Advanced analytics expertise in a democratised AI world

Note: This post was adapted from a version originally written for versor.com.au

It has been said that Artificial intelligence (AI) is well on the way to being democratised in the workplace.

But what does this really mean for businesses, employees and customers?

For business, democratisation of AI is reflected in the ease of access to all-singing, all dancing cloud-based analytics platforms which allow for cheaper, more robust access to powerful machine learning algorithms and common predictive models. It is seen in the increasing ease with which an average analyst can gain access to pre-built tools to semi-automatically generate fully productionised solutions. The need for low-level code has been greatly minimised, and in some cases this has led to virtual plug-n-play environments where to build a solution one has only to know what sort of model is being built, and not what drives it under the hood. A business which leverages these environments benefits by being able to be increasingly confident of its promises to the market, as well as paying a much lower cost for productivity improvements.

For the employee, AI democratisation means more reliable tools and applications right throughout the enterprise, leading to increasing levels of confidence on production lines and the ability to create products and solutions with previously unseen levels of accuracy and consistency. It means much more can be achieved in less time, and that output quality is much improved.

Finally, for the customer, democratisation of AI allows for an increase in the supply of high quality products and solutions supported by reliable processes. This is irrespective of whether we are talking about products built in the factory setting or about using models embedded in customer-facing applications. It allows customers to increasingly expect reliability and consistency as a given from all areas of the economy.

All of these things are true – to an extent. The concern is that not knowing what is under the hood, how particular models work, actually limits the effectiveness of AI solutions. If an analyst doesn’t understand how accurate a particular model being used is, or under what circumstances the model should or should not be used, or in fact if the data the model is based on can be relied upon, then all sorts of conclusions may be drawn from the output which are actually not valid. It really does take an expert to be able to design machine learning solutions, whether the components of the solution are pre-built or coded from the ground up.

Expert consultants will pride themselves on constantly learning. It is also important to be surrounded by a team who are engaged in either academic study or commercial certifications that directly relate to the work they do. Knowing when to use particular models and how to most effectively apply them is essential, as is being able to work across a number of major platforms. This allows the team to dive right in and write sophisticated code from scratch when the need arises.