Make the decision
Choose the smallest and fastest AI capability that meets the task’s quality, privacy, and reasoning requirements.
Pathway 1 · Lesson 0230 min
The AI Landscape
Choosing the right model for every task
Learning Objective
By the end of this lesson, you will be able to: Distinguish between major AI model families, match model types to appropriate tasks, and explain the speed-vs-depth trade-off in model selection.
7×
growth in demand for AI-fluent workers
Engage
"Using the wrong AI tool for a task is like using a hammer to paint a wall. The AI landscape has never been more diverse — or more confusing. How do you choose?"
Explore
An overview of the major AI model families and how they differ — a useful orientation before diving into the details.
Explain
Why Model Choice Matters
Not all AI models are equal — and using the wrong one wastes time and money while producing inferior results. The AI landscape has fragmented into specialised families, each optimised for different tasks.
The key insight: there is no single best model. There is only the best model for your specific task. A reasoning model that excels at complex logic is overkill (and slow) for writing a meeting summary. A fast general model that's great for summaries will struggle with multi-step mathematical reasoning.
The AI Model Landscape
| Model Type | Best Examples | Ideal For | Speed | Depth |
|---|---|---|---|---|
| Text Generation | GPT-4o, Gemini 1.5, Claude 3.5 | Writing, summarisation, Q&A, emails | Fast | General |
| Reasoning | o1, o3, DeepSeek-R1 | Complex logic, maths, multi-step analysis | Slower | Deep |
| Code | Copilot, Cursor, Codex | Code generation, debugging, automation | Fast | Domain-specific |
| Multimodal | GPT-4V, Gemini Pro | Image + text, document analysis, visual reasoning | Medium | Cross-format |
| Specialised | Med-PaLM, BloombergGPT | Medical, legal, financial domain expertise | Varies | Narrow but precise |
💡
The Speed vs. Depth Trade-Off
Reasoning models (like o1 or o3) 'think' before they answer — they spend more tokens on internal reasoning before producing a response. This makes them significantly more accurate on complex tasks but slower and more expensive. For a quick email draft, this is overkill. For a complex financial analysis or legal reasoning task, it's worth every second.
Explain
RAG, Fine-Tuning, and Agents — When to Use Each
Beyond choosing a model, you'll encounter three approaches to extending AI capability:
RAG (Retrieval-Augmented Generation): Give the AI access to your specific documents at runtime. Best for: answering questions about internal knowledge bases, policies, or recent data. No retraining required.
Fine-tuning: Retrain a model on your specific data to change its behaviour permanently. Best for: when you need consistent style, tone, or domain expertise baked in. Expensive and time-consuming.
Agents: Give the AI tools (web search, code execution, APIs) so it can take actions, not just generate text. Best for: multi-step tasks that require real-world interaction.
✅
Practical Rule of Thumb
Start with the fastest, cheapest model that can do the job. Only upgrade to a more powerful (slower, more expensive) model when the output quality is genuinely insufficient for your needs. Most daily tasks — emails, summaries, drafts — don't need the most powerful model available.
Elaborate · Worked Example
Worked Example: Choosing the Right Model
Context
Sarah is a strategy consultant with three tasks to complete today.
The Situation
Task 1: Draft a follow-up email to a client meeting. Task 2: Analyse a 50-page financial report and identify the top 3 risks. Task 3: Review a legal contract for unusual clauses.
The Outcome
Task 1 → Fast text generation model (GPT-4o, Gemini Flash). Speed matters, depth doesn't. Task 2 → Reasoning model (o1) with the document uploaded. Depth matters, speed is secondary. Task 3 → Specialised legal model or reasoning model with legal context. Domain accuracy is critical.
Try It · Reflect & Apply
Apply It: List 3 tasks you do regularly at work. For each one, which model type would you choose and why? Think about whether you need speed, depth, or domain expertise.
Your notes are saved in your browser for this session.
Evaluate · Knowledge Check
1 / 3You need to solve a complex, multi-step business problem that requires careful logical reasoning. Which model type is most appropriate?
Fluency Practice Studio
Model Choice Studio
Practise at work
Take three recurring tasks—an email, an analysis, and a document question. For each, write the required output, consequence of being wrong, sensitivity of the data, and the model type you would choose.
Quality check
Can you justify your choice by task consequence rather than by popularity or novelty? High-consequence work needs stronger verification, not simply a more impressive model name.
Spaced revisit
At the end of the week, review one model choice you made. Did the model save time, preserve quality, and respect the data boundary?
External video references
Keep Learning on YouTube
Open in a new tab
Ready to move on?
Mark this lesson complete and continue to "Prompt Engineering Foundations".