It is projected that by 2030, companies will be investing $42 billion annually in generative artificial intelligence (genAI) projects, including chatbots, research, writing, and summarization tools. While genAI is expected to boost productivity, determining its return on investment (ROI) could be challenging.
Rita Sallam, a distinguished vice president analyst at Gartner, stated that capturing and measuring productivity improvements has been a struggle for many clients. Expressing ROI for genAI is difficult due to its indirect or non-financial impacts that create future financial outcomes.
For instance, using genAI to automate code generation can enhance a software developer’s productivity, leading to faster time to market for new features and increased customer satisfaction. However, assigning a value to these intangibles is complex.
According to Sallam, the benefits of genAI could result in cost reduction, increased work capacity, earlier revenue generation, and higher customer spend. However, the line of sight to cost reduction is not direct.
While 55% of organizations experimented with genAI in workflows last year, AI use remains limited in scope. Nevertheless, AI deployments are expected to increase rapidly this year as projects gain experience and proficiency.
Despite the potential of genAI, Gartner predicts that by 2025, 90% of enterprise deployments will slow down due to costs exceeding value, and 30% of projects will be abandoned after proof of concept. Additionally, more than 50% of enterprises that have built large language models (LLMs) from scratch will abandon their efforts by 2028.
Bret Greenstein, Data & AI leader at PriceWaterhouseCoopers (PwC), emphasized that measuring ROI is challenging. However, adapting an LLM to perform a function or process makes it easier to compare its performance against earlier processes.
Ultimately, ROI is a financial ratio that compares an investment’s gain or loss relative to its cost. Therefore, when investing in genAI, the benefits should outweigh the costs.
…
2024-03-06 09:41:02
Link from www.computerworld.com