Applied AI & AI in business /

What is deep learning and how can I make use of it in my business?

September 7 2020/12 min read
  • Reynaldo Boulogne
    Reynaldo Boulogne

Making sense of deep learning

Deep learning is a type of machine learning that is able to work with image, video, audio and natural language data.

Ok, so what is machine learning? In a nutshell, it’s a group of techniques that allow computers to come up with a way to solve a problem by themselves only by looking at data, rather than being pre-programmed with a set of answers as is the case in software development.

For example, in traditional software development if you wanted to create a solution that is able to detect whether there is a car present in an image or not, you would have to explain (code) how a car looks from every possible angle and in every possible color, shape, lighting conditions, background, and countless more variables. This wouldn’t be humanly possible.

Instead, with machine learning and specifically deep learning, since we’re talking about images in this example, you can show the computer thousands of car images and let it analyze them by itself to look for patterns. By repeatedly doing this, it begins to recognize the features that make up a car and develops a model of how a car looks like. Once it has built this model, next time it comes across a car picture it should be able to identify it.

Differences between Software development and Machine learning

In short:

  • Machine learning systems learn through examples, so making sure the system is shown the right data for the problem at hand is pivotal.
  • Machine learning systems don’t know or understand anything nor can they think, plan, reason, or have intuition. What they can do is automatically find and create a model for very complex patterns in data that we wouldn’t be able to do by hand.
  • Deep learning and neural networks are (for the most part) the same thing. So neural networks are also a form of machine learning and thus, all the ideas expressed in this article apply to them.

The opportunities of deep learning for businesses

So where can I use this in my work / business? As we mentioned earlier, deep learning systems can create models that a computer can understand for things that we can’t capture in code or through hand-made models.

This might seem a bit too abstract, so let's look at what this means in practice.

Sample Task 1:

Detect defective products on a production line through images.

Example of defect detection using image data

How would you solve this using...

  • Traditional software development: Manually explain (code) all the possible types, shapes, orientations, lighting conditions,  and countless more variations of the defects that might appear in the production line. This is painstaking and error prone manual work.
  • Deep learning: Show the deep learning systems images of the possible defects and let it automatically find the patterns it needs to detect the defective products in the future.

Sample Task 2:

Determine how similar the content of two documents is. 

To illustrate this, we’re going to use some made-up news headlines.

Example of semantic similarity search

How would you solve this using...

  • Traditional software development: Manually explain (i.e. model / code) all the possible way in which you can express an idea in writing. A starting point could be, for example, creating  a thesaurus to match similar words; creating lists of nouns, verbs, adjectives, etc.; encoding sentence structures, etc. As you can probably guess, this would not only be painstaking work, but also probably not give good results.
  • Deep learning: Show the deep learning system thousands of sentences from different contexts (newspapers, articles, books, social media messages, etc.) and let it automatically find the patterns it needs to build a model of how written language works and what is considered to be similar content.

Sample Task 3:

Determine the type of failure of a motor by sound only.

Example of an audio classification problem

How would you solve this using...

  • Traditional software development: Manually explain (code) all the possible ways each of the failure modes could sound like while at the same time excluding all the possible background noise that could be present during normal operation and which might interfere with the identification of the failure mode sound. This is painstaking and error-prone manual work.
  • Deep learning: Feed the deep learning systems audio clips of the different kinds of failure modes in normal operation (i.e. including normal background sounds) and let it automatically find the patterns it needs to differentiate the different failure modes in the future. To get a sense of how this works, you can try out this Audio Analysis for industrial maintenance tutorial which also uses sound data.

With these examples under our belt, we can take our understanding of deep learning a bit further:

Think of deep learning as a way to automatically make models that a computer can understand, for things that you can see, hear or use language for. And it does this only by being shown examples (i.e. data) of the task you want it to model.

This is immensely powerful! 

If you know a task that relies on a qualified expert looking at or hearing out for something to make a decision, or a task where someone needs to extract specific information from language, they can now potentially be also done by a computer.

So let’s go back to the main question: where is this relevant in my work/business? Why should I care?

The simple answer is that once a computer is able to do the above it opens up a whole new world of automation opportunities in your business.

  • Do you have a quality control process where a qualified expert needs to do a visual inspection of the product? You can automate that using deep learning.
  • Do you need a team going through dozens of documents to find answers to specific questions? You can automate that using deep learning.
  • Do you have an operator that knows what to do just by listening to how a machine sounds? You can embed that knowledge into a system using deep learning.

To get your creativity going, we'll finish off with a few more use cases.

Example cases using image data


Classify any image/video based on what is depicted in them.



Highlight different objects in an image / video.



Find images that look similar to each other. Examples:

* Image source:

Object detection

Find any kind of object / item in an image

Example cases using text data

Text classification

Sort text into different categories based on the content / meaning of the text.


Sentiment analysis

Popular subcategory of text classification, which sorts messages based on the sentiment contained in the message.

Text similarity

Surface sentences or paragraphs that have similar content / context, even when the actual wording is different.


Entity extraction / Named entity recognition

Locating and classifying named entities into predefined categories.

Text generation

Automatically create complete and coherent sentences or paragraphs about any topic.

Example cases using audio data

Sound classification

Sort audio sample based into different categories based on the sounds in the sample.


  • Audio analysis for industrial maintenance (Tutorial)
  • Build your own music critic (Tutorial)
Speech to text

Convert spoken language directly to written text

Some closing thoughts

As we mentioned before, data is the key that will make these applications possible. Chances are that you won’t have data for a task that you never thought could be automated. But rather than this being the end of your AI journey, it is the beginning. Finding that use case where automating it would result in substantial gains for your business, will be the catalyst for starting to collect the data you need to build the deep learning solution.

If you think we can help you on your journey, go to our solutions page or get in touch with our services team. Alternatively you can try out our platform to try building something yourself!

  • Reynaldo Boulogne

    Reynaldo Boulogne

    With over 15 years of experience, Reynaldo has worked within the intersection of business and technology across multiple sectors, most recently at Klarna and Spotify. He is passionate about innovation, leadership, and building things from scratch. Reynaldo is also a former Vice-chairman of the Stockholm based AI forum, Stockholm AI.

02/ More on Business & Applied AI