It’s not all about making money…at least not when it comes to AI. According to a study by professional services firm Deloitte, most companies’ adoption of AI tools has no impact on their bottom line. But researchers still praise the technology.
According to Deloitte’s “State of AI in the Enterprise” report: [PDF]74% of organizations want their AI initiatives to increase revenue, but only 20% actually do so.
The consulting firm’s findings echo a recent PwC Business Leaders survey that found only 12% of CEOs saw both cost savings and revenue increases as a result of their investments in AI.
Deloitte explains that “money isn’t everything”.
”[S]Success with AI is about more than just increasing efficiency or increasing revenue. “It is important to achieve strategic differentiation and maintain a durable competitive edge in the market,” the report states.
Not that companies’ AI investments were completely wasted. Of the 3,235 business and IT leaders worldwide who participated in Deloitte’s survey, 25 percent said AI is having a transformative impact on their organizations, up from 12 percent a year ago.
When asked what benefits AI is currently delivering in practice, 66% said it is improved productivity and efficiency. How does it work when only 20% report sales growth and the answer is unanswered? We note that a study published last year by the nonprofit METR found that AI coding tools reduce developer productivity, despite expectations to the contrary.
Employee access to AI tools is growing, even without a compelling economic rationale. Today, fewer than 60 percent of employees have access to IT-approved AI tools, up from 40 percent a year ago. However, less than 60 percent of these AI-enabled workers use AI tools as part of their daily workflow.
“This suggests that while access is expanding, enterprise AI is underutilized and its productivity and innovation potential is still largely untapped,” the report posits.
That said, it seems likely that more AI pilot projects will move into production. Today, 25 percent of organizations say they have moved more than 40 percent of their AI experiments into real-world use. This number is expected to reach 54% of organizations within the next 3-6 months.
As Deloitte sees it, the adoption of AI within companies is likely to have an impact on employment.
“More than a third (36%) of companies surveyed expect at least 10% of their operations to be fully automated within a year,” the report states. “The majority of companies surveyed (82%) expect at least 10% of their operations to be fully automated in three years.”
However, this expectation has not been accompanied by major organizational changes. Approximately 84% of respondents said they are not redesigning roles based on AI capabilities.
Even people in these positions remain unconvinced by AI technology. According to the report, only 13% of non-technical workers are highly enthusiastic about and willing to take advantage of AI. 55% accept this technology, 21% want to avoid it and 4% actively distrust it.
Ultimately, companies working on AI must convince employees that the technology has benefits beyond automating jobs.
“Organizations that are successful with AI aren’t just investing in automation and algorithms; they’re also investing in people,” Jim Rowan, U.S. head of AI at Deloitte, said in a statement. “As AI continues to create new ways of working, this dual focus on improving the capabilities of both people and AI tools will enable teams to embrace reimagined business models and establish a foundation for competitive advantage.”
Another concern is “sovereign AI.” This means that companies manage AI software and data according to local laws and regulations and are not dependent on foreign vendors or infrastructure. 83% of responding companies say sovereign AI is at least moderately important to their organization, and 43% say it is very or extremely important.
When it comes to agents (AI models that are given access to tools), usage is currently modest but expected to increase in the future. Currently, 23% of businesses report using agents at least moderately, and that number is expected to reach 74% in two years.
Slow adoption may prove beneficial, as only 21 percent of companies report having a mature governance model in place for autonomous agents.
said Ali Salafi, CEO and co-founder of Kovant, an enterprise agent platform. register In an interview, he said that the problem with the way people use AI is that they see it as some kind of fancy workflow automation.
“There’s research that shows doing that doesn’t actually make you much more productive as an individual,” he says. “People start using it, but as soon as they get tired of it, they go back to their old ways.”
He says the big change for companies to start seeing bottom-line results will come from giving AI agents tasks as if they were colleagues, and allowing those agents to perform automatically.
“We are working with this large manufacturing company,” Salafi explained. “They have about 7,000 suppliers. And every time they needed to replenish something, they had to coordinate with so many suppliers. In fact, it’s the most tedious job ever for anyone. But they basically put in place this agent worker or a team of agent workers who monitor inventory levels. As soon as inventory drops below the predicted requirement level, a preliminary email is sent to the supplier saying, ‘Can you tell me if you can supply this and at what price?’
As a result, a summary report is sent to Microsoft Teams and must be reviewed and approved by the company’s planner. Salafi said if approved, the agency would send purchase orders and follow up with suppliers until the goods arrive at the warehouse.
“So all of a sudden you’re actually saving about 95% of that manual, tedious work,” he said.
Regarding Deloitte’s report, he said the consultancy’s focus on governance can be addressed in product design, or in carefully designing AI workflows.
“They value governance, but they actually need an agile model of governance,” he said. “You build governance around people, just like when you hire people. You have to work on classifying the information first. If you’re actually just opening up the whole world to agents, which is, of course, a statistical model, there can be problems. So this is more of a design issue in my mind than a need for a large-scale governance architecture.”
Sarrafi also said that Deloitte’s findings about employee hesitance toward AI can be partially attributed to unwieldy enterprise tools. “Most of the AI tools built for enterprises and the applications built for employees aren’t really on par with what they expect in terms of user experience,” he said, adding that people don’t want to switch between multiple tools.
Successful implementations tend to allow people to interact with agents through familiar tools like Microsoft Teams and Slack, he said.
“I won’t name the product, but most of the current enterprise AI tools are one to two years behind consumer AI tools in terms of user experience,” Salafi said. ®
