AI is having an impact across all sectors and companies are realising that they need software capabilities to manage AI deployments, as these are not available in every company or business function.

This is according to management consulting firm Boston Consulting Group's (BCG’s) 'Applied AI Index 2026' report, which shows that 48.5% of respondents are now delivering some or significant value through the use of AI.

The report shows that 7.5% of companies are now delivering significant value from the use of AI compared with companies that are lagging in AI deployment, up from 5% in the 2025 report.

Further, 41% of companies that are actively scaling AI use are also outperforming companies that are not adopting AI as rapidly, the report shows.

The report was based on a survey of 1 330 C-level and senior respondents and found that 7% of roles in companies were now AI-related roles. This was set to increase to 21% of roles in organisations being AI-related by 2030, said BCG MD and senior partner and report co-author Jeff Walters.

Defining how workflows would change and what AI agents must be built to handle which tasks were roles that did not yet exist in all companies. They were realising that tech-native skillsets were now necessary, as AI was able to accelerate knowledge work even if companies were not tech-natives, he said.

The need for software capabilities was highlighted by the finding that only 5% of companies had a full set of AI controls in place currently, while 42% expected to have autonomous AI agents in their enterprises by 2030, said BCG Tech and Digital Advantage global leader Vlad Lukić on September 30.

About 22% of the value companies currently derived from AI is from agentic AI systems, which were expected to account for about 40% of all AI value by 2030, he said.

Some of the companies in the index that were deriving the greatest value from AI were already deploying autonomous AI agents, which indicated the gap to what was needed by 2030, he pointed out during an online briefing to partners, academics and media.

The six controls that provided full control over AI systems included, firstly, memory and data access management, and this level of maturity was not yet in place in the majority of companies, said Lukić.

The second control lever required for autonomous AI is effectively managing and orchestrating how to keep humans in the loop as the volumes AI agents handle outpace human capabilities.

The third required AI control BCG's team recommends is testing and shut-off controls and rollback gates.

The closely related fourth control is the interface controls for handing out tasks to and engaging with AI processes.

Similarly, the fifth control lever is to establish clear ownership over AI agents. Agentic workflows are not limited to the normal company function silos and, typically, AI agents work across four to six different corporate functions.

A new set of leadership instructions must be introduced to manage AI agents, he said.

The sixth AI control lever companies required was managing security controls effectively on an ongoing basis as the operations of businesses changed, Lukić said.

“Companies are preparing to hand agents real decision-making authority and very few have the controls to do so safely. This gap is the defining challenge of the next two years. It won't be solved by regulators or technology vendors. The responsibility sits with the companies deploying these systems,” said Walters.

Further, the report shows that AI spending has grown to 3.3% of revenue from about 1.7% in late 2025.

These investments have also spread well beyond the enterprise IT function, as more than 80% of AI spending now sits outside enterprise IT. AI is no longer just a technology initiative, and has become a core business investment.

However, effective management of AI-related costs was also a best practice of the companies that were deriving the most value from AI systems, said BCG MD and senior partner and report coauthor Nicolas De Bellefonds.

Among the 7.5% of companies in the survey categorised as leading, or future-built, 70% actively managed the costs of AI systems, including by encouraging AI use as well as by capping and monitoring its use, he said.

Meanwhile, the deployment of AI was expected to impact on companies' workforces and, even with the anticipated increase in AI-related roles, companies were indicating that they could reduce their workforce by 10% to 15% by 2030 through the use of AI, said Walters.

However, the impact is expected to be uneven, with most companies that are delivering value from AI expecting that mostly mid-level and orchestration positions will become redundant.

While entry-level work can be automated using AI systems, companies are still recruiting entry-level workers, although with a focus on AI-related skills.

Additionally, while there would be a significant impact on the workforces in companies, 70% of the companies deriving value from AI were retraining and upskilling their employees, while 40% of the companies seeing value from AI were putting in place strategic workforce management processes, he said.

Upskilling and strategic workforce planning had become more important, and companies were revisiting how their AI talent was sourced. With the expected three-fold increase in the requirement for AI talent, and with the market competing for AI talent, 41% of companies were rethinking how they attracted AI talent, said Walters.

Further, another important finding in the report is that there are companies that are leading in the use of AI in all sectors.

As more tasks were automated and managed by AI agents, roles would change. It mattered more what companies were doing with AI than what sectors they operated in, and there were leading AI companies in every industry, said De Bellefonds.