There are many reasons why companies are struggling to use artificial intelligence, Deloitte research says

Many companies are struggling to move their artificial intelligence (Gen AI) operations from early stages to production, according to a report from consulting firm Deloitte.

“70% of respondents said that their organization has moved 30% or less of their Generative AI testing into production,” according to lead author Jim Rowan in the company’s last minute panel. The State of Generative AI in the Enterprise’ series of articles. .

Also: Companies are doubling their efforts to deploy Gen AI, Bloomberg research says

Lack of progress in productivity is different from the complexity of work around technology. “Two out of three companies surveyed said they are increasing their investment in Generative AI because they have seen strong early benefits,” Rowan and colleagues said.

The challenge of moving Gen AI projects from the presentation of ideas into production is what Rowan and team call “efforts at scale”.

The survey, conducted between May and June, received responses from 2,770 directors and C-suite level respondents across six companies in 14 countries. The survey also included interview responses from 25 interviewees, are C-suite leaders in AI and data science leaders in large companies.

The research shows “many reasons” companies are struggling to measure Gen AI. The group is, generally speaking, “learning from experience that large-scale deployment of Generative AI can be challenging and multifaceted,” the report says.

Why companies are struggling with Gen AI became clear when Rowan and team asked survey respondents to rate where they believed their organizations were “most ready”. Less than half of the respondents felt that their company was sufficiently prepared for the most important capabilities.

On average, 45% of respondents said they were well prepared for “technical infrastructure,” and 41% said they thought the organization was well prepared for “data management.”

The areas that have not been well prepared, these responses show, are “strategies”, and 37% think that their companies are very prepared, followed by “risks and governance” and “talent”, where only about one way in five ways respondents indicate readiness in each area.

Also: One-third of all Gen AI jobs will be outsourced, Gartner says

Some of the qualitative comments from the interviewees revealed more details about where the lack of support was. For example, a former vice president of data and intelligence for a media company told Rowan and the team that “the biggest challenge” for the company “is really the amount of data we have access to and the lack of management skills. appropriate data”.

The CEO continued: “There is no common data catalog. There is no metadata that catalogs and labels the data across the company. We can only go as fast as we can name the data.”

Rowan and colleagues suggested in the report that the quality of data hinders many companies: “Data-related issues caused 55% of the organizations we surveyed to avoid some Generative AI use cases.”

The research highlighted governance issues including both AI risks and governance risks. On the one hand, the company is struggling with “new risks emerging specifically for new tools and capabilities” that are unlike the risks from any previous technology. The risks include the weaknesses of current Gen AI, such as “model confusion, visibility, privacy concerns, reliability and protection of new combat surfaces”.

Also: 5 ways CIOs can manage business demand for generative AI

Uncertainty about the new law is also causing companies to pause and think, Rowan and the team said in the report: “Teams are very uncertain about the regulatory framework that may exist in the future ( depending on the countries in which they operate).”

In response to both concerns, companies are adopting different strategies, Rowan and colleagues found. These features include: “disable access to Generative AI tools specifically for users”; “apply guidelines to prevent users from entering administrative data into public LLMs”; and “build private cloud fences and security measures to prevent data leakage into the cloud.”

The lack of focus on Gen AI projects contrasts with other recent studies that have shown a strong focus on deploying emerging technologies. For example, a recent Bloomberg Intelligence report on AI found that the rate of companies deploying artificial intelligence programs from “copilot” doubled between December of last year and July 2024, hitting 66% of domestic All services are accountable.

However, a Deloitte study may help explain why Gartner’s recent report on Gen AI and the company predicted that one-third of Gen AI projects will be abandoned before they leave the certification stage. and production.

Even as US CIOs are “working” to deploy Gen AI, and increasingly “exploring” copilot technology and the like, Deloitte’s study shows that they are running into many obstacles to doing it.