When To Use – or Not Use – AI in the Enterprise

People are waking up to the potential of AI to empower our everyday processes.

The question is, what do we need to do to make true implementation of AI a reality, what are the challenges we face, and what adjustments do we need to make?

Why AI can be a headache to adopt

Introducing a “real” intelligence into a setting where other intelligent entities, such as humans, are present, can cause lots of problems.

We are used to processing things in a specific way where we trust each other’s reasoning behind decisions, but once an intelligence is introduced that processes information in a different way than us, that fundamental trust is challenged. I doubt that any of us would pick a co-worker that wouldn’t join you for lunch, couldn’t crack a joke, and would only think, speak and eat ones and zeros. While that is a caricatured way of looking at adding an artificial intelligence into a human setting, it highlights a potential schism that AI can cause.

The “mind” of an AI is like a black box, it cannot be decoded, understood or trusted through the regular social techniques we all rely on in our everyday work. So, no matter the expected accuracy and ability of an AI, its unfamiliarity to us as humans means we feel that we have very little reason to trust one.

While it is unlikely this will change any time soon, and it may seem impossible to educate a workforce on the inner workings of an AI, there are much more accessible elements that can be easily understood. By humanizing the input and output and explaining the reasons behind each decision, algorithms become easier to understand. And by relating the actions of an AI to the actions of their human counterpart, AI can become a much less intimidating prospect for a business to adopt.

When to use AI

As with any tool in our toolbox, we should be aware that some tools are a lot more efficient for some tasks, and others could completely ruin whatever you are trying to build. The same applies to data and how we approach it. At its core, AI is “just” another approach to deciphering data. AI has its clear benefits, it can learn from past experiences, it can examine all known data points, and it can enable us to make more calculated decisions.

However, AI also suffers from a limitation, as its ability is dictated by the amount and quality of data that we feed into the training of our systems. There is a pretty good reason why most natural language processing solutions (NLP) are provided by some of the biggest data processors out there (Google, OpenAI etc.), and that’s because they need an extreme amount of text data to become efficient.

So, to answer the question of when should AI be used − it should only be used when the data suits the tool. Ultimately, if there is an insufficient amount of data or the quality is poor, it should be clear that AI is not the correct tool for the job.

Empowerment of AI

There are prime examples of how AI has empowered businesses, giving them increased potential when adopted. Recently seen in the investment and venture capital space: Revolut’s CEO Nik Storonsky has announced his intention to create a venture capital fund that will have its research and profiling handled by an AI.

It’s a perfect example of the role that AI has within modern applications, as it is more efficient than a human, but it is not being applied as a fix all that will manage large swathes of the organization. AI certainly has a place within the current mechanism, and can be inserted without creating disruption, misunderstanding and unfamiliarity. It is unknown whether Stronksky’s proposal will be a success, but what it most clearly demonstrates is that AI and business have an intertwined fate.

AI has long been touted as the technology of the future, and thanks to the standards set by Hollywood, its associated abilities are often far beyond reason and current capabilities.

In reality, AI has become a supplement to pre-existing processes, removing the tedious and inefficient tasks, and replacing them with a much more reactive alternative.