There is no doubt that artificial intelligence is disrupting and transforming industries across every discipline. This is due to the tireless efforts of AI pioneers in developing tools that can be effectively deployed. But one thing that we tend to take for granted is the success of AI ventures.
Developing AI tools is an expensive and time-consuming process, involving extensive trial and error. In fact, the leading pioneers of AI did not have it easy; they managed to rise to their positions only after repeatedly. A successful AI pioneer learns three important lessons:
Despite the very much valid hype surrounding AI, AI projects rarely succeed. On the other hand, the successful ones reap amazing returns; an example is Amazon’s AI. Says Charles Elkan, former head of ML, “Our biggest [AI] win [at Amazon] was several billion dollars in revenue, which translates to maybe $100 million in profit because the retail business is actually a relatively low-margin business.”
The profitability of AI businesses was discussed by a recent study out of MIT’s Sloan Management Review.
“Many AI initiatives fail. Seven out of 10 companies surveyed report minimal or no impact from AI so far. Among the 90% of companies that have made some investment in AI, fewer than 2 out of 5 report business gains from AI in the past three years. This number improves to 3 out of 5 when we include companies that have made significant investments in AI. Even so, this means 40% of organizations making significant investments in AI do not report business gains from AI.”
So, what makes the AI industry tick? How do businesses profit in a domain riddled with pitfalls? We discuss below.
Although AI is seeing a lot of investment in recent years, the industry is still evolving and sees businesses failing every day. The top reasons for failure are:
Businesses often fail because they do not see beyond innovation. It is necessary to envision what problem the innovation is going to address once it is built.
Companies fail when they are incompetent in data collection, storage and accessibility. Bad and unstructured data also impedes progress.
Most companies do not have the cultural capital to understand and embrace AI into their business. Lack of data or AI literacy is a handicap to the development and success of AI capabilities.
Businesses suffer from inability to identify the right use cases for AI.They tend to either overvalue or undervalue the benefits of AI because of lack of information.
Failing is very frequent in AI. Many businesses don’t give the time and patience AI requires after one or two failures.
The MIT report mentioned above also remarks that the average number of attempts it takes to gain benefits from AI deployments is eleven!
Some businesses make it because they had a strategy and a clear vision of the goal to be achieved. Picking measurable problems helps to succeed better; companies can predict the impact of AI the better for deployment.
The first thing to consider is the problem to be solved. It is always better to pick ones that already have a business case so that funding is quicker. Setting up for success involves doing just the opposite of what was discussed above.
ROI on AI is always difficult to calculate, so enterprises need a staged approach to get a clearer picture of the risks involved in large-scale deployments. The three stages can be as follows:
Before deploying a new AI project, it is advisable to use a limited pilot to scout the area to qualify hypotheses and gather the knowledge to calculate the value of scaling up.
This checks the effectiveness of two approaches simultaneously. A/B testing can quantify the value given by AI to the business by pitting it against the status quo.
A model is always best for demonstrating the benefits of the AI for safeguarding against potential threats to employees, customers or equipment, and driving conviction in the model inside the organization.
There are four ways to calculate ROI.
BEA measures the number of sales needed to regain investment in terms of units sold. This is best for market-focused projects.
This measures the time needed to regain investment in terms of months or years, and is best for investment-heavy projects that turn profitable over long periods of time.
NPV gives the most complete view of the ROI by measuring the 10-year impact of the present project, expressed in today’s total dollar amount. This metric is in total dollars and demonstrates the total project value to the business.
IRR measures the ROI of a project over its life cycle in percentage. This is great for projects where the company reports to investors or borrows money, and is most suitable for fundraising at the center.
AI can prove beneficial beyond just goal-fulfillment, in unexpected ways. These extra benefits add to the value of a project and figure as part of the final output. Sometimes, the project may result in a completely different output altogether. Hence it is important to educate all stakeholders that there are no surefire guarantees for the achievement of targets due to the experimental nature of AI.
According to McKinsey Global Institute, AI could potentially deliver additional global economic activity of around $13 trillion by 2030, amounting to 1.2% additional GDP growth per year. If this becomes real, its impact would be as great as the biggest industrial revolutions in history, from steam engines to IT. Now is the best time for enterprises to join the AI bandwagon, provided they first figure out the technicalities of problem assessment.