The 5 AI Archetypes

How to navigate the growing AI landscape

Read time: 6 minutes

Hey there,

Chasing every new hype in AI is a time-killer.

That's why I've created a framework that maps modern AI capabilities to business problem domains, organized around the data they use. I call it the 5 AI Archetypes.

Once you understand these archetypes and what they're all about, you'll find it much easier to identify the right use cases for AI in your business.

Let’s dive in!

A brief history on AI

The term AI has been around since the mid-1950s. Since then, two main types of AI research areas have emerged: Artificial General Intelligence (AGI) and Narrow AI.

AGI tries to develop intelligent systems capable of solving any task they encounter, a concept similar to how human intelligence works. Narrow AI, on the other hand, refers to systems that have been specifically trained to perform a particular task. At the heart of Narrow AI is a technology called machine learning, which allows computers to infer patterns and rules from data without being explicitly trained on those rules.

It's important to clarify that all AI applications currently visible in the business world are instances of Narrow AI. AGI is still a significant area of research, and it's not clear when or if we'll ever get there.

Consequently, the five AI archetypes I'm about to present are all examples of Narrow AI. These are not hypothetical or research-based cases; they are ready-to-use technologies that can be applied in your business today.

This diagram illustrates the five AI archetypes and their corresponding data types:

The 5 AI Archetypes and their data types

Let's learn more about them and what each is capable of!

Archetype 1: Supervised Machine Learning

Supervised Machine Learning is a process that can automatically learn patterns from your historical data, typically in tabular form. This can then be used to build a model that can predict a numeric or categorical variable.

Originally, building these models used to be a lot of work.

Today, training these models has become actually rather easy thanks to advanced technology like Automated Machine Learning (Auto ML) which takes over a big part of the model training and deployment process. Check this Auto ML beginner’s guide if you like to learn more.

The real challenge though is collecting the right data, cleaning it, and preparing it. This is something that Auto ML cannot do for you, as the chart below shows:

Auto ML workflow

The input data type for supervised machine learning is typically tabular data, and most business use cases still rely on tabular data - that's why this archetype is so valuable. You train a model on historical data, and you can use that model to make predictions on new data points - sweet!

Popular use cases in this space include:

  • Predicting categories / classe

  • Predicting numerical data

  • Imputing missing values

  • Time-series forecast

  • Recommendations

Data input: tables

Archetype 2: Natural Language Processing (NLP)

Text is still the primary form in which human experience, knowledge, and feedback (e.g., customer reviews) is stored. NLP gives machines the ability to analyze, interpret, and generate this data.

This ability can be used to perform a wide range of tasks such as sentiment analysis, entity recognition, key phrase extraction, summarization, answer retrieval, or translation.

For example, modern NLP services are able to capture a good amount of context. For example, given the sentence "Next summer I want to visit Barcelona and Lion", a good entity extraction service would be able to recognize that "next summer" refers to a time/date dimension, and the words "Barcelona" and "Lion" both refer to locations - and not to an animal, despite the typo here (the city's name is Lyon).

NLP entity recognition despite typo

NLP technology has made huge leaps in recent years, especially with recent developments in the area of large language models such as chatGPT (see more below).

As you can imagine, this makes it a top skill to have in your belt.

Typical business use cases include:

  • Analyzing customer feedback

  • Categorizing content

  • Chat bots

Data input: text

Archetype 3: Audio and Speech

Audio and speech processing in business applications can typically be broken down into two fields: Text-to-Speech (TTS) and Speech-to-Text (STT).

Sometimes these fields are seen as subcategories of NLP. I like to list them out as a separate archetype, though, because they work with different data types.

Text-to-Speech is the task of converting text into a human-sounding audio stream (speech). This technology can be used to generate narratives, read reports, or create voice-controlled applications.

Speech-to-Text is the task of recognizing and understanding audio data (usually spoken language) and converting it to text. It can be used to transcribe audio recordings, telephone conversations, customer service audio, etc.

Business use cases include:

  • Voice interfaces

  • Voice-based assistants

  • Call center transcriptions

Data input: audio files

Archetype 4: Computer Vision (CV)

Simply put, computer vision allows machines to "see" images or documents in much the same way that humans do. This allows them to analyze, recognize, or extract meaningful information.

This can be used for a variety of tasks, such as object detection, face recognition, text extraction, or landmark recognition.

For some industries (such as automotive), computer vision is a big deal. But for many industries, computer vision is just a more advanced tool for analyzing PDF files - which is not to discount its value!

Computer Vision example: Extracting relevant information from a PDF

For example, a computer vision-based AI service would be able to process unstructured PDF documents and extract relevant information from them in structured form (e.g. tables) by "looking" at these documents and learning repeating patterns.

Popular analytics use cases include:

  • Entity detection for upstream processing

  • Image classification

  • Text extraction

  • PII removal

Data input: images / documents

Archetype 5: Generative AI (Mixed Data Types)

Generative AI refers to a set of technologies that allows users to generate digital content (image, video, or audio), typically from, but not limited to, text inputs. The most powerful models currently are LLMs - Large Language Models - which generate text from text.

ChatGPT is the most popular example, but there are many others. For example, tools like Synthesia allow you to generate natural looking video avatars from text inputs.

Modern Generative AI systems, such as GPT-4 aim to be multimodal. This means that they are trained not only on one type of data (e.g., images or text), but on a combination of different data types. This generally results in more powerful models that allow users to input not only text, but also images, for example, when interacting with the model later.

Here’s an example of how GPT-4 can explain the contents of an image:

Source: OpenAI

The capabilities of modern Generative AI services have become so good that they stretch out into other more narrow AI tasks.

For example, a capable large language model like chatGPT lets you also do very specific tasks such as sentiment analysis or entity extraction.

Be aware, though, that at their core these models still provide "just" text completion - where the suggested text completion in the case of sentiment analysis might be the sentiment label (such as positive or negative).

If your head hurts, don't worry. Just keep in mind that Generative AI models may be more powerful than you'd think - also, feel free to check out this guide to AI terminology here.

Generative AI use cases span a broad spectrum, including:

  • Text generation / completion

  • Speech synthesis

  • File compression

  • Video avatars

  • Coding

Data input: mixed

That’s it!

Use these 5 AI archetypes to never get lost in the AI jungle again. And the next time you hear about a new AI hype, think about which archetype might be involved.

As always, thanks for reading.

I hope you now have a better understanding of these archetypes and feel more confident navigating the growing AI landscape.

Want to learn more? Here are 3 ways I could help:

  1. Read my book: If you want to further improve your AI/ML skills and apply them to real-world use cases, check out my book AI-Powered Business Intelligence (O'Reilly).

  2. Book a meeting: If you want to pick my brain, book a coffee chat with me so we can discuss more details.

  3. Follow me: I'm regulary sharing free content on LinkedIn and Twitter.

AI-Powered Business Intelligence Book Cover

If you liked this content then check out my book AI-Powered Business Intelligence (O’Reilly).