Gergely Szertics
Dozent
Gergely earned a Master’s degree from Corvinus University in Organization and Management. He then ...
Professional Education

AI in Business: What data source would you plug into your head?

April 23, 2018
0
9
1999
< >

Artificial Intelligence is reshaping our world. Sophia shows us already that we will have to live with robots that have citizenships in certain countries, self-driving cars are cautiously developing despite recent fatal accident and Alpha Go with Deep Mind’s engine is beating the best of humans in complicated games like Go. But it is not just general science and thrilling headlines any more. AI is here to reshape business and enable new competitive advantages.

Examples of Artificial Intelligence applications in business

Artificial Intelligence is not only used in areas where one would expect it with a body or as a Chabot. The most broadly used business application nowadays is called machine learning (ML) which is the capability of finding and understanding patterns from a huge amount of data. Here are some examples of how ML can be useful in business. It can…

  • help finding similar customers to your best customers, based on the digital footprint of both of them, thus generating new leads, like Lead crunch.
  • help transcribing all sales calls and analyse the text to give suggestions of how to learn from the best sales people, like Gong.
  • help analysing all manufacturing results and machine settings in order to improve them and to reduce waste products or picking time in warehouses, like Hitachi.
  • pick up patterns of actions from CRM and suggest the next best move in the case of a specific customer like in InsideSales.
  • experiment with posts, timings for different target audiences in different social media channels and optimise the time of sending and the content to specific target groups like Albert.
  • profile sales agents and customers and start assigning clients to agents where the probability of an effective conversation is the highest, like Afiniti.
  • suggest answers to customer service agents they should reply to a customer either on chat or email. If the algorithm is confident enough to have figured the right answer and if it is allowed to, it can even respond automatically like Digital Genius.
  • even find a pattern that UPS should extract the option of turning left from route planning and they could deliver 350 000 packages more every year.

These examples above use some kind of machine learning technique. But how can we imagine on how this works in general?

Machine learning: the pattern recognition engine

 Machine learning (ML) is such a generally used term nowadays that we tend to merge it with the notion of Artificial Intelligence (AI). Although AI as a term is more broad than ML and also contains expert systems or reasoning engines and other, it is true that ML is the most rapidly developing field of AI. Also exotic terms like “deep learning”, “neural networks”, “reinforcement learning” are related to this topic. These have all contributed to the technology that could beat humans in the GO boardgame. But how can a non-IT mortal, for example a responsible business person, relate to all this if we don’t want to learn coding, but we would still want to live with the opportunities it brings? Let’s understand machine learning through our own human experiences with an analogy.

If you look at the following picture: Can you differ the male from the female face at once?

Sure you can. Could you explain the exact parameters of your decision? I seriously doubt it. You might come up with guesses. But your brain has an area which is responsible for pattern recognition. It does not have to do reasoning but identify differences and make predictions. From a very early age we were in millions of situations where we had to make a guess about a face: boy or girl?… and we faced the consequences in the form of acknowledgement, being ridiculed or offended mates. This was important feedback and our brain picked up on these and refined our guesses. That is what we call reinforcement learning within machine learning: a lot of experience with a lot of feedback.

An analogy like that can be used as a general principle to understand how to relate to this kind of AI: you need a huge amount of examples to learn from. Thereafter you can classify things, predict trends and behaviours. Have you ever felt that there is something wrong on how a word is written because it just doesn’t look right? That is pattern recognition.  Do you have any idea about how a quarrel will end or how a colleague will react to a promotion? Do you have a relative, whose murmuring only you understand and others have a hard time decoding? That is also pattern recognition.

Even a large chunk of natural language processing (NLP) is based on this statistical approach. But here is a simple question: “How many pairs of animals did Moses take to his ship?” If you started counting the different species one could have packed on a boat you’ve been tricked by your pattern recognition engine: Moses didn’t take any animals to any ships. It was Noah. Pattern recognition neglects the little defaults that occur in a pattern and focuses on the generally conceivable idea.

Pattern recognition can have great benefits and can give us invaluable insights and intuition that we do not have to understand, also it can be misleading sometimes. But why is it important in the context of AI in business?

Data streams and (artificial?) expert intuition

 Expert intuition is based on years of working within a specific area. For example if a man is working with a CNC machine for ten years, he will have an idea on everything, that seems strange with it.  He might not be able to explain, he just knows. Or: if a sales representative has been in contact with about 1000 clients, he or she has a pretty good intuition about the 1001st during the first five minutes. A customer service agent is significantly faster in replying after having met 10.000 complaints. A sales manager can instantly see the patterns of each employee from their CRM reports. A clever broker “has a feeling” about what is going to be the next big hit, an experienced investor “has a smell” for good investments. These are all great examples of well trained neural networks in between our ears.

We love to hire experienced people for the same reason we are afraid of them: they operate with patterns. They are fast in solving things, but they can also be biased by the specific experiences that they have been through. Imagine to plug wires into people’s heads to process more data and identify more patterns. What if we could plug all the sensor data, and sounds and temperature and yields and maintenance history of 3000 CNC machines to the head of an operator? How much better would his intuition be, when a machine is going to break down? How much better could a sales manager be if he could not only go through all the reports, but also all of the CRM data? How well could a sales person suggest products to a need if she had experienced 100.000 conversations from all 100 sales reps?

This is the main idea behind using machine learning driven pattern recognition in business. We can feed in a lot of data to different engines which are able to learn specific patterns and give hints upon categories, give predictions, even act autonomously, like a trading algorithm. One way thinking about using AI in an organisation is to think about which data streams you would like to feed directly into your brain because you think there might be an important pattern in there, like habits of your customers, stock levels, costs of procurement and yields or customer service conversations. These are the areas where you should train a machine learning algorithm. This is the potential you would like to exploit with AI: what would you ask yourself if you could channel all that data to a person’s head? Some of these questions can be answered already.

If you start playing around with this thought experiment, you will start realizing your company’s enormous data capital, which is not that easy to grasp. Despite the difficulties of collecting and accessing it, data capital is going to be the key asset in an AI driven future.

Learn more about the FS programmes.

Artificial Intelligence is reshaping our world. Sophia shows us already that we will have to live with robots that have citizenships in certain countries, self-driving cars are cautiously developing despite recent fatal accident and Alpha Go with Deep Mind’s engine is beating the best of humans in complicated games like Go. But it is not just general science and thrilling headlines any more. AI is here to reshape business and enable new competitive advantages.

Examples of Artificial Intelligence applications in business

Artificial Intelligence is not only used in areas where one would expect it with a body or as a Chabot. The most broadly used business application nowadays is called machine learning (ML) which is the capability of finding and understanding patterns from a huge amount of data. Here are some examples of how ML can be useful in business. It can…

  • help finding similar customers to your best customers, based on the digital footprint of both of them, thus generating new leads, like Lead crunch.
  • help transcribing all sales calls and analyse the text to give suggestions of how to learn from the best sales people, like Gong.
  • help analysing all manufacturing results and machine settings in order to improve them and to reduce waste products or picking time in warehouses, like Hitachi.
  • pick up patterns of actions from CRM and suggest the next best move in the case of a specific customer like in InsideSales.
  • experiment with posts, timings for different target audiences in different social media channels and optimise the time of sending and the content to specific target groups like Albert.
  • profile sales agents and customers and start assigning clients to agents where the probability of an effective conversation is the highest, like Afiniti.
  • suggest answers to customer service agents they should reply to a customer either on chat or email. If the algorithm is confident enough to have figured the right answer and if it is allowed to, it can even respond automatically like Digital Genius.
  • even find a pattern that UPS should extract the option of turning left from route planning and they could deliver 350 000 packages more every year.

These examples above use some kind of machine learning technique. But how can we imagine on how this works in general?

Machine learning: the pattern recognition engine

 Machine learning (ML) is such a generally used term nowadays that we tend to merge it with the notion of Artificial Intelligence (AI). Although AI as a term is more broad than ML and also contains expert systems or reasoning engines and other, it is true that ML is the most rapidly developing field of AI. Also exotic terms like “deep learning”, “neural networks”, “reinforcement learning” are related to this topic. These have all contributed to the technology that could beat humans in the GO boardgame. But how can a non-IT mortal, for example a responsible business person, relate to all this if we don’t want to learn coding, but we would still want to live with the opportunities it brings? Let’s understand machine learning through our own human experiences with an analogy.

If you look at the following picture: Can you differ the male from the female face at once?

Sure you can. Could you explain the exact parameters of your decision? I seriously doubt it. You might come up with guesses. But your brain has an area which is responsible for pattern recognition. It does not have to do reasoning but identify differences and make predictions. From a very early age we were in millions of situations where we had to make a guess about a face: boy or girl?… and we faced the consequences in the form of acknowledgement, being ridiculed or offended mates. This was important feedback and our brain picked up on these and refined our guesses. That is what we call reinforcement learning within machine learning: a lot of experience with a lot of feedback.

An analogy like that can be used as a general principle to understand how to relate to this kind of AI: you need a huge amount of examples to learn from. Thereafter you can classify things, predict trends and behaviours. Have you ever felt that there is something wrong on how a word is written because it just doesn’t look right? That is pattern recognition.  Do you have any idea about how a quarrel will end or how a colleague will react to a promotion? Do you have a relative, whose murmuring only you understand and others have a hard time decoding? That is also pattern recognition.

Even a large chunk of natural language processing (NLP) is based on this statistical approach. But here is a simple question: “How many pairs of animals did Moses take to his ship?” If you started counting the different species one could have packed on a boat you’ve been tricked by your pattern recognition engine: Moses didn’t take any animals to any ships. It was Noah. Pattern recognition neglects the little defaults that occur in a pattern and focuses on the generally conceivable idea.

Pattern recognition can have great benefits and can give us invaluable insights and intuition that we do not have to understand, also it can be misleading sometimes. But why is it important in the context of AI in business?

Data streams and (artificial?) expert intuition

 Expert intuition is based on years of working within a specific area. For example if a man is working with a CNC machine for ten years, he will have an idea on everything, that seems strange with it.  He might not be able to explain, he just knows. Or: if a sales representative has been in contact with about 1000 clients, he or she has a pretty good intuition about the 1001st during the first five minutes. A customer service agent is significantly faster in replying after having met 10.000 complaints. A sales manager can instantly see the patterns of each employee from their CRM reports. A clever broker “has a feeling” about what is going to be the next big hit, an experienced investor “has a smell” for good investments. These are all great examples of well trained neural networks in between our ears.

We love to hire experienced people for the same reason we are afraid of them: they operate with patterns. They are fast in solving things, but they can also be biased by the specific experiences that they have been through. Imagine to plug wires into people’s heads to process more data and identify more patterns. What if we could plug all the sensor data, and sounds and temperature and yields and maintenance history of 3000 CNC machines to the head of an operator? How much better would his intuition be, when a machine is going to break down? How much better could a sales manager be if he could not only go through all the reports, but also all of the CRM data? How well could a sales person suggest products to a need if she had experienced 100.000 conversations from all 100 sales reps?

This is the main idea behind using machine learning driven pattern recognition in business. We can feed in a lot of data to different engines which are able to learn specific patterns and give hints upon categories, give predictions, even act autonomously, like a trading algorithm. One way thinking about using AI in an organisation is to think about which data streams you would like to feed directly into your brain because you think there might be an important pattern in there, like habits of your customers, stock levels, costs of procurement and yields or customer service conversations. These are the areas where you should train a machine learning algorithm. This is the potential you would like to exploit with AI: what would you ask yourself if you could channel all that data to a person’s head? Some of these questions can be answered already.

If you start playing around with this thought experiment, you will start realizing your company’s enormous data capital, which is not that easy to grasp. Despite the difficulties of collecting and accessing it, data capital is going to be the key asset in an AI driven future.

Learn more about the FS programmes.

Comments
Write comment