Leverton ’s research shows that AI is first used where it’s relatively easy to implement — to automate business processes and serve customers. But many organizations are moving toward using AI to analyze data, writes independent analyst Joe Mackendrick on ZDNet.
Leverton interviewed 100 executives to explore the motives for acquiring AI. Why do businesses buy AI systems? In which of them do they invest the most? When you come up with new concepts for leadership and board, what should you focus on?
As it turned out, when buying or creating AI systems, managers think primarily about automating business processes and customer support. A significant part of the budget is spent on this. The main areas of application were:
- automation of business processes (49%);
- customer support / chatbots (47%);
- data extraction (43%);
- contract analytics (28%);
- voice and video processing/visualization (25%).
Automation of business processes and customer support are most accessible during the initial use of AI. “The most common scenarios for the use of AI are those for which the solutions are already ripe, the problems solved are less complicated and the guidelines are written,” the authors of the study note.
AI is a technology that can provide the end-user with a new and useful experience. “By 2021, most large companies will provide customers with a certain type of AI-based service,” Overton said. According to IDC forecasts, in 2022 the costs of AI systems will more than double, reaching $ 79.2 billion, and the aggregate annual growth rate (CAGR) for the period 2018-2022 will be 38%.
However, lately, AI is starting to be used for data mining and analytics. “Such scenarios are significantly more complex, especially when working with unstructured documents,” Leverton analysts say. – These documents may include thousands of different versions of the same question with different grammar, syntax, languages , and layouts. AI needs millions of records to be effective. As a result, current solutions are at an early stage of development, and their implementation is a longer and more difficult process. ”
In these early stages, AI doesn’t make a lot of money. On average, 37% of companies plan to spend less than $ 250 thousand on AI next year. And only 15% of companies – more than $ 1 million, about 20% of these companies – on automating business processes. “Since this area directly correlates with the effectiveness of the back office and, ultimately, with the return on investment, it makes sense for organizations to allocate a larger share of the IT budget for this technology,” the study said.
Company executives were asked how the real implementation of AI relates to their expectations. They were most surprised at the level of involvement necessary for implementation, as reported by 30% of respondents. Approximately the same number indicated that the implementation of AI was more complicated than expected.
The hardest thing in terms of time (46% said it took more time than planned), complexity (42%) and involvement (38%) were to process voice, video, and images.
Data extraction came in second place. 35% said they spent more time than expected, and the same amount – that it required more involvement.
There is also a personnel problem, but its importance decreases as AI matures. According to the data acquired by 42% of the “lagging” and 70% of the “leaders.”