Published on the 02/07/2025 | Written by Heather Wright

While ‘silicon ceiling’ sees frontline AI adoption stagnate…

AI agents are raising red flags for employees concerned about unclear accountability should mistakes occur.

Boston Consulting Group says 35 percent of respondents in a recent survey expressed concerns about accountability of AI agents, with 46 percent concerned about the smart digital assistants making decisions without human oversight and 32 percent concerned about bias or unfair treatment being introduced through their use.

“You can’t simply roll out genAI tools and expect transformation.”

BCG surveyed more than 10,600 frontline workers, managers and leaders globally (once again, A/NZ was missed from the survey).

It found 72 percent of respondents are now ‘regular’ AI users. How much of that is specifically genAI isn’t defined, but the report does note 72 percent of respondents say their company is deploying genAI tools. But it also found that among frontline employees, adoption has stalled at 51 percent – down one percentage point on 2024’s results.

The report, now in its third year, notes despite all the hype, it is still very early days for AI agents.

Just 13 percent of respondents report that they are currently integrated into broader workflows within their organisations. But 56 percent say they are being used experimentally, in pilots, or under human supervision, with just 31 percent saying they haven’t deployed any AI agents yet.

Other reports have flagged concerns for companies around privacy, security and integration as key implementation roadblocks for agentic AI.

A Cloudera report showed data privacy (53 percent), integration with legacy systems (40 percent) and high implementation costs (39 percent) are all pain points

The BCG report, AI at Work: Momentum Builds But Gaps Remain, shows there is plenty of uncertainty on the employee side too:  While 77 percent of respondents think AI agents will be important in the next three to five years, only 33 percent say they have a proper understanding of what they actually are.

When employees are more familiar with AI agents, their enthusiasm for them increases, with 71 percent of those who understand and can explain how agents work saying they see them as a valuable tool that can support and collaborate with human workers. That’s a sentiment shared by just 25 percent of those who have heard about the technology but aren’t sure what it actually does.

Sylvain Duranton, BCG X global leader and co-author of the report, says the results highlight that companies can’t simply roll out genAI tools and expect transformation.

Instead, real returns come when businesses invest in upskilling their people, redesigning how work gets done and aligning leadership around AI strategy.

The 2025 results show confidence around the wider AI use at work is up on previous years, but still only sits at 26 percent – up 10 percentage points on 2024. Concerns are down marginally to 28 percent.

Managers are still increasing their adoption – up to 78 percent from 64 percent last year – but leaders’ use of the technology has dropped back three percentage points to 85 percent.

But even with that uptake, BCG says many companies are failing to unlock AI’s full potential – something the management consulting firm says requires going beyond mere deployment to reshape entire workflows.

The report outlines three stages of AI – deploy, reshape and invent – and says companies creating the most value with AI focus 80 percent of their investment on reshape and invent in a few core processes.

Reshape is the redesigning of end-to-end workflows and processes to reimagine functions, such as adding AI into critical processes in HR, while invent sees companies building and innovating new business models and products to drive growth, creating new revenue streams with AI-powered services for example.

“Companies actively reshaping their workflows with AI benefit in many ways that generate value for the organisation. Their employees save significantly more time than those in companies where the technology is less integrated into the workday. In addition, employees’ decision making sharpens and they work on more strategic tasks.”

But those results don’t just happen. The 50 percent of companies surveyed who are in reshape mode do a better job of tracking value created by AI and spend more time training employees, who are more likely top say their leaders support them.

There is, however, a flip side, with employees at organisations undergoing comprehensive AI-driven redesign far more concerned about job security (46 percent) than those at less-advanced companies (34 percent). And it’s not just the frontline employees, with leaders and managers more likely to worry about losing their job in the next 10 years.

“In other words, the work of allaying employee fears is ongoing,” BCG says, adding that appropriate training and upskilling can help reduce employee concerns.

It says the results show companies need to stop underestimating the importance of training and commit appropriate levels of investment, time and leadership support.

Tracking the value generated with AI through improvements in productivity, quality and employee satisfaction is also crucial, along with investing in employees with upskilling and reskilling to reshape workflows.

And, if all else fails, when it comes to AI agents, you may want to check out our earlier story on why AI agents aren’t the answer to every problem – and you actually may not need them right now.