- Many AI models report claims that they are not as effective as they are on sale
- 95% of the companies surveyed were barely affected by LLMS
- Specialization is key to successful AI adoption
A new study from MIT's Nanda initiative claims that the majority of Genai initiatives seeking to drive rapid revenue growth are “flat.”
Of the sampled people, 95% of companies deploying generative AI stalled at “having little unmeasurable impact” on profits and losses.
This seems to be all games, as 5% of companies benefiting from generative AI are superior. These are primarily startups, age 19 or 20, earning $20 million in revenue per year.
The key to success with AI models is specialization. A successful deployment involves selecting “one pane point”, doing this well, and using tools to carefully partner with the company.
Professional vendors are successful in around 67% of the time, but internally built models are often only successful in only about a third. Highly regulated sectors like the financial industry confirm that many organizations have built their own AI systems, but the research suggests that companies tend to fail far more in doing so.
When a line manager is given the authority to promote adoption, he sees success as he can choose tools that can be adapted over time.
As most Genai budgets are dedicated to sales and marketing, allocation is also important, but the biggest ROI was seen in back office automation.
This is not the first time that research has suggested that AI models are not needed. A considerable number of companies have introduced low-level workers layoffs and brought in AI systems, but more than half of the UK companies that replaced workers with AI have regretted the decision.
Tangible benefits from these models are increasingly difficult to find, the security risks associated with the model are related to the organization, making AI models much more difficult to reach ESG goals.
