Delivering adaptive value with enterprise AI

The introduction of enterprise artificial intelligence

As business moves into a new decade, the 2020’s, enterprises continue to look for the leg up that will push them above the competition. For years now, artificial intelligence has been one of the key technologies that can help push business to the next level, becoming smarter, more efficient, and eventually more profitable.

But despite the growth of the term “artificial intelligence” into our modern lexicon, it’s not actually as commonly used as one would think among modern industries. A survey for the O’Reilly AI Adoption in the Enterprise report found that just under 75 percent of respondents said their business was either evaluating ‘AI’ or not yet using ‘AI’, leaving just a quarter of industries such as financial services, healthcare, telecommunications, and electronics and technology with fully-fledged and operational artificial intelligence systems.


Best use cases for AI

The first step in implementing artificial intelligence into a company’s business plan is to gain a better understanding of what AI is and the best ways it can be leveraged. AI generally refers to the equipping of machines with the ability not just to execute commands or follow a defined path from the programmer, but to absorb data in large quantities, analyze it, and then make determinations on actions based on that data. In essence, the programmer gives the computer a task and then feeds it as much data as possible in order for the computer to teach itself how to complete the task. This brand of programming is generally referred to as machine learning.

As a result of the machine learning model, typical use cases for AI revolve around deciphering patterns within a large abundance of data. Simple examples include differentiating a cat and a dog, determining the meaning of a word from its context within a sentence, and picking out words from an audio recording. Such examples are extremely rudimentary, but expand into more pragmatic uses such as more effective customer service via machine or picking out a wanted man’s face from a crowd. In fact, customer services systems are among the most common examples of entry-level artificial intelligence, thanks to their secondary nature compared to other facets of a company’s operations.


Big data: the fuel for the fire

Because the machine learning system relies so heavily on a strong and constant influx of data, ensuring you have or can create a large data pool from which to draw is crucial when starting out. You must be sure you know what kind of data you’ll be collecting and what kind of data will be most useful in optimizing your computer’s learning trail, which draws back to having a well-defined use case in mind before beginning. Relatedly, you must consider what features of your data are going to be the most useful in training your machine-learning model, and how you might best format that data so that the model may be as efficient as possible.

The extent to which artificial intelligence relies on data is such that even an enterprise with an extremely highly functioning data science team will need to take another look at how their data is structured and how their data is tagged before using it to train their AI investment. The data must be labelled, the data set must be complete, and even then, it might be such that the data isn’t sufficient for the machine to make large strides forward. The data you use to train the machine may not be able to compare to the data that the machine sees when used for real-world analysis. It is impossible to overemphasize the importance of building this perfect data set.


Build or buy?

If a company wishes to institute artificial intelligence, there are currently three major options through which to do so: building customized AI solutions and creating tailored algorithms to your enterprise, buying solutions through specialized companies, or using public solutions through cloud API’s.

Building your own machine learning system in house would be a greater initial investment but could easily save you money over time. The first necessary expenditure is a GPU that has the strength to train complicated neural networks, which are almost always necessary to compute the math behind the machine learning program. Afterwards, there are many different options when it comes to writing the algorithms, including frameworks such as TensorFlow, which allows the user to write in Python, Java, or C++, PyTorch, and Keras. Developing a custom AI solution allows you to better use your own data, as you can create processes that are tailored to the data you will use, which allows for more efficiency and, eventually, better predictive power.

Buying a Software as a Service (SaaS) package is probably the quickest and easiest way to jump into the world of artificial intelligence and see the most immediate benefits. They are prepackaged and ready to use, reducing initial investment cost and providing broad value fairly quickly across an organization. But the downsides are limits in the ways you can configure your systems and in the types of data you can input, thus resulting in long-term limits in ROI potential. As well, they are harder to integrate into the every- day business processes a company would hope to improve. Thus, it is hard for these solutions to provide a competitive advantage.

Finally, some solutions through large, public, cloud-based API’s such as those offered by Google, Microsoft, and Amazon can offer very powerful artificial intelligence capabilities. They are easily learned and can handle the scale of some of the more common use cases such as speech, language processing, and computer vision. By using the cloud, there is no installation of hardware, and no prior expertise in AI. But what separates these solutions from building personalized ones is their inability to use specific data; they can only use publically available datasets or those given to them by the provider. So while they may be perfectly equipped to handle the common use cases as mentioned above, they are not as well-enabled to help businesses w more specialized needs. Despite their limitations, cloud API’s can still be incredibly useful, and can be used in conjunction with more personalized systems to achieve the maximum program efficiency and quality while still achieving the desired, specific conclusion.



At the end of the day, despite all the technological support and data science that will go into the development of the machine learning system, the most important thing is creating a strategy that will ultimately allow you to incorporate AI into your organizational processes. Ensuring buy-in from upper management, even in the early days when the system is in development and not yet profitable, and meaningfully conveying the eventual positive growth that artificial intelligence is equipped to bring is crucial when embarking on the AI journey. Show a completely thought-out plan alongside AI processes with clear and useful objectives in order to find relatively short-term results, a map for how to build the AI platform out in the future, and support from all stakeholders within the organization.