AI Workloads at a Glance
From chatbots to crop analysis — AI workloads come in many flavours. This module maps out every type the exam tests, including the newest: agentic AI.
What are AI workloads?
An AI workload is a specific type of job AI can do.
Just like people have different skills — cooking, driving, painting — AI has different workloads. A text analysis workload reads and understands documents. A vision workload analyses images. A generative workload creates new content. Each workload uses different models and techniques.
The exam will give you a scenario and ask: “Which AI workload is this?” This module teaches you to identify all six types instantly.
The six AI workload types
| Feature | What It Does | Example Scenario |
|---|---|---|
| 🤖 Generative AI | Creates new content — text, images, code, audio | Drafting email responses, generating marketing images, writing code |
| 🎯 Agentic AI | Makes decisions and takes actions autonomously | An AI agent that books meetings, files expense reports, or triages support tickets |
| 📝 Text Analysis | Extracts meaning from text — sentiment, entities, keywords | Detecting negative reviews, extracting names from contracts, summarising articles |
| 🎙️ Speech | Converts between spoken language and text | Meeting transcription, voice assistants, real-time translation |
| 👁️ Computer Vision | Analyses and understands images and video | Quality inspection on a factory line, facial recognition, reading street signs |
| 📄 Information Extraction | Pulls structured data from unstructured sources | Reading invoice numbers from PDFs, extracting patient info from forms |
Generative AI workloads
Creates new content that didn’t exist before.
| Input | Output | Use Case |
|---|---|---|
| Text prompt | New text | Email drafts, summaries, translations |
| Text description | New image | Product mockups, marketing visuals |
| Text prompt | Code | GitHub Copilot writing functions |
| Text + data | Reports | Automated financial summaries |
Priya scenario: Priya uses a generative AI model to turn her lecture notes into study flashcards. She pastes her notes, and the model generates Q&A pairs.
Agentic AI workloads
Takes actions autonomously — not just generating content, but actually doing things.
| What Makes It “Agentic” | Example |
|---|---|
| Plans multi-step tasks | ”Book a meeting with everyone on the project team next week” |
| Uses tools to accomplish goals | Searches calendars, sends emails, creates documents |
| Makes decisions based on context | Chooses the best meeting time based on availability |
| Iterates when the first attempt fails | If a slot is declined, it suggests alternatives |
DataFlow Corp scenario: DataFlow deploys an agent that monitors their ticketing system. When a high-priority ticket arrives, the agent:
- Analyses the issue description
- Searches the knowledge base for solutions
- Drafts a response
- Escalates to a human if confidence is below a threshold
Generative AI vs Agentic AI — what's the difference?
This is an exam favourite:
| Generative AI | Agentic AI |
|---|---|
| Creates content when asked | Takes actions autonomously |
| Responds to a single prompt | Plans and executes multi-step workflows |
| Doesn’t use external tools | Uses tools (APIs, databases, calendars) |
| Output is content (text, image) | Output is actions (bookings, emails, updates) |
Key: If the AI is doing something (not just creating something), it’s agentic.
Text analysis workloads
Understands and extracts meaning from text — without creating new content.
Covered in detail in the next module. Key techniques:
- Keyword extraction — identifying important terms
- Entity detection — finding names, dates, locations, organisations
- Sentiment analysis — determining positive/negative/neutral tone
- Summarisation — condensing long text into key points
Speech workloads
Converts between spoken and written language.
Two directions:
- Speech-to-text (STT) — also called speech recognition
- Text-to-speech (TTS) — also called speech synthesis
Covered in detail in Module 8.
Computer vision workloads
Analyses and understands visual content — images and video.
Key capabilities:
- Image classification, object detection, OCR
- Face detection, image description generation
Covered in detail in Module 9.
Information extraction workloads
Pulls structured data from unstructured documents, images, audio, and video.
This is different from text analysis — extraction focuses on specific fields and structured output, not understanding meaning.
Covered in detail in Module 11.
🎬 Video walkthrough
🎬 Video coming soon
AI Workloads at a Glance — AI-901 Module 6
AI Workloads at a Glance — AI-901 Module 6
~12 minFlashcards
Knowledge Check
MediSpark wants an AI system that monitors incoming patient messages, identifies urgent cases, searches the medical database for relevant protocols, and automatically schedules priority appointments. Which workload type best describes this?
GreenLeaf uses AI to scan photos of crops and identify which plants have disease symptoms. Which AI workload type is this?
Next up: Text Analysis — keywords, entities, sentiment, and summarisation techniques.