8.1 Map Processes to Microsoft AI Apps and Services
Key Takeaways
- Microsoft's July 22, 2026 AB-731 skills list includes "Map business processes and use cases to Microsoft’s AI apps and services": leaders match the work, not a favorite product logo.
- Everyday knowledge work in Word, Excel, Teams, Outlook, and Copilot Chat maps to the Copilot family; Researcher and Analyst are first-party Copilot agents for deep research and numeric analysis, not replacements for vision or search platforms.
- Microsoft Copilot Studio is the low-code studio for custom agents and workflows; Microsoft Graph is the permissioned gateway to Microsoft 365 work data and the path Copilot connectors use to bring external systems in.
- Azure Vision in Foundry Tools maps to image and optical character recognition (OCR) work; Azure AI Search maps to grounding, hybrid retrieval, and chat-with-your-data; Microsoft Foundry is the custom-agent and model platform when Copilot as shipped cannot run the process.
- A vice president can use a decision table: start with Copilot where users already work, extend with Studio and Graph when knowledge or actions are missing, and choose Foundry Tools when the process is vision, enterprise search, or a custom model.
8.1 Map Processes to Microsoft AI Apps and Services
Quick Answer: Start with the business process, then pick the Microsoft capability that already sits closest to the user’s work and the data the process needs. The Copilot family covers everyday Microsoft 365 work. Microsoft Copilot Studio covers custom agents. Microsoft Graph covers permissioned work data and connectors. Researcher and Analyst cover deep research and numeric analysis inside Copilot. Foundry Tools—especially Azure Vision in Foundry Tools and Azure AI Search—plus Microsoft Foundry cover vision, enterprise retrieval, and custom applications Copilot does not ship.
Microsoft’s AB-731 study guide (skills measured as of July 22, 2026) lists two related mapping skills: map processes to Copilot, and map processes to Microsoft’s AI apps and services. This section is the broader map. Chapter 7 already treated Copilot-in-apps. Here a transformation leader must still know Copilot, but must not force a warehouse photo pipeline, a multi-million-document knowledge base, or a specialty model into Copilot Chat just because Copilot is familiar.
AB-731 is a business-decision-maker exam. You are not asked to write an application programming interface (API) call. You are asked to sit with a chief financial officer (CFO), an operations vice president, human resources (HR), legal, and security and name a first-choice service, a wrong-fit service, and the handoff when the process outgrows what Copilot ships.
Read the process before you read the product sheet
Ask four leader questions before anyone opens a portal:
- Where does the work already happen? If the user lives in Word, Excel, Teams, or Outlook, Copilot in those apps is the default. If the work is a website, a factory camera, or a custom line-of-business (LOB) app, Copilot may be the wrong front door.
- What must the system see? Mail, files, meetings, and people point to Graph. Spreadsheets and CSV files point to Analyst. Public web plus work files point to Researcher. Images and printed labels point to Azure Vision. A large proprietary corpus that must ground a chatbot points to Azure AI Search.
- Does the process need an action, not only an answer? Summarizing is Copilot. Creating a ticket, updating an order, or running a workflow is Copilot Studio, plugins, or a Foundry agent with tools.
- Is this a one-off knowledge-worker task or a reusable operating process? A single brief is Copilot. A repeating help desk, onboarding, or intake process is Studio or Foundry.
VP decision table: first-choice Microsoft AI map
Use this table in an intake workshop. It is independent OpenExamPrep teaching that uses Microsoft’s current product names from the July 22, 2026 blueprint and Microsoft Learn. It is not a claim that OpenExamPrep is approved by Microsoft.
| Business process (examples) | First-choice Microsoft capability | Why it maps | Wrong fit if you stop here |
|---|---|---|---|
| Draft, summarize, and catch up inside Word, Excel, PowerPoint, Outlook, Teams | Copilot family: Microsoft 365 Copilot in apps and Copilot Chat | The user is already in Microsoft 365; Graph can ground work answers on content the signed-in user may see | Vision on photos; custom RAG over millions of blobs; write-back to an unsourced LOB system |
| Deep, multi-step research that must mix the web and work files, with citations and a shareable report | Researcher agent in Copilot | Microsoft documents Researcher for complex research across web and work content, with structured, source-cited reports | Cleaning messy workbooks; OCR; a custom model you host |
| Clean, trend, and reason over numbers in Excel, CSV, or similar tables | Analyst agent in Copilot | Microsoft Support describes Analyst as a data-analysis assistant; the Researcher FAQ points Excel-style work to Analyst | Narrative competitive research that needs web plus meetings |
| A reusable assistant with your instructions, SharePoint or connector knowledge, and tools | Microsoft Copilot Studio (often as a declarative agent in Microsoft 365 Copilot) | Studio is Microsoft’s graphical low-code studio for agents and workflows published into Copilot, Teams, and other channels | Specialty models, custom orchestrators, or hosting outside Copilot’s model |
| Permissioned mail, files, calendar, Teams, people; or bringing Salesforce, ServiceNow, Jira, Confluence into Microsoft 365 | Microsoft Graph plus Microsoft 365 Copilot connectors (formerly Microsoft Graph connectors) | Graph is the gateway to Microsoft 365 data; connectors index or federate external content into Graph-powered experiences including Copilot and Microsoft Search | Image pixels; a Foundry-only chatbot that never needs Microsoft 365 |
| Read text from images, detect objects or faces, caption media, digital asset management | Azure Vision in Foundry Tools | Microsoft’s Vision overview lists Face, image analysis, and OCR / Read for printed and handwritten text | Spreadsheet analysis; Copilot email summary |
| Ground agents and chat apps on enterprise or web content with full-text, vector, hybrid, or agentic retrieval | Azure AI Search | Microsoft positions Azure AI Search as the retrieval layer that connects data to models and underpins Foundry IQ knowledge bases | A one-page Word draft that Copilot in Word can already write |
| Custom agents, a catalog of models, hosted code, tracing, and enterprise control when Copilot extensibility is not enough | Microsoft Foundry and the broader Foundry Tools family | Foundry unifies agents, models, and tools; Microsoft’s Copilot extensibility FAQ treats Azure AI services / Foundry as the path for bespoke apps, not for extending Copilot’s own orchestrator | Everyday Copilot in Outlook |
Keep the names straight. Microsoft’s Foundry documentation records the rename Azure AI Studio / Azure AI Foundry → Microsoft Foundry and Azure AI Services → Foundry Tools. Older Learn pages still say Azure AI Foundry. On AB-731, prefer the blueprint names Microsoft Foundry, Foundry Tools, Azure Vision in Foundry Tools, and Azure AI Search.
How the pieces relate, not compete
Leaders fail the mapping skill when they treat the stack as a single chatbot with extra stickers. The pieces compose:
- Copilot is the productivity surface. It uses Graph to see work content the user can already open.
- Researcher and Analyst are first-party agents inside Copilot. They do not replace Studio, Vision, or Search.
- Copilot Studio is how you add a named assistant with your knowledge and actions, including agents that run in Microsoft 365 Copilot.
- Graph and Copilot connectors widen what Copilot, Search, and Researcher can see. Researcher FAQ states Researcher can fetch data from Graph connectors.
- Azure AI Search is the retrieval engine for custom RAG and many Foundry agents. Copilot’s Semantic Index is a Microsoft 365 grounding path, not a substitute for a Foundry search service you own.
- Azure Vision turns pixels into text and tags. Those tags often land in Search so the asset library is queryable.
- Microsoft Foundry is where you choose models, host agents, and apply platform controls when the process is an application, not a Copilot conversation.
Leader scenarios: operations, finance, HR, legal, security
Operations VP (warehouse). Associates photograph damaged cartons. The process is “read the label and file the photo,” not “write a memo.” Map Azure Vision (OCR) and, if the library must be searched by logo or object, Azure AI Search. Copilot in Outlook can still draft the carrier email after the vision step.
CFO (close commentary). Controllers live in workbooks. Map Analyst to clean and trend the close file, Copilot in Excel for in-grid help, and Copilot in Word for the narrative. Do not map the messy workbook to Researcher. Microsoft’s Researcher FAQ says Analyst is better suited for Microsoft Excel-related tasks.
HR (policy questions). If the need is “answer from the published handbook the employee may see,” start with Copilot work grounding and, if you need a dedicated assistant, a Studio / declarative agent on SharePoint. If you are building a custom policy chatbot outside Copilot, map Azure AI Search plus Foundry. Legal still owns citations and human review.
Legal (market-entry brief). Counsel needs a cited synthesis of internal strategy decks, email, meetings, and public regulation. That is Researcher, not a 30-second Copilot chat, and not Vision.
Security. Mapping is also a boundary choice. Graph honors existing permissions. A custom Search index you copy without security trimming is a different risk. Foundry adds identity, role-based access control (RBAC), and network controls you must staff. Copilot extensibility that stays on Copilot’s orchestrator inherits Microsoft 365 compliance patterns; a custom engine you host does not get that inheritance for free.
A five-step mapping workshop you can run without code
- Name the process in a verb phrase. “Approve a vendor exception,” “find the current refund rule,” “read lot numbers from photos,” “prep a board research pack,” “trend monthly recurring revenue.” Product names come second.
- Inventory the artifacts. Files, mail, meetings, workbooks, photos, ticket systems, public web. Each artifact class points at a row in the VP table.
- Decide whether Copilot as shipped is enough. If licensed users can finish the job in Copilot Chat, in-app Copilot, Researcher, or Analyst, buy that path (section 8.3). If they need extra knowledge or actions, extend. If they need a different model, orchestrator, or application, build on Foundry.
- Name the owner and the control. HR owns handbook sources. Finance owns the close file. Security owns whether web search is on. Studio and Foundry agents need an operating owner, not an orphan demo.
- Write the anti-map. Explicitly list what you will not use. Example: “We will not put carton OCR in Researcher.” Anti-maps stop budget from collapsing onto the last vendor demo.
What this skill is not
This skill is not “always Copilot.” It is not “always Foundry.” It is not a coding lab. Microsoft’s audience profile for AB-731 says candidates are not expected to write any code. The exam still expects you to recognize Azure Vision, Azure AI Search, and Microsoft Foundry by the jobs they do, because the same skill area also asks you to identify Foundry Tools capabilities.
If two services could work, pick the one that preserves permissions, reduces new hosting, and meets the user where they already work—then escalate only when that choice cannot see the data or take the action.
Official pages opened for this section
- Study guide for Exam AB-731 — skills Map business processes and use cases to Copilot and Map business processes and use cases to Microsoft’s AI apps and services (July 22, 2026)
- What is Researcher Agent in Microsoft Copilot?
- Microsoft Copilot Researcher agent FAQ
- Get started with Analyst in Microsoft Copilot
- Microsoft Graph overview
- Overview of Microsoft Copilot Studio
- What is Azure Vision in Foundry Tools?
- Introduction to Azure AI Search
- What is Microsoft Foundry?
A manufacturing vice president wants associates to photograph warehouse labels, extract printed lot numbers, and later search the photo library by logo and object. Which mapping is the sound first choice?
Which Microsoft capability is the permissioned gateway to mail, files, meetings, and people in Microsoft 365, and the path Copilot connectors use to bring external line-of-business content into Microsoft 365 experiences?
A customer-support vice president needs a chatbot that grounds answers in a large proprietary knowledge corpus using hybrid retrieval, not a one-off Word draft. Which map should you teach?