2.1 Analyzing Requirements and Identifying Components

Every intelligent application starts as a pile of business requirements, and the AB-410 exam tests whether you can translate those requirements into the right Microsoft Power Platform components with the least custom code. The design domain carries roughly 25-30% of the exam, so build a repeatable method: decompose the requirement, identify the personas, choose the app or automation type, pick the data platform, and only then decide where AI genuinely adds value.

The Component Toolbox

Power Platform gives makers five primary building blocks. Learn the signal phrases that map a requirement to each one:

Canvas apps — when the experience is the requirement

Canvas apps give you pixel-level control in Power Apps Studio at make.powerapps.com. Choose a canvas app when the requirement emphasizes a tailored interface: mobile-first field apps using the camera, GPS, and barcode scanner controls, kiosk screens, or branded task flows. Canvas apps can start from data, from a blank screen, from a Figma file, or from an image or natural-language description through AI-assisted app creation. Because canvas apps connect to more than 1,000 connectors as well as Microsoft Dataverse, they are the right choice when data must stay in an external system such as SQL Server or Salesforce. Watch delegation: queries the data source cannot execute itself only process the first 500 records by default (configurable up to 2,000) — a classic exam trap for apps built on SharePoint lists, which also enforce a 5,000-item list view threshold.

Model-driven apps — when the data is the requirement

Model-driven apps generate a responsive interface automatically from your Dataverse tables, views, forms, charts, and dashboards. Choose one when the requirement centers on structured data and process: case management, opportunity pipelines, inspection back offices. You get business process flows, business rules, and Dataverse row-level and column-level security essentially for free. The hard rule: model-driven apps require Dataverse. If the requirement says data must remain in an on-premises SQL database with no replication, a model-driven app alone cannot satisfy it — pair a canvas app with the SQL connector, or project the external data into Dataverse as virtual tables without copying it.

Power Pages — when the audience is external

Power Pages is the answer whenever the requirement mentions customers, partners, vendors, or anonymous public users. A Power Pages site is an external-facing, low-code website — built in the Power Pages design studio — that reads and writes Dataverse data through lists, basic forms, and multistep forms, secured with table permissions and web roles. Internal employees should never receive a Power Pages site for a back-office process, and external customers should never receive a model-driven app; the exam swaps these two constantly.

Cloud flows — when nobody should be in the loop

Cloud flows in Power Automate carry the automation layer. Memorize the three types: automated flows fire on an event (a new Dataverse row, an arriving email), instant flows are triggered manually from a button or an app, and scheduled flows run on a recurrence. Approvals landing in Teams or Outlook are cloud flows, not apps.

Choosing the Data Platform

Data placement drives almost every downstream decision:

• Choose Dataverse when you need a relational schema, row-level and column-level security, auditing, offline capability, application lifecycle management (ALM) through solutions, or server-side business logic.
• Choose connectors when the data must remain at its source because another team owns the system of record and needs real-time fidelity.
• Choose virtual tables when you want Dataverse behaviors (security, model-driven UI, relationships) over data that physically lives outside Dataverse.

Remember that Dataverse, SQL Server, and all custom connectors are premium, so data choices also carry a licensing signal the exam expects you to notice.

Where AI Adds Value

Identify the AI insertion point by asking where a human currently reads, decides, or retypes:

• Generation — drafting emails, record descriptions, or entire app tables from natural language (Copilot in Power Apps).
• Summarization — condensing case histories or long text columns into digestible briefs.
• Prediction and classification — AI Builder models that score likelihood or categorize text and images.
• Extraction — AI Builder document processing for invoices, receipts, business cards, and identity documents.

AI is a component choice like any other: pick it because the requirement demands it, not because it is fashionable. A scenario that only needs CRUD screens earns no extra credit for an agent.

Mapping Personas to Apps

Strong exam answers start from the persona, not the technology. Consider a requirement that reads: field technicians must capture equipment photos and meter readings on phones, service managers must triage and schedule work, customers must check ticket status without calling, and finance must approve discounts. Decomposed correctly, that is a canvas app for technicians (camera, offline-friendly controls), a model-driven app for managers (Dataverse tables, business process flow, dashboards), a Power Pages site for customers (table permissions scoped to their own tickets), and an approval cloud flow for finance. One requirement, four components — the exam rewards exactly this decomposition.

Exam Traps to Avoid

• Proposing a model-driven app when the data cannot move into Dataverse.
• Giving external users anything other than Power Pages.
• Ignoring the 500/2,000 delegation limit on non-delegable queries.
• Forgetting that premium connectors (SQL, Dataverse, custom) require premium licensing for every user of the app.
• Over-engineering: recommending a full Dataverse migration when a connector satisfies the requirement as written.