AI’s Productivity Boost Still a Long‑term Dream, According to Deutsche Bank’s Jim Reid

When the buzz around generative AI reached a fever pitch, many investors started picturing a near‑instant surge in corporate efficiency – like flipping a switch and watching profits soar. Jim Reid, an analyst at Deutsche Bank, pulls the rug out from under that fantasy. In a recent note, he argues that the promised productivity windfall is still a few years away.

Reid points to the classic technology‑adoption curve: early excitement, a lag as firms wrestle with integration, data quality, and talent gaps, and only then the actual payoff. He cites a Gartner forecast that suggests a tangible impact on earnings per share won’t become evident until around 2027‑2028. In other words, we’re not looking at a 2024 miracle.

Why the delay? First, the cost of building or buying AI tools remains high. Companies need to invest not just in software licences, but in the infrastructure – cloud compute, storage, and secure pipelines – that can support large models. Then there’s the talent bottleneck: data scientists, prompt engineers, and AI‑ops specialists are scarce, and onboarding them takes time.

Second, the data problem can’t be ignored. Most enterprises still sit on messy, siloed datasets. Training a model on such a patchwork often yields unreliable outputs, meaning firms spend months fine‑tuning, testing, and re‑training before seeing any real improvement.

Finally, regulatory and ethical concerns add another layer of caution. In sectors like finance and healthcare, any AI‑driven decision must pass strict compliance checks, slowing down rollout even further.

That said, Reid isn’t dismissing AI altogether. He believes the technology will eventually become a “productivity multiplier” – but the multiplier effect will be gradual, akin to how the internet reshaped business over a decade rather than overnight.

For investors, the takeaway is simple: temper expectations, watch for genuine pilot successes, and focus on companies that are methodically building AI capability rather than those shouting the loudest.