The AI Productivity Gap: When CEO Enthusiasm Outruns Evidence

The News

A fresh debate has broken out over whether the people loudest about artificial intelligence are also the least equipped to judge it. On the latest episode of TechCrunch's Equity podcast, hosts Anthony Ha, Kirsten Korosec and Sean O'Kane picked apart a provocation from Box founder Aaron Levie: that chief executives are "uniquely prone to AI psychosis" precisely because they sit so far from what Levie called the last mile of work.

The argument is blunt. Leaders who no longer write code, answer support tickets or close the books are the ones promising the steepest productivity gains from AI. The further a boss sits from the task, the easier it becomes to believe a demo reflects daily reality.

The panel pointed to consumer pushback as evidence that not every AI feature is welcome. Privacy-focused search engine DuckDuckGo has seen installs climb 30 per cent, a jump the hosts tied partly to users fleeing AI additions they never asked for. Google, they argued, has bolted generative answers onto its core search product in ways that dent trust rather than build it.

Why It Matters

Strip away the colourful framing and this is the defining business question of 2026: is the money pouring into AI buying real output, or simply executive conviction? Capital expenditure on data centres and model licences has reached eye-watering levels, yet hard productivity numbers remain stubbornly elusive.

There is precedent. Economist Robert Solow famously quipped in the late 1980s that the computer age was visible everywhere except in the productivity statistics. The personal computer eventually delivered, but only after a decade of reorganised workflows. The risk today is that boards book the gains before the reorganisation happens, then quietly write them down.

The DuckDuckGo signal matters because it inverts the usual story. For two years the assumption was that more AI meant more engagement. A measurable migration toward a deliberately AI-light product suggests a ceiling on how much automation users will tolerate before they walk. That is a demand-side warning few investor decks have priced in.

Indian Angle

Nowhere is the gap between AI promise and proof more commercially loaded than in India's services economy. TCS, Infosys, Wipro and HCLTech have built investor narratives around AI-led productivity, pledging clients more output from smaller teams. If Levie is right that the gains are routinely overstated by those furthest from delivery, the pressure lands squarely on Indian engineering managers asked to make the spreadsheet match the pitch.

NASSCOM has repeatedly urged caution, noting that enterprise generative-AI adoption in India is still mostly at pilot stage rather than full production. Indian CIOs, who answer to tight margins and rupee-denominated budgets, tend to demand proof of return before scaling. That scepticism, often mistaken for slowness, may turn out to be a competitive advantage if the global hype cycle corrects.

There is a consumer parallel too. India's largest internet audiences are price and data sensitive, and a backlash against intrusive AI features, of the kind boosting DuckDuckGo abroad, could shape how domestic platforms roll out automation. For founders and funds alike, the lesson is the same one Indian enterprise buyers already practise: measure the last mile before you celebrate it.

FAQ

What is "AI psychosis" in this context?

It is a deliberately provocative phrase for the tendency of senior executives to overestimate AI's productivity benefits because they are removed from the actual work. The claim, made by Box founder Aaron Levie, is that distance from delivery breeds unrealistic optimism.

Why does the DuckDuckGo figure matter?

A reported 30 per cent rise in installs suggests some users are actively avoiding products laden with AI features. It is an early signal that automation has a tolerance ceiling, which complicates the assumption that adding AI always lifts engagement and revenue.

How does this affect Indian IT firms?

Indian services giants have tied growth stories to AI-driven efficiency. If real-world gains lag the marketing, delivery teams face pressure to reconcile client expectations with what the technology actually saves, making measurable proof of return more important than ever.

Where can I read the original discussion?

The debate featured on TechCrunch's Equity podcast, published on 31 May 2026, with the full argument and panel discussion available through the original article linked below.

This story was reported by TechCrunch. Read the full original coverage at TechCrunch.