How much value is AI really creating?

{"text":[[{"start":5.35,"text":"The conversation about AI’s usefulness has matured considerably over the past year. Outright denial of its capabilities has diminished as more people have seen for themselves what it can do in their line of work. The battle is now over exactly how much value it provides."}],[{"start":22.299999999999997,"text":"One particular point of tension between AI’s boosters and detractors has been the disconnect between reported increases in coders’ output and the apparent lack of a corresponding boom in product or value creation. A new paper leaves both sides able to claim vindication. "}],[{"start":38.349999999999994,"text":"The study by MIT’s Mert Demirer and co-authors tracked software developers’ work before and after they adopted AI tools. Importantly, they measured this at several different levels, from the amount of code written, to the number of discrete files edited, to the number of projects or features worked on, to actual releases of new software. "}],[{"start":58.599999999999994,"text":"They found an explosive impact at the top of this funnel — coders created or edited almost 300 per cent more files — but that boost was halved to 150 per cent by the time they got to the number of discrete pieces of work submitted for review, and that in turn shrunk fivefold to a roughly 30 per cent uplift in the number of full software releases."}],[{"start":null,"text":""}],[{"start":80.05,"text":"A 30 per cent uplift in producing a company’s core product is significant, but the findings nonetheless demonstrate how perceptions and even some direct measures of AI’s impact on productivity can be far out of step with the value it ultimately adds. What feels like — and indeed measurably is — an explosive boost for a particular task often translates into a much more modest gain once that work has passed through all the human bottlenecks associated with reviewing and releasing production-grade work."}],[{"start":110.65,"text":"Moreover, when the researchers looked at whether AI-assisted increases in software production have led to increased consumption, they found little evidence. The marked increase in mobile app releases over the past year has not been accompanied by any increase in downloads — most of the new apps fail to capture even a modest audience."}],[{"start":null,"text":""}],[{"start":129.75,"text":"Notably, the finding that productivity and value creation have been much weaker than some assumed landed at a time when Uber CEO Dara Khosrowshahi revealed the company had blown through its entire AI budget for 2026 in one quarter, and was planning to switch much of its AI use to lower-cost models, reserving frontier tools for special cases. Then came new research on AI use for legal work, which found that pairing cheap open-source AI agents with top-end models acting as sporadic “advisers” delivered better results at much lower cost."}],[{"start":165.8,"text":"It would not be unreasonable to see all of this as evidence that AI’s capacity to deliver genuine value has been vastly exaggerated, or at least that splurging on the latest models is often unnecessary. But Demirer and his co-authors feel the more likely explanation is that current organisational structures and marketplaces are not set up to take advantage of real underlying gains. That view is supported by the evidence from past technological revolutions, where the real jumps in productivity and job displacement came from new companies and processes rather than incumbents grafting new technology on to existing workflows."}],[{"start":203.4,"text":"In the case of electricity in the late 19th and early 20th century, productivity gains were modest where factories simply replaced giant steam engines with giant electric motors but left the rest of the machinery and layout unchanged. The boom arrived decades later when engineers fitted individual workstations with their own small motors."}],[{"start":null,"text":""}],[{"start":224.35,"text":"The fact that incumbent software and knowledge work companies are finding only modest productivity gains by incorporating AI into existing workflows and organisational structures, while usage, revenue and productivity explode at Anthropic and OpenAI — companies built around AI, with products written and reviewed by it — is perhaps early evidence of the same dynamic playing out here, only much faster."}],[{"start":null,"text":""}],[{"start":250.25,"text":"I suspect both camps are correct. A lot of corporate AI use and spending today is inefficient. But realised productivity gains are capturing the interaction of powerful new tools with poorly suited structures and processes. Those frictions and bottlenecks will only ease over time."}],[{"start":267.65,"text":"john.burn-murdoch@ft.com, @jburnmurdoch"}],[{"start":null,"text":""}],[{"start":279.3,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1780706962_1173.mp3"}