Azure AI Services: Vision, Search & Beyond

Specialised AI for specific problems

Service-by-service breakdown

Azure AI Vision

Vision gives AI the ability to understand images and video.

Capability	What It Does	Business Example
Image analysis	Describes image content, detects objects, reads text	Product catalogue automation — AI tags product photos
OCR	Extracts printed and handwritten text from images	Digitise paper forms, scan receipts, read whiteboards
Object detection	Identifies and locates specific objects in images	Retail shelf monitoring — detect out-of-stock products
Face detection	Detects faces (not identification — privacy-safe)	Count foot traffic in retail without identifying individuals
Custom Vision	Train custom image classifiers	Manufacturing quality control — detect defective products. Note: Microsoft is migrating away from Custom Vision — new projects should use Florence/Azure AI Vision model customisation in Foundry.

Azure AI Speech

Speech bridges the gap between human voice and digital text.

Capability	What It Does	Business Example
Speech-to-text (STT)	Converts spoken audio to text	Meeting transcription, call centre analysis
Text-to-speech (TTS)	Converts text to natural-sounding voice	Voice interfaces, accessibility, automated announcements
Speech translation	Real-time translation of spoken language	Multilingual meetings without interpreters
Speaker recognition	Identifies who is speaking	Call centre authentication by voice

Azure AI Language

Language processes and understands text at scale.

Capability	What It Does	Business Example
Sentiment analysis	Determines positive, negative, or neutral tone	Monitor customer feedback across reviews and social media
Named entity recognition	Identifies people, places, organisations, dates	Extract key information from contracts and legal documents
Key phrase extraction	Pulls out the most important terms from text	Summarise support tickets to identify trending issues
Text summarisation	Condenses long text into key points	Auto-summarise lengthy reports or articles

Azure AI Translator

Translator is a separate service from Azure AI Language, purpose-built for text translation.

Capability	What It Does	Business Example
Text translation	Translates text between 100+ languages	Multilingual documentation, global communication
Document translation	Translates entire documents while preserving formatting	Localise contracts, manuals, and marketing materials
Custom Translator	Train custom translation models with your terminology	Industry-specific translations (legal, medical, technical)

Azure AI Search

AI Search is the enterprise search engine that powers intelligent information retrieval.

Capability	What It Does	Business Example
Full-text search	Traditional keyword search across indexed content	Search across millions of documents
Vector search	Semantic search that understands meaning, not just keywords	”Find policies about flexible working” finds results even if they don’t contain those exact words
Semantic ranking	Re-ranks results by relevance using AI	Most relevant document appears first, even in large result sets
RAG pattern	Retrieves relevant documents to ground AI responses	Copilot or Foundry app answers questions using YOUR documents as evidence
Multi-source indexing	Indexes content from SharePoint, Blob Storage, SQL, and custom sources	One search experience across all your data

ℹ️ What is RAG and why does it matter?

RAG (Retrieval-Augmented Generation) is the pattern that connects AI models to your data:

1. User asks a question
2. AI Search retrieves the most relevant documents from your data
3. Those documents are sent to the AI model as context
4. The model generates an answer grounded in YOUR data, not just its training knowledge

RAG is how enterprise AI avoids hallucination — by grounding responses in real documents. Azure AI Search is the most common retrieval engine for RAG in the Microsoft ecosystem.

Question

What is Azure AI Search and what role does it play in the RAG pattern?

Click or press Enter to reveal answer

Answer

Azure AI Search is an enterprise search service that indexes content from multiple sources. In the RAG (Retrieval-Augmented Generation) pattern, AI Search retrieves the most relevant documents for a user's question, which are then sent to the AI model as context — grounding the response in real data instead of general knowledge.

Click to flip back

Question

Name the four main Azure AI Services categories.

Click or press Enter to reveal answer

Answer

1) Vision — image analysis, OCR, object detection. 2) Speech — speech-to-text, text-to-speech, translation. 3) Language — sentiment analysis, entity recognition, summarisation. 4) Translator — text translation across 100+ languages. 5) AI Search — enterprise search, vector search, RAG retrieval. Plus Document Intelligence for extracting data from forms and invoices.

Click to flip back

When to use Azure AI Services vs Copilot vs Foundry

Need	Azure AI Services	Copilot for M365	Microsoft Foundry
Image analysis and OCR	Yes — Vision API	No	Can integrate Vision APIs
Speech transcription	Yes — Speech API	Teams has built-in transcription	Can integrate Speech APIs
Sentiment analysis at scale	Yes — Language API	No	Can integrate Language APIs
Enterprise document search	Yes — AI Search	Copilot Chat uses Graph (not Search)	Uses AI Search for RAG
Draft emails and documents	No	Yes — built-in	No
Custom AI model training	No (pre-built APIs)	No	Yes — fine-tuning and custom models
Meeting summaries	No	Yes — Teams Copilot	No

The key insight: Azure AI Services, Copilot, and Foundry are complementary. A single solution might use all three:

• Copilot for everyday productivity
• AI Search for document retrieval (RAG)
• Vision for image processing
• Foundry for custom model orchestration

Question

How do Azure AI Services, Copilot, and Foundry work together?

Click or press Enter to reveal answer

Answer

They are complementary layers. Copilot handles everyday M365 productivity. Azure AI Services provide specialised capabilities (vision, speech, language, search). Foundry orchestrates custom AI applications that may use both AI Services and custom models. A single enterprise solution often uses all three.

Click to flip back

🔄 PacificSteel uses Vision for quality inspection

Tomás, DT Lead at PacificSteel Manufacturing, is piloting AI-powered quality inspection on the production line.

The problem: Steel sheets are inspected manually. Inspectors check 500 sheets per shift and miss about 3% of defects. A missed defect costs $12,000 in rework and customer returns.

The solution: Azure AI Vision with custom image classification.

1. Training: Tomás’s team photographs 10,000 steel sheets — 8,000 good, 2,000 with various defects (scratches, pitting, discolouration). They train a custom vision model in Azure AI Services.
2. Deployment: Cameras on the production line capture images of every sheet. The Vision API classifies each sheet in real time.
3. Integration: Defective sheets are automatically flagged and diverted. The system logs defect types and locations for trend analysis.

Results after 90 days (in this scenario):

• Defect detection rate: improved from 97% to over 99%
• Time per inspection: under 1 second (vs 45 seconds manual)
• Estimated annual savings: in the range of $500,000–$1,000,000 in reduced rework (depending on defect rates and production volume)

💡 Why this couldn't be done with Copilot

This is a classic example of where Azure AI Services are the right choice:

• Real-time image classification is not something Copilot does
• The solution needs to run on the production line, not inside Microsoft 365
• It requires a custom-trained vision model specific to steel defects
• The API-based architecture integrates with existing manufacturing systems

If the exam describes a specialised AI need (vision, speech, document processing) on a production or operational system, the answer is almost always Azure AI Services — not Copilot.

Question

A company needs real-time image classification on a production line. Which Microsoft AI service should they use?

Click or press Enter to reveal answer

Answer

Azure AI Vision (part of Azure AI Services). It provides image analysis, object detection, and custom image classification via APIs. This is a specialised AI capability that Copilot and Copilot Studio don't offer — they focus on productivity and chatbots, not real-time vision processing.

Click to flip back

Knowledge Check

Elena's consulting firm has a client — a legal firm that wants to automatically extract key information (parties, dates, amounts, clauses) from thousands of contract PDFs. Which Azure AI Service should Elena recommend?

AAzure AI VisionBAzure AI SpeechCAzure AI Document IntelligenceDAzure AI Search

Check Answer

Knowledge Check

Ravi wants to build an AI assistant for TechVantage that answers employee questions using information from SharePoint, Confluence, and an internal SQL database. Which Azure AI capability enables searching across all these sources?

AAzure AI LanguageBAzure AI VisionCAzure AI Search with multi-source indexingDAzure AI Speech

Check Answer

---

Next up: Matching the Right AI Model to Your Business Need — learn how to choose between large models, small models, and open-source options.