
Enabling concurrency control and setting the limit to 1 makes the flow process one item at a time in order, preventing duplicates and conflicts.
Extending the timeout or configuring a retry policy does not address the root cause of an authentication failure or duplication.
Troubleshoot flow runs
Q6.In a canvas app, you are building a process that classifies images submitted by users using an existing AI Builder model.
The app must return predictions in real time while the user is interacting with it.
How should you natively consume this AI model?
Show answer
The correct answer is A.
To natively use an AI Builder model in a canvas app and generate predictions during interaction, add the AI model as a data source and call it with Power Fx’s Predict function (or a model-specific function such as AIClassify).
This lets you dynamically classify an image at a moment such as a button press.
Replacing it with a static formula (B), retraining (C), and exporting the solution (D) do not fulfill the goal of generating predictions at run time.
Use AI Builder in Power Apps
Q7.You have built an AI assistant for marketing support.
Users report that “even when I send exactly the same request, the text that comes back varies greatly every time.”
Which parameter should you adjust to increase the consistency of the responses?
Show answer
The correct answer is A.
Temperature is the parameter that controls the randomness of generative AI output; the lower the value, the more consistent and predictable the responses become.
Conversely, raising the value increases creativity and variability.
Storage capacity, network bandwidth, and authentication method do not affect the model’s generative behavior itself.
When you want to tune response consistency, lowering Temperature is the standard approach.
Azure OpenAI parameters and prompt design
Q8.A company runs a Power Platform solution that includes prompts, cloud flows, and Dataverse components.
Multiple development teams are each responsible for a separate feature.
Each team has the following requirements.
Be able to develop and update each feature of the same solution independently, team by team.
Ensure that one update does not overwrite another team’s changes.
Ensure that production deployment is controlled and predictable.
In line with Microsoft best practices, recommend an ALM approach that supports independent development and controlled deployment. What should you recommend?
Show answer
The correct answer is C.
Separating the development environment for each team prevents accidents where shared components are overwritten and enables independent development and testing.
Furthermore, deploying to production as a managed solution achieves a controlled and predictable deployment.
A single environment or making changes directly in production (A, B, D) invites overwrite risk and disorderly deployment, which is contrary to Microsoft’s recommendation.
Power Platform ALM overview
Q9.A company is developing an AI chatbot using Azure OpenAI.
The security administrator requires that “customer data must never be used to train public foundation models.”
Which statement correctly describes how Azure OpenAI handles data?
Show answer
The correct answer is B.
Azure OpenAI provides enterprise-grade data protection, and customer prompts, responses, and training data are never used to train OpenAI’s public models.
Data is isolated within the organization’s resources and separated from the public model improvement process.
That prompts are automatically used for training (A), shared externally (C), and that fine-tuning data becomes publicly searchable (D) are all incorrect.
Azure OpenAI data privacy
Q10.A company is developing an AI solution that extracts information from images, understands text, and generates responses.
This architecture should combine multiple AI capabilities through managed services.
Which is the recommended design approach?
Show answer
The correct answer is A.
In modern intelligent apps, rather than relying on a single capability, a composable design that combines specialized AI services optimized for each use case is recommended.
For example, Vision services handle images, Language services handle text understanding, and generative models handle response generation.
This composition excels in scalability, maintainability, and flexibility.
A single unmanaged model (B), avoiding APIs (C), and permanently storing everything in the prompt (D) are all inappropriate.
What are Azure AI services