TrustedTech report warns of 'AI Skills Underclass' gap

Nearly half of the U.S. workforce is being left behind on AI training, with the gap falling hardest on the employees most likely to be affected by the technology, according to new workforce research from TrustedTech, a Microsoft-focused cloud solutions and IT modernisation provider. The report found that 78% of organisational decision-makers feel confident using AI at work, compared with just 43% of the workers below them.

TrustedTech operates as both a B2B and B2C business, and the research marks the second wave of its AI workforce study, following an earlier report published in May. For July's release, the company surveyed 2,001 employees across the UK and the U.S.

Andy Nolan, Vice President of Technology at TrustedTech, oversees the company's internal systems, including its CRM, ERP, billing and subscription management platforms, as well as its development staff building tools for customers. He said the report grew out of TrustedTech's own internal experience with AI adoption, after the company noticed staff using AI tools without guidance or oversight and assumed other businesses were likely facing the same issue.

The report describes what the company calls the "AI Skills Underclass": entry and mid-level workers navigating AI adoption largely alone. Nolan said the term reflects a lack of structure in how many employees use the technology day to day.

Nolan said this does not mean roles such as data entry are without value or a path forward. He argued that even seemingly simple functions carry knowledge that is difficult to automate away entirely, and that workers in these roles can be retrained rather than made redundant.

"You could be very reductionist and say it's moving data from one place to another, but there is so much that goes into that - the tribal knowledge, the things that maybe AI could eventually be trained on. But those people still have that knowledge, and there's always going to be an opportunity to retool them elsewhere in the workplace with the appropriate skilling," added Nolan.

Beyond the divide between leaders and staff, the report found that 36% of employees are self-teaching AI skills, compared with just 23% who have received any formal training from their employer.

Nolan said he had expected this formal, employer-led training to be far less common than the report found, and was equally surprised by how much more comfortable senior decision-makers were with AI compared with the workers below them.

"I'm very surprised that [business leaders] are much more accepting of the output - I feel like business leaders are typically a little bit more pragmatic, especially with the historical hallucinations and strange things with AI, where numbers aren't quite right - it's never quite perfect," said Nolan.

The findings build on TrustedTech research published earlier in May this year, which identified senior leaders as the biggest source of "shadow AI" risk within organisations. Nolan said the same leaders who set AI policies are often the first to bypass them.

"There's a convenience factor and a competitive factor, where it's like, I can just ignore these rules and these policies and do whatever I want with my own tool," said Nolan. "I think that's indicative of this very confident and comfortable trend from leaders of pushing forwards - which comes back to this underclass of AI skills, because leaders are pushing forwards and moving around any roadblocks they themselves put up, whereas the underclass employees don't have that same luxury or confidence to do that."

Nolan said organisations should start by mapping where AI is already being used internally, since usage often varies widely across teams and tools, before deciding on policy. He said the priority should be matching training to the vulnerabilities of specific roles, rather than to seniority, and recommended that companies conduct an independent assessment rather than relying solely on an internal review.

"The biggest thing that companies can do to prepare themselves for organisational gaps in AI training this coming year is get a readiness assessment. There is a lot that is uncovered, and there is a lot that you won't know in the bottom depths of your organisation," he said. "Getting an assessment is the right way to do this - it's going to raise your entire workforce, the skills and the capabilities. Absolutely, get an assessment today, it is crucial because there are so many gaps that you don't know are there."