The term ‘Artificial Intelligence’ sparks interest and brings to mind images of highly intelligent robots and autonomous vehicles. Even the numbers support the rise of AI – By 2020, 85% of customer interactions will occur without human involvement. (Gartner); By the end of 2018, ‘customer digital assistants’ will be able to identify customers by their face and voice across various platforms and partners. (Gartner). AI has many applications in areas like artificial agents, the internet of things, marketing, financial trading, and healthcare. However, can businesses truly take advantage of AI’s growing capabilities? Why is it that small and medium businesses find adopting AI to be a distant goal?
At a basic level, Artificial Intelligence can be divided into vertical and horizontal types. Vertical AI is used by services that perform specific tasks, like automating repetitive work, scheduling meetings, and managing e-commerce chatbots. Examples include Kayak, Skyscanner, and X.ai. On the other hand, horizontal AI can manage various tasks but doesn’t dive deep into any one task. Cortana, Siri, and Alexa are examples of horizontal AI. They can answer many questions but can’t finish a job entirely. For instance, if you ask Siri to plan a vacation, it can provide some destination ideas. In contrast, a chatbot like Kayak can search for and organize a trip based on your budget and dates. Currently, horizontal AIs can serve as connectors to integrate individual vertical AIs and complete larger tasks from start to finish.
But adopting these AI services can be a tedious and expensive affair for an organization especially when it is driven by big data and complex technologies. Let’s take a look at some of the factors affecting the adoption of AI –
1. Lack of AI talent
According to Charles Green, Director of Thought Leadership at Belatrix Software “It’s a huge challenge to find data scientists, people with AI experience, or people with the skills to analyze and use the data, as well as those who can create the algorithms required for machine learning”. A.T. Kearney surveyed 430 senior executives of companies possessing advanced analytics capabilities. Two-thirds of companies answered that they can’t hire enough people who can generate insights from corporate data. This represents a dearth of talent and relevant skills. AI requires people with a firm understanding of advanced programming (Python, Java, C/C++), business strategy and digital technology. The companies often observe that candidates possess right technical acumen but lack the ability to associate it with business strategy.
2. Tangible return on investments
It’s difficult for a business to measure and predict the returns on investment in Artificial Intelligence and Machine Learning technologies. AI is generally linked with the improvement in quality and efficiency of the business in the long run. For example, an e-commerce website invests heavily in AI technology to recommend better products to the users. But the results for improved customer satisfaction and revenue would not be visible immediately.
3. Affordability of organizations
Developing and maintaining an AI solution from scratch will require data engineers, a project manager and expensive infrastructure. According to O’Reilly, it was found that the average base salary of a data scientist in the United States in the year 2014 was $105,000. Including all the other expenses, benefits, bonuses the figure can rise up to $144,000. In essence, a full data science team for AI project isn’t something newer companies or start-ups can afford.
4. Problem of Big Data
The big data conundrum is laced with two problems – abundance and lack of data. Abundance of data forces an organization to rely on sophisticated data analytics solutions and get insights for decision making. Whereas small set of data is insufficient to implement AI correctly. Applications such as speech and image recognition require deep learning and are successful when the machines are fed with huge amount of data. Here, big firms such as Google and Facebook are at an advantage. Other companies which might have excellent ideas but implementing them would be a mammoth task due to lack of data. Problems are more prominent in sectors where data is tough to gather such as healthcare. Computer vision used to predict cancer cells in the body requires more information than humans to understand concepts or recognize features.
Thus a shallow but expensive talent pool, unclear return on investments and lack of data are the major challenges faced by businesses in adopting AI. Companies looking to strengthen their business with AI should start with building a unique digital brand and establish partnerships with universities to bridge the expectation gaps. AI can be made affordable by partnering with technology firms having knowledge and infrastructure required to build robust AI models. Companies facing data hurdle should develop efficient machine learning systems which are capable of working with less data. Use of advanced data modeling techniques such as Bayesian program synthesis can make recognition of concepts in images or text much more efficiently.