accessibilityalertarrow-downarrow-leftarrow-rightarrow-upchevron-downchevron-leftchevron-rightchevron-upclosedigital-transformationdiversitydownloaddrivedropboxeventsexpandfacebookguideinstagramjob-pontingslanguage-selectorlinkedinlocationmailmenuminuspencilphonephotoplayplussearchsharesoundshottransactionstwitteruploadwebinarwt-arrowyoutube
Article Article

For artificial intelligence to change your business, you need to get your house in order

AI can’t just be plugged in, you have to give it what it needs first: data & training

Artificial intelligence is already reshaping business processes across industries, some faster than others. It is a mythical term, laden with our expectations that it will boost our quality of life and solve many problems.

For business users already familiar with the practical realities of artificial intelligence, they understand that it is not an off-the-shelf solution which companies can simply plug and play.

They know that to effectively implement some of the most innovative technologies known to humankind, business need to change the way the gather, organise and store data, and more importantly these businesses need to change the way they operate internally to incorporate the technology into long term processes as the business model evolves and grows, always responding to changing market demands. Essentially, they need to adopt agile iterative ways of working to enable continuous evolution of the machine learning solution at hand.

From mythical beast to practical business tool

What the heck is artificial intelligence anyway?

AI is actually a simple concept in its most basic form (with unbelievably complex implementation and fascinating underlying technologies and models), but it is primarily built around data. It has many forms and the discussions around it are dizzying, but the usefulness for businesses is much simpler to distil.

It is a general-purpose technology and a tool, essentially comprised of pieces of computer software running in the cloud. AI must be trained to perform the function you want it to accomplish (whatever that is), and it needs data. AI lives on huge amounts of data. This is what it learns from as it is programmed to do what you want it to perform.

AI is driven by a fundamental paradigm shift. Whenever you have a vast number of possibilities which are too large for humans to fully grasp (such as chess or the game GO), computers can implement techniques that go through this vast amount of data efficiently and can achieve better outcomes than humans.

This is because the AI has learned all past outcomes (games in this example) and then honed its craft by playing against itself. It is similar for how it would learn to help design a new airplane wing or develop a drug compound.

“AI is not about replacing people, it is about enhancing their ability to perform their jobs or freeing them up to do essentially human tasks, such as building stronger relationships with clients.”

AI is a (very clever) puppy that needs trained. You can use it to recognise patterns in CT scans to locate brain tumours in patients, but it needs to see a whole lot of brain scans first (like millions) to learn exactly what you want it to see. Once it understands what you are looking for, it can perform remarkably complex tasks very well and in some cases
better than humans.

This driving paradigm to train AI with data is integral to the business directives of companies on the forefront of the AI revolution, like Google and Facebook, who possess a) huge volumes of well-structured data and b) the technical muscle to implement artificial intelligence technologies on a high level to exploit that data to add value.

AI includes derivative technologies like machine learning, neural networks, and deep learning, but while these distinctions are important, benefitting from them as a business is a simpler task than understanding their technical nuances.

Puppies can be trained to do lots of tricks

The business applications are virtually endless, but you have to invest on the front end

AI can be trained to do remarkable things, across industries and adapted to your specific model, but you have to work for it. AI is not just about digitising processes, it is about adding value.

You can always identify businesses process that can be vastly improved by AI by not only improving efficiencies but by finding new value streams.

“AI is not just about automation. If you are only automating, then you are commoditising, and giving up on innovation. It is not a cost savings exercise. In its essence, AI should be about finding new value streams.”

It is always those who understand that way to capitalise on new technology is not to commoditise but to improve value added – these are the ones who win.

Business users often think they buy solutions off the shelf and it will solve their problems, but you need to change your working processes, by gathering data continuously and improving your work processes – that is what it means to be agile after all.

Did we mention data? Data, data, data…

Businesses need to establish a data pipeline and bring in data from business operations effectively. That data then needs to be organised and housed in an effective way so that AI technologies can learn from it. Think: “well-structured data in a good form”.

Then companies need to tune their models on a regular basis to leverage these technologies. That is what Google and Facebook are doing every day. It needs to be monitored, and it has to be adjusted according to the changing business needs.

3 APPLICATIONS THAT ARE ALREADY AT YOUR DOORSTEP

#1 Natural language processing

The ability to process and analyse human language is a game changer

Companies have large amounts of unstructured data in the form of things like Word documents. AI can make them searchable, structure them, have a conversation around them, find forgotten data, and much more.

Although old in its origins, the deployment of natural language processing is a particularly exciting and vastly relatable application of AI (it is about something human after all). Imagine an application that could listen to your meetings, take notes, capture action points, set goals, and capture this in a database where it is searchable.

Natural language processing will likely prove to be a significant business disruptor. Or what about AI that can recommend minimum sentencing guidelines for a court case and give explanations as to why, awaiting approval/input from the presiding judge to confirm? The applications for interacting on such a level with something as human as
language are extraordinary.

A technology called Generative Pre-trained Transformer 3 (GPT-3) from OpenAI has been cited as actually creating text that is difficult to distinguish from content written by humans.

#2 Generative design

AI and creativity are friends

Generative design is a good example to demonstrate that it is not about replacing your design department, but about improving quality about equipping them with tools to make new things happen.

Generative design is very useful in industrial applications. Wings of planes have been designed by these statistical designs for a long time. A computer runs a simulation, greatly simplifying the design process.

An AI application can be given guidelines for a new airplane wing design and can then generate and design dozens of options.

“The actual designer can choose some of those options, and then the computer will design more based on that feedback, which entails a creative process between the designer and the computer. The designer can choose a last few and then do the final touches.”

#3 Predictive technologies

Predictive technologies can address critical business and real-world needs

Who doesn’t want a predictive model that can tell you the likelihood of something happening and when? The possibilities are endless, such as predicting failure in a manufacturing product, predicting a political or environmental event, and so forth.

AI technologies can extrapolate based on previous data to predict how and when events may occur.

In 5 years in R&D, these applications will have further interesting possibilities, such as in molecular research, where models can speed up searches rapidly.