AIAI Agent Dev Labaiagentdev-lab.com
Sign in to save progressSign in
Menu

AB-100 · LESSON 1

Plan AI-powered business solutions25–30% · 3 topics

Learn to analyze requirements for AI-powered business solutions, design an overall AI strategy, and evaluate costs and benefits.

1.1Analyze requirements for AI solutions

Analyze requirements for AI-powered business solutions

Concept

Requirements analysis clarifies the business problem, users, success metrics, data and constraints (security, regulation, budget, timeline), and decides whether AI should solve it and which kind of AI (generative, agents, predictive, traditional automation) fits.

How it works / steps

  1. Interview stakeholders and visualize the business flow and problems
  2. Define success metrics (KPIs) and current baselines
  3. Confirm where usable data lives, its quality and its permissions
  4. Organize regulatory, security and data residency requirements
  5. Separate the parts AI suits from parts where rules or traditional automation are enough

Workplace example

Finding that 60% of order inquiries are routine, propose a design where an agent handles routine ones and staff handle exceptions, with handling time as the KPI.

Common confusions and exam tips

Mini quiz

Q1. At a management meeting, you are considering “which work to start using agents for.” Which is the most appropriate first evaluation axis?

  1. Compare model and license costs per task and start with the work that is cheapest to adopt
  2. Collect requests from each department and start developing agents for the work with the most requests
  3. Classify work into task automation, data analysis and decision support, and evaluate scope of impact and need for human involvement
  4. List all work by usage frequency and replace the most used work with agents first to see effect and cost
Show answer and explanation

Answer: C Classify work into task automation, data analysis and decision support, and evaluate scope of impact and need for human involvement

Set priorities by use type and scope of impact. Design decision support on the premise that people make the final decision. [Other options] Usage frequency, request counts or cost alone do not reveal scope of impact or need for human involvement. Evaluate first by task type, impact and human involvement.

Q2. In which situation should creating a custom AI model be considered?

  1. When evaluation confirms prebuilt models, prompts and RAG cannot meet requirements (accuracy, format, cost and so on)
  2. When the company has AI specialists and wants them to gain model development experience
  3. When business data is not yet organized and you want model training to compensate for quality
  4. Before trying prebuilt models, when you think a proprietary model specialized in your data would be a competitive advantage
Show answer and explanation

Answer: A When evaluation confirms prebuilt models, prompts and RAG cannot meet requirements (accuracy, format, cost and so on)

Custom models carry high cost and operational burden. Consider them after confirming prebuilt options fall short. [Other options] Deciding before evaluating prebuilt models is premature. Staff development is not a reason to adopt. Training on unorganized data does not improve quality.

Reference links

1.2Design the overall AI strategy

Design overall AI strategy for business solutions

Concept

An AI strategy is a plan for which AI goes into which work, in what order and on which platform. Combine Microsoft 365 Copilot (ready-to-use productivity), Copilot Studio (low-code agents), Microsoft Foundry (pro-code development) and built-in Dynamics 365 agents from a buy, build and extend perspective.

How it works / steps

  1. Link business goals to AI initiatives
  2. Prioritize use cases by impact and feasibility
  3. Buy where off-the-shelf products suffice; build where differentiation is needed
  4. Decide a common foundation for data platform, governance and security
  5. Include skills development and adoption support in the plan

Workplace example

Set a three-layer strategy: Microsoft 365 Copilot for company-wide productivity, Copilot Studio agents for departmental work, and Foundry for a proprietary demand forecast.

Common confusions and exam tips

Mini quiz

Q1. Which framework should be referenced to progress company-wide AI adoption in stages?

  1. The reliability pillar of Power Platform Well-Architected
  2. The AI adoption process in the Cloud Adoption Framework for Azure
  3. The AI workload guidance in the Azure Well-Architected Framework
  4. Only the Microsoft 365 Copilot deployment guide
Show answer and explanation

Answer: B The AI adoption process in the Cloud Adoption Framework for Azure

The Cloud Adoption Framework is the official framework organizing strategy, planning, readiness and governance in stages. [Other options] Well-Architected guides workload design quality, not stages of company-wide adoption. The Copilot deployment guide is a product rollout procedure.

Q2. You want to add knowledge of internal application procedures to Microsoft 365 Copilot, which employees use for email and document writing. Which decision is most appropriate?

  1. Fine-tune your own model in Azure to learn the application procedures
  2. Build a separate chatbot in Copilot Studio and offer it apart from Microsoft 365 Copilot
  3. Put the procedure documents in OneDrive and guide employees to search and refer to them
  4. First consider whether extending Microsoft 365 Copilot (agents, added knowledge) is enough
Show answer and explanation

Answer: D First consider whether extending Microsoft 365 Copilot (agents, added knowledge) is enough

If existing Microsoft 365 Copilot can be extended, it is faster and cheaper than custom builds. [Other options] A separate bot falls outside the daily Copilot flow. Training your own model is excessive. Guiding document searches adds no knowledge to Copilot.

Reference links

1.3Evaluate the costs and benefits of AI solutions

Evaluate the costs and benefits of an AI-powered business solution

Concept

Costs include licenses (Copilot, Copilot Studio capacity), Azure consumption (tokens, search, compute), and development, operations and training. Benefits are measured by time saved, quality improvements and effects on revenue or customer satisfaction, and judged by ROI and total cost of ownership (TCO).

How it works / steps

  1. Measure pre-adoption baselines (work time, volume, quality)
  2. Estimate license, consumption, development and operating costs
  3. Convert benefits to money and calculate ROI and payback period
  4. Measure in a pilot and update the estimate
  5. Also record qualitative benefits such as employee satisfaction and risk reduction

Workplace example

Expecting an inquiry agent to save 400 hours a month, estimate payback in 8 months after usage fees and development costs, and verify it in a pilot.

Common confusions and exam tips

Mini quiz

Q1. When evaluating the ROI of an AI solution, what should be included in TCO?

  1. Licenses, initial development costs and the cost of rollout briefings for user departments
  2. License and development costs plus the labor-cost equivalent of work time saved
  3. Licenses, development and operating costs, plus effort for human review and exception handling
  4. Only development costs and the cost of training user departments
Show answer and explanation

Answer: C Licenses, development and operating costs, plus effort for human review and exception handling

TCO includes ongoing costs such as human review effort, not just initial and operating costs. [Other options] Time saved is on the benefit side, not TCO. Licenses and initial costs alone miss operations and human review effort.

Q2. For accounting’s invoice processing, you are considering buying an off-the-shelf SaaS, building in Copilot Studio or extending existing Dynamics 365 features. Which approach is appropriate?

  1. Have user department staff try the three options and choose the one that is easiest to use
  2. Compare the three quotes, choose the one with the lowest initial cost and adjust requirements after adoption
  3. Choose the option that uses the most of the latest generative AI features to prepare for future expansion
  4. Compare build, buy and extend by requirement fit, TCO, implementation time and maintainability
Show answer and explanation

Answer: D Compare build, buy and extend by requirement fit, TCO, implementation time and maintainability

Build / buy / extend decisions compare requirements, total cost and maintainability. [Other options] Initial cost alone does not reveal requirement fit or TCO. Many features do not indicate requirement fit. Usability alone cannot compare maintainability or implementation time.

Reference links

Lessons are written by this site based on public information and do not guarantee exam content. Product specifications change, so check the latest information at the reference links.

Notes (unofficial, original questions, no pass guarantee)