Gen AI makes no financial difference in 95% of cases

The value of the United States of Technology in the United States decreased at the end of the trading yesterday, with a 1.4 % NASDAQ NASDAQ index. Among those who lost the value of Palantir, a decrease of 9.4 % and arm preparations by 5 %. According to Financial times [paywall]Tuesday witnessed the largest decrease for one day in the market since the beginning of August.

Some merchants put the fall in a report [PDF] By Amnesty International, Nanda, who has noticed the high failure rate of many artificial intelligence projects in commercial organizations. Project Nanda grew up at the Massachusetts Institute of Technology Institute and describes itself as an organization that builds a “agent web”. The paper has been placed, since publishing, behind a wiping wall, but it is available for downloading from this site.

Research authors mention only 5 % of the Gen Ai pilots reach production and actually achieve measurable cash value, as the vast majority of projects create a little impact on profit and loss measures. Nanda’s research consists of 52 organized interviews with decision makers of institutions, researchers analyzed 300+ initiatives and public advertisements of artificial intelligence, and a survey completed by 153 company leaders. She measured the return on investment for six months after she left GEN AI projects the experimental situation.

The paper says that while many institutions spread artificial intelligence in the front office or job functions facing customers, successful projects tend to find them between the work of the back office. It is in the worldly tasks of the back office where savings are accumulated, to a large extent of the low need for third -party and BPOS agencies. The poll found that there is a little effect by artificial intelligence projects on the total internal levels of employees.

While 90 % of the employees mentioned that they had benefited personally from the use of AIS available to the public, usually in the form of large language models such as ChatGPT, these self -gains do not translate at the institution level. About 40 % of the companies surveyed are paid for a subscription in LLMS.

Many failed project owners referred to the lack of contextual awareness that the artificial intelligence models show – i.e. adaptation to conditions, change over time, and remembering previous inquiries. Nanda states that forming a partnership with an organization that can provide such a system and ensure its adaptation to the specific conditions of the institution is the decisive element of success. The paper highlights many “interviews” quotes that include between 60 % -70 % of the agreement with data, “[The AI system] He does not learn from our notes, “and” a lot of the hand context required each time. “

The most positive vertical of Gen Ai was Media & Telecom, followed by professional services, health care, medicines, consumer, retail and financial services. The paper says that the rate of launch of the energy and materials project in the Obstetical Energy Sector is currently minimal. With regard to commercial units, sales and marketing are the place where most projects are based or were in the field of financing and less popular purchases as a place where artificial intelligence projects can start.

The area in a model organization where artificial intelligence is widely spread in sales and marketing, although financing and purchases is the least popular location. The complex tasks are those that are expected to be completed by artificial intelligence; Managers will set projects such as customer management to artificial intelligence only 10 % of time, while tasks such as summarizing a report or writing an email will go to humans in 70 % of occasions.

The language of the published report and its lack of academic rigor indicates that its origin and goal is more like marketing than intellectual and technological debate. The authors of the paper urges strategic partnerships with a seller who is familiar with the chances of the success of artificial intelligence projects, a partnership that is Nanda, by chance, capable of forming half of them. There are “unprecedented opportunities for sellers who can provide artificial intelligence systems capable of learning and integrated deeply,” as stated by the paper conclusions.

Newspaper headlines from the Nanda report issued a discreet reading among decision makers assigned to the AI ​​applications, however, the basic messages of the paper due to the intentions behind their publication are weakened. Stock prices this week could have been affected by partisan surveys of authors with a clear skin in the game, but it seems likely that the Nanda Publishing is simply reflecting the fears of floors circulating about the effectiveness of the obstetric intelligence as a commercial tool.

(Photo))