Quick answers
What are the main types of AI?
Six categories cover most of what business leaders meet in 2026: large language models (text), image/audio/video generation, computer vision, predictive models (classical machine learning), reinforcement learning, and agents. Neural networks and deep learning aren't separate types - they're the underlying architecture for most of the categories above.
What's the difference between LLMs and predictive AI?
LLMs (like ChatGPT and Claude) generate text by predicting the next chunk of language - they're flexible but can confidently produce things that aren't true. Predictive models forecast or classify from structured data (sales numbers, transactions, customer attributes) - they're narrower but more measurable, cheaper to run, and easier to explain to a regulator. For numerical forecasting, a classical predictive model is usually a better fit than an LLM.
What is an AI agent?
An AI agent is a system that doesn't just produce output - it takes actions. It uses tools, browses the web, runs code, fills in forms, and executes multi-step tasks. Agent modes are now built into Claude, ChatGPT, and Gemini directly. They're powerful but still unreliable for high-stakes work, so the right pilot is something where the cost of getting it wrong is low and supervision is acceptable.
Are neural networks a type of AI?
No - they're the architecture that most modern AI is built on, not a separate category. LLMs are neural networks. Image generators are neural networks. Modern computer vision is neural networks. 'Deep learning' just means a neural network with many layers. When someone says a product 'uses neural networks,' that tells you about the underlying engineering, not what the system actually does.
Why does the AI category matter for business decisions?
Because different categories fail in different ways. A predictive model might give a wrong forecast - that's a calibration problem you can measure and manage. A generative model might confidently produce something untrue - that's a verification problem, which is a different beast entirely. The questions you ask, how you score the project, and the controls you put in place all depend on which type of system you're dealing with.
Most leaders I talk to want to make sensible decisions about AI without becoming an engineer first. They've used ChatGPT, sat through a vendor demo, and heard "agentic" four times this week without being sure what it means.
That's reasonable. ChatGPT is the most accessible AI tool that's ever existed - but it's one window onto a much wider room. If you're deciding where AI fits in your organisation, you need to see the rest of it.
You don't need to become technical. You need to recognise what's in front of you - the category tells you what kind of risk and opportunity comes with it.
Here's a working map of the AI categories worth knowing in 2026.
The map, at a glance
Six categories. Put the thing in front of you into one of them and you can ask sharper questions.
Generate and work with text. ChatGPT, Claude, Copilot, Gemini.
Create new visual and audio content. Midjourney, DALL-E, Sora, ElevenLabs.
Recognise what's in images and video. Quality inspection, OCR, security.
Forecast and classify from data. Demand planning, churn, fraud detection.
Learn by trial and error. Robotics, dynamic pricing, model training.
Take multi-step actions, not just outputs. Computer-use agents, agentic workflows.
01Large Language Models
The chatbots - ChatGPT, Claude, Gemini, Copilot, and open-weight models like Llama, Mistral, DeepSeek and Qwen. Built on the transformer architecture, what they fundamentally do is predict the next chunk of text, very well. You meet them in drafting, summarising, extracting structure from messy documents, brainstorming, and answering questions over your own files - already in someone's workflow whether IT knows or not. Reasoning models like OpenAI's o1 sit here too, trained to spend more compute on harder questions.
02Image, audio and video generation
DALL-E, Midjourney, Flux, Sora, ElevenLabs, Suno, Runway. Most modern tools are built on diffusion models now (the older GAN generation still gets mentioned, but diffusion took over for the high-quality work). Video has moved fast - what was clearly synthetic a year ago now passes a casual glance. You meet these in marketing imagery, product mockups, design exploration, training materials, and increasingly voice and video.
Honest note: copyright and authenticity questions are still live and vary by jurisdiction. Treat the output as a draft to review, not a finished asset to publish.
03Computer vision
Recognising what's in an image or video: quality inspection on a line, document parsing, OCR, security, medical imaging. This is the unhyped success story - working reliably in narrow tasks for over a decade. In manufacturing, pharma, or anything involving physical inspection or document handling, it's often where AI pays back fastest, because the use cases are well-defined and the failure modes well-understood.
04Predictive models (classical ML)
The unglamorous workhorses: demand forecasting, churn prediction, fraud detection, recommendations, credit scoring. The maths underneath - regressions, decision trees, gradient-boosted models - is decades old and well understood. If you're mid-sized or larger, these have run in your business for years (often now re-badged as "AI"). For structured numerical data, a classical model is often a better fit than an LLM: cheaper to run and easier to explain to a regulator. The hard part is rarely picking the model - it's orchestrating it into your operation.
05Reinforcement learning
Systems that learn by trial and error, optimising toward a goal through repeated feedback. Famous for AlphaGo; practical in dynamic pricing, robotic control and route optimisation. It's also key to how the latest reasoning models are trained. You're less likely to meet it directly as a leader, but it's worth recognising - it's a learning approach, not a use case: the same approach can train a robot to walk and a chatbot to be more helpful.
06Agents
The fastest-moving category of the past year. Agents don't just produce output - they take actions: use tools, browse the web, run code, fill in forms, execute multi-step tasks. A year ago this was experimental; now agent modes are built into Claude, ChatGPT and Gemini directly, and "agentic workflows" are showing up in customer service, software development, research and operations. This is where most enterprise AI investment is heading.
Honest note: agents are still unreliable for high-stakes work. Treat one like a capable junior who needs supervision and clear constraints - the right pilot is something where the cost of getting it wrong is low and the cost of supervision is acceptable.
07A note on neural networks
You'll see "neural networks" everywhere - they're not a separate category alongside the six above, they're the architecture under most of them. LLMs are neural networks; image generators are neural networks; modern computer vision is neural networks. "Deep learning" just means a neural network with many layers - a description of structure, not a different kind of system.
08What this means for your decisions
Why does the category matter to a leader who won't write the code? Because different categories fail in completely different ways - and the controls you need depend on which you're dealing with.
Might give a wrong forecast - a calibration problem you can usually measure and manage over time.
Might confidently produce something untrue - a verification problem, a different beast entirely.
You don't need to be an ML engineer - you need a working mental map. If you can place the thing in front of you into one of these categories, you can ask sharper questions, set better expectations, and decide whether to pilot it, scale it, or walk away. (We score projects with the RATES framework in workshops.)
A working map beats a technical degree.
Go deeper on opportunity, risk and the adoption roadmap in the free AI Toolkit - three short whitepapers, no opt-in.
Book a Focus Call