Overview of AI Tools and Platforms

The AI Ecosystem

The number of AI tools has exploded since 2022. Navigating this landscape is itself a skill. This lesson maps out the key categories of tools and the leading options in each, so you can make informed choices when building AI automation systems.

---

Category 1: Foundation Model APIs

These are the underlying LLMs you call to process language. You send a prompt, they send back a response.

OpenAI

The most widely used provider. Models include:

• GPT-4o: Fast, multimodal (text and images), excellent general capability
• GPT-4o mini: Cheaper, faster, suitable for high-volume tasks
• o3 and o1: Reasoning-focused models for complex, multi-step problems

Anthropic (Claude)

Known for safety focus and long-context capabilities. Models include:

• Claude Opus: Highest capability, best for complex reasoning
• Claude Sonnet: Balanced performance and cost
• Claude Haiku: Fastest and cheapest, for high-volume simple tasks

Google (Gemini)

• Gemini Ultra: Google's most capable model
• Gemini Pro: General purpose, available through Google AI Studio and Vertex AI
• Strong integration with Google Workspace

Open Source Models

Run on your own infrastructure, no API costs, full data privacy:

• Meta Llama 3: Strong open-source model
• Mistral: French company, excellent models for European data compliance
• Gemma: Google's open-source offering

When to use open source: When data privacy is critical (healthcare, legal), when you need to fine-tune on proprietary data, or when API costs are prohibitive at scale.

---

Category 2: Orchestration Frameworks

Frameworks for building multi-step AI workflows, including RAG (Retrieval-Augmented Generation) systems and agents.

LangChain

The most popular framework. Provides:

• Chains: Sequential AI operations
• Agents: AI systems that can use tools and make decisions
• Vector store integrations: Connect to databases of embedded documents
• Memory: Maintain conversation history

LlamaIndex

Focused on data ingestion and retrieval. Excellent for building systems where an AI needs to reason over large document collections.

Direct API calls

For simple use cases, calling the API directly (without a framework) is often the clearest and most maintainable approach. Only add orchestration when you genuinely need the extra capabilities.

---

Category 3: No-Code / Low-Code Workflow Automation

These tools let you build AI-powered workflows with visual, drag-and-drop interfaces. Useful for automating workflows without heavy coding.

n8n

Open-source, self-hostable workflow automation. Connects hundreds of services. Has native AI nodes for calling LLMs, processing documents, and building chatbots.

Zapier

The most widely used integration platform. "If this happens in App A, do that in App B." Now includes AI capabilities for text processing.

Make (formerly Integromat)

Similar to Zapier but with more complex logic capabilities. Popular in European markets.

When to use no-code tools

• Rapid prototyping and proof-of-concept
• Business users building their own automations
• Workflows connecting many SaaS (Software as a Service) tools

When not to use them: Complex logic, high-volume processing, custom AI behaviour, or when you need full control over the system.

---

Category 4: Vector Databases

Essential for Retrieval-Augmented Generation (RAG) -- the technique of giving LLMs access to a searchable knowledge base.

How RAG works:

1. Take your documents (PDFs, articles, knowledge base)
2. Convert them to vector embeddings (numerical representations of meaning)
3. Store in a vector database
4. When a user asks a question, find the most relevant documents
5. Include those documents in the LLM prompt as context

Popular vector databases:

• Pinecone: Fully managed, easy to start with
• Weaviate: Open source, strong hybrid search
• Chroma: Lightweight, great for development
• pgvector: PostgreSQL extension -- if you already use PostgreSQL, this is the simplest option

---

Category 5: Deployment Platforms

Where your AI systems run in production.

---

Choosing the Right Stack for Your Project

With so many options, how do you choose? A simple decision framework:

Budget small + fast iteration needed: Use OpenAI API + direct calls + no-code for integrations

High data volume + cost sensitive: Evaluate open-source models (Llama, Mistral) on your own infrastructure

Knowledge base search: Add RAG with Chroma (development) or Pinecone (production)

Complex multi-step workflows: Add LangChain when direct API calls become hard to manage

Non-technical team needs automation: Use n8n or Zapier for the integration layer

---

Keeping Up With a Fast-Moving Field

The AI tools landscape changes rapidly. New models release monthly. New frameworks emerge constantly. New capabilities become available that were impossible six months earlier.

How to stay current:

• Follow the official blogs of Anthropic, OpenAI, and Google DeepMind
• Read newsletters like "The Rundown AI" and "TLDR AI"
• Build projects that use the tools -- hands-on experience is the best filter
• Join communities (Discord servers, Twitter/X, LinkedIn) where practitioners share real-world usage

The most important skill in this field is knowing how to quickly evaluate new tools against your specific use case. Not every new model or framework needs to be adopted -- but you need to know enough to make that judgement.