Many expectations have been loaded onto AI-based technologies over the years, from questioning the dreams of androids to algorithms that seem to know our shopping desires. But like all technology trends, there are misunderstandings, incorrect attributions and simplifications associated with the field of artificial intelligence.
Artificial intelligence, AI, is a tool to help process large amounts of data and find rules, changes or patterns that are difficult or time-consuming for humans to detect. AI is an integral part of the toolbox that makes up the approach we know as advanced analytics. Advanced analytics enables companies to make informed decisions throughout their business. However, there is also much more to this toolbox.
The task of all advanced analytics solutions should be to support business and knowledge management. Which tools are used in each solution should be decided on a case-by-case basis.
What is going on with advanced analytics, and what does artificial intelligence mean in the context of knowledge management?
Advanced analytics and artificial intelligence are both umbrella terms. A data scientists' toolbox goes far beyond the technologies that can fit underneath artificial intelligence, so it's better to use a broader term such as advanced analytics.
Machine Learning, or ML, is one aspect of artificial intelligence. When a company says that it uses artificial intelligence in its business, it almost always means machine learning of some kind. In its simplest form, machine learning uses a large amount of data to define several rules or regularities that can be used to create models to support decision-making or streamline the production process. This technology is primarily based on statistical methods and is commonly supervised – you determine the rules - or unsupervised – the algorithm finds its way.
Machine learning in isolation is rarely the best option for our customers. One of the biggest challenges in machine learning is explainability. Models work particularly well in situations where incorrect predictions do not lead to significant business problems. For example, high-quality product recommendations for customers in an online store have significant added value, while the occasional incorrect or low-value recommendation rarely impacts. Knowing how these product recommendations worked is less critical than gauging which models produce more value over time. On the other hand, it is vital to understand the entire reasoning chain in a paper machine's control algorithms, and then there is much less room for error or opaque predictions.
Advanced analytics may not be a well-defined scientific term, but it describes a data scientist's toolbox of technologies and skills in practice. Advanced analytics consists of business and data analytics, data science, machine learning, artificial intelligence, and the application of related technologies to these.
In addition to the expected machine learning applications, this term includes many statistical methods and techniques to refine insights and answers from raw data.
What information can be identified or produced through advanced analytics?
Data analytics is often divided into four categories in order, from simplest to most complex. They also settle into chronological order, from depicting what has already happened to forecasts into the unknown future.
- Descriptive tells you what has happened or what is happening now. The key is to collect data as close to real-time as practical and present it in an understandable and actionable format.
- Diagnostic explains the reasons why things may have happened the way they happened. We can drill down to the root causes of events by digging up the elements relevant to the outcome.
- Predictive tells you what is likely to happen. By combining historical data from internal and external data sources with real-time or recent data, you can identify trends and better predict the future.
- Prescriptive tells you what to do to be more likely to achieve the desired result. This forecast recommends measures or strategies that improve the likelihood of achieving what is wanted.
Machine learning and artificial intelligence can be used in all of these and become more prominent contributors the more automated the analytics become. Sometimes, cognitive analytics is also added, but we find this a contributor at all levels – hand-in-hand with those other technological umbrellas.
Artificial intelligence makes a lot possible, but it's not the answer to everything
However, business-useful information can also be generated and presented without machine learning. For example, well-designed data visualization in BI dashboards or reports are relatively simple to implement, but their business benefits are enormous.
The most important question for the functioning of any advanced analytics solution is how great the business benefit it offers may be in relation to its overall investment – in terms of money, time, and business commitment.
Prescriptive analytics with machine learning can potentially provide the most significant business benefit in many situations; however, its implementation also requires the most significant investments. This investment is regardless of if you choose a bespoke development or an off-the-shelf packaged solution.
It is genuinely practical intelligence to use the best tool for the situation.