What is AI?

 What is AI exactly ? What role does and can it play, and how can it help? How do I ‘handle’ AI, and what can be a possible correct approach? How can I implement AI? Also, finally, how will it impact my job?


If you ask ten different experts to define AI, you’ll get ten different variations of a definition. One of the better ones – mainly because of its simplicity – is the one from Demis Hassabis, co -founder of Google DeepMind. Demis defines Artificial Intelligence as “the science of making machines smart”. These machines then augment human knowledge and capabilities. Another useful definition on the other hand – because of its completeness – is the one that Forrester upholds. Forrester defines Artificial Intelligence as:

A self-learning system that is able to interact with humans naturally, understands the environment, solves problems, and performs tasks that normally require human intelligence, qualities, and abilities without the need to code instructions and rules.


Intelligence is an umbrella concept for the algorithms, technologies and techniques that make machines smarter and give them superhuman capabilities. These superhuman capabilities will one day bring us singularity.

AI can take on many forms: conversational interfaces that use NLP and NLG (Natural Language Processing – think chatbots and voice interfaces, and Natural Language Understanding), image recognition, machine learning. In those forms, AI can help in optimizing offers, recommend the next best actions, create content , optimize performance etc. It can crunch through large volumes of data. IBM’s Watson for example can read and fully understand about half a million pages in under 15 seconds.

Very much simplified we can visualize AI and its surrounding components as follows:

AI and its surrounding component

Forrester's definition is an excellent standard to benchmark AI against, and to help distinct real AI from fake. In the last five years, the search term Artificial Intelligence has been playing catch-up with its most popular counterpart "big data" topping it amongst the most popular search terms according to Google Trends. Globally the Artificial Intelligence topic reached the maximum interest score of 100 frequently since mid-2017. In Belgium, this happened only in October 2018. When comparing it to one of the closest counterparts' search topic "big data", big data has never gone above an interest score of 15. Because of this increased attention - just like back in the days - when many software vendors claimed to be "marketing automation", nowadays, many software vendors will try to hijack the popularity of the topic of AI to attract attention. Wrongly.

Real AI for example has feedback loops built into the system

As the name suggests, a feedback loop is any process where the outputs of a system are plugged back in and used as iterative inputs. Feedback loops exist pretty much all around us. In business, for example: taking customer feedback (the output of a product or service) and using it to improve future processes is a generally used feedback loop.

Accelerated advances in AI are supporting businesses to act more on that feedback data, and we can now analyze ruthless amounts of data and allow organizations to adjust algorithms, workflows, and processes on the fly.

Previously, companies relied on manually driven methods of obtaining, uploading, and analyzing data. Over time, the general adoption of mobile apps and an increasing number of IoT devices helped them capture more data than ever. Now, we are capable of leveraging AI to capture and analyze data at a speed and scale that previously was unthinkable.


How does a feedback loop work?

  1. The system observes user actions and system events to capture data.
  2. It analyzes observed data against historical trends and data from other sources.
  3. It predicts outcomes based on observed data and associated analysis.
  4. It recommends specific actions automatically and adjusts future ones for a better outcome.

These cycles are a loop without an end, as the system continues to refine its recommendations based on the latest outputs.

Implementing feedback loops in your organization

It might be that one of your goals is to drive automation, but you can’t simply flip a switch and check back next year. Tomorrow ’s workforce will become blended between humans and AI, and the most successful companies will include human input and intervention along the way as they transition into the hybrid model.

Before you can successfully implement AI-driven feedback loops in your organization, you need to:

  1. First of all, discover the key metric or outcome you would like to improve with automation.
  2. Figure out the critical variables and elements that affect these outputs.
  3. Audit your current technology systems to see where your data currently resides and break down any information silos that might ‘clutter ’ results.
  4. Provide critical feedback on early responses, leading to smarter, more effective automation.

Once you have the right systems in place, AI can help you create feedback loops where your product or service continuously improves with use.

Want to know more about the evolution of AI? Read this article on how AI is outpacing Moore's Law.