BCG says most companies see weak AI returns

That makes the issue less about access to models and more about execution. BCG argues that the companies extracting value are the ones that move fastest from learning to daily performance, rather than treating training as a box-ticking exercise tied to completion rates.

The report’s message matters because AI spending has become one of the biggest line items in corporate technology budgets, with management teams under pressure to prove that infrastructure, software and workforce investment can justify the cash burn. If those investments do not lead to measurable gains, margins can come under strain before revenue benefits arrive.

The consultancy said the main obstacle is often not technical but organizational, with workers reluctant to shift from doing the task themselves to supervising AI-assisted workflows. In one example cited by BCG, a large global tech company said engineering teams using AI inside live projects lifted productivity by about 30% and increased features shipped by 27%.

The report also warns of a temporary productivity dip when firms change how work gets done, a phase that can last weeks and trigger a short-term drop in output. BCG says companies that keep incentives and performance metrics unchanged during that transition often push teams back to the old process before the new one can gain traction.

For investors, that raises the bar for vendors and adopters alike. Microsoft, Oracle and Salesforce are among the companies most exposed to the AI build-out narrative, but the market will increasingly want proof that customers are turning pilots into measurable gains rather than simply spending more on experimentation.

BCG says the fix is to embed training in real work, focus on a few high-impact processes, use AI to personalize learning and bring middle managers into the transformation. The broader takeaway is that the next phase of the AI trade will hinge less on adoption headlines and more on whether companies can turn capability into profit.