The Business of Data Analytics and Artificial Intelligence- Part I

From a pure technological viewpoint – artificial intelligence looks to be the future. The future of operations, decision making, intelligence and with that the future of software.  

For simplification in these articles, Ai would encompass machine learning and deep learning and their application in problem solving but data analytics and statistical computing would be employed differently. I will also not be discussing the general or narrow view of AI and algorithms but the business costs of developing a viable Ai platform for your business or setting up an Ai company altogether.  

"This series is about the costs associated with setting up an Ai infrastructure, and not the various complex statistical models."

Ai is not Traditional Software 

Application of Ai and data analytics has shown phenomenal capacity to solve a range of difficult problems and has helped us build scale and efficiency in three primary business domains – financial markets, institutional sales and Out-of-home advertising – each of them we would discuss in further details. The insights are result of years of work and research that Hashbrown Systems has put in with our clients.   

The first fundamental difference is that artificial intelligence is as much about working with code as much as working with data. That simply doubles your development cost. The rules for Ai are being written as we speak but the cost of development of an AI solution viz-a-viz a traditional software is remarkably higher.  

Also due to substantial cloud infrastructure and highly skilled manpower required, the gross margins for an AI company are lower. The transformative effect of Ai is not in question, but the associated costs in setting up the infrastructure ensures that it is not everyone’s cup of tea.  

Ai is not traditional service 

Application of Ai involves combining elements of software with data services. So, a particular Ai app might look and feel like a traditional software with user interfaces, data management and application programming interfaces (APIs), in the core lies trained data models residing on some cloud.  

These models perform functions, run algorithms, interpret images, do natural language processing and other complex tasks that require statistical computing. The beauty of software is that it can be developed once and sold multiple times and you can scale it super-linearly but Ai due to its inherent requirement of significant data-centric operation entails costs that goes beyond usual support functions and calls for specialized manpower.  

The linear and not super linear nature of Artificial Intelligence makes it not a great fit for every business case. Noticeably. the success of Ai lies in the long tail (something we will discuss subsequently), so an effective cost basis is vital for an Ai to succeed for your business. 

In the next episode we will take a closer look into cloud infrastructure and specifically the substantial and hidden cost of it