Module 1: Understanding AI for Marketing
Module 1 · Lesson 2
The Main Types of AI Models & Tools Marketers Need to Know
The AI market can look confusing because hundreds of products describe themselves as AI. The easier way to understand it is to group tools by the type of model or capability doing the work.
By the end of this lesson
You will understand the major categories of AI models used by marketers, what each category is best suited to, and how to choose between general-purpose AI platforms and specialist marketing tools.
The model landscape
Language → Image → Video → Audio → Predictive
01
Language Models
Work with text, files, reasoning, research, analysis and conversation.
02
Image Models
Create, edit and interpret visual content.
03
Video Models
Generate or transform moving-image content.
04
Audio Models
Generate speech, transcribe conversations and work with sound.
05
Predictive Models
Estimate outcomes using patterns in data.
1. Large language models
The general-purpose AI workbench
Large language models — usually shortened to LLMs — are the technology behind many of the AI assistants marketers now use.
Examples include platforms such as ChatGPT, Claude, Gemini and Microsoft Copilot.
Although they began primarily as text systems, modern platforms can increasingly work across multiple types of information.
Strategy
Reason through problems
Analyse options, structure plans, challenge assumptions and build decision frameworks.
Content
Create and transform text
Generate drafts, rewrite copy, summarise material and produce variations.
Analysis
Interpret information
Work with documents, research, campaign reports, spreadsheets and qualitative feedback.
LLMs are general-purpose systems. Their strength is flexibility, but they still need good context, validation and human judgement.
2. Multimodal AI
AI that can work across more than one type of information
Increasingly, the distinction between text, image and audio systems is becoming less rigid.
A multimodal model may be able to interpret:
Text
Documents and prompts
Read reports, briefs, articles, transcripts and written instructions.
Visual
Images and screenshots
Interpret creative work, charts, website layouts and product photography.
Audio
Speech and recordings
Transcribe, summarise and potentially respond to spoken information.
For marketers, multimodal capability is important because marketing rarely exists in only one format.
3. Image-generation models
From creative concepts to finished assets
Image-generation systems can create visual material from written instructions or edit existing images.
Marketers may use them for:
Concepts
Rapid creative exploration
Test multiple visual directions before investing in final production.
Assets
Campaign imagery
Create supporting visuals for social, ads, presentations and content.
Editing
Transform existing creative
Modify backgrounds, layouts, formats or visual styles where rights and brand rules allow.
Visual quality is not the only consideration. Brand consistency, factual accuracy, copyright, permissions and authenticity still matter.
4. Video-generation models
AI is expanding into moving creative
Video AI can support tasks ranging from simple editing and captioning through to synthetic scenes, avatars and generated footage.
| Use case | Potential marketing application |
|---|---|
| Text-to-video | Create visual concepts, short-form creative or illustrative scenes |
| Video editing | Resize, repurpose, remove backgrounds or generate variations |
| AI avatars | Training, explainers and controlled presentation formats |
| Captioning and translation | Repurpose video for additional channels, languages and audiences |
5. Audio and speech models
AI can listen as well as speak
Speech-to-text
Transcription
Turn interviews, meetings, calls and videos into searchable written information.
Text-to-speech
Voice generation
Create narration or spoken versions of approved content.
Analysis
Customer insight
Analyse call transcripts or spoken feedback for common themes and customer concerns.
6. Predictive machine learning
Not all important AI generates content
Predictive models use historical patterns to estimate what may happen next.
Marketers encounter this type of AI throughout advertising, ecommerce, CRM and analytics.
Propensity
Who is likely to act?
Estimate the likelihood that a user will convert, purchase, churn or respond.
Value
Which customers matter most?
Estimate customer value or prioritise high-potential prospects.
Forecasting
What might happen next?
Estimate future demand, revenue, traffic or campaign outcomes.
Predictive AI and generative AI solve different problems. Asking a chatbot to guess next quarter's sales is not the same as building a forecast from appropriate historical data.
7. Recommendation systems
AI choosing the next best piece of content or product
Recommendation systems rank or select items according to predicted relevance.
| Environment | Possible recommendation |
|---|---|
| Ecommerce | Products someone may be likely to purchase |
| Content | Articles or videos a user may engage with |
| Offers, products or content selected for a subscriber | |
| Advertising | Creative or messages predicted to perform better for a user or segment |
8. Specialist marketing AI tools
A specialist interface may sit on top of a general AI model
Many products marketed as specialist AI tools do not necessarily contain a completely unique foundation model.
Instead, they may combine existing AI models with:
Workflow
Purpose-built processes
The software structures AI around a particular marketing job.
Data
Connected information
The product may integrate CRM, SEO, advertising or analytics data.
Controls
Marketing-specific guardrails
Templates, brand rules and approval workflows can make outputs easier to operationalise.
A specialist tool is valuable when the workflow and integrations save more time than you could achieve with a general-purpose AI platform alone.
9. AI agents
From producing an answer to completing a workflow
An AI assistant usually responds to a request. An AI agent may be designed to take a goal and complete multiple steps towards it.
A marketing agent might:
Review campaign data → identify an anomaly → investigate possible causes → prepare a summary → create recommended actions → send the report for approval.
Agentic systems can be powerful because they reduce manual coordination between tasks.
They can also increase risk because the system may take multiple actions before a human reviews the result.
General-purpose vs specialist tools
Which should you use?
| General-purpose AI | Specialist AI tool | |
|---|---|---|
| Flexibility | Usually very high | Usually narrower |
| Setup | Often requires better prompts and workflow design | May provide a ready-made process |
| Integrations | Varies by platform | May connect deeply with specialist systems |
| Best for | Research, analysis, reasoning and flexible creation | Repeatable domain-specific workflows |
Interactive tool selector
Which AI category fits the job?
Quick challenge
You need to forecast next quarter's demand. Which approach is strongest?
Knowledge check
Check your understanding
1. What are large language models particularly useful for?
2. What makes a model multimodal?
3. Which AI category is most directly associated with estimating future outcomes from historical patterns?
4. Why might a specialist AI marketing tool be valuable?
Key takeaway
Choose the model category that matches the problem
AI marketing now includes language models, multimodal systems, image and video generation, audio AI, predictive models, recommendation systems, specialist tools and agents.
The strongest marketer does not ask which AI brand is best in general. They ask which capability is best suited to the task.
In Lesson 3, you will learn how AI models actually produce outputs — and why they can sound confident while still being wrong.