On the question of utility, AI seems to be on shaky ground

1 hour ago 3

ARTICLE AD BOX

The biggest AI developments, decoded. 27 May 2026.

Cognitive warmup. Considering what AI companies tell us, chatbots are a mix of precise elders’ wisdom and an astrologer that never goes wrong. Now I’ll tell you a tale of an AI executive whose chatbot couldn’t deliver on either. It also becomes very clear: AI bros simply cannot read the room. Perhaps the dazzle of circular funding is blinding them to realities. When Google CEO Eric Schmidt took the stage to deliver the commencement address at the University of Arizona last week, he was repeatedly drowned out by boos when the topic of the speech veered toward AI. I wouldn’t hold it one bit against the young adults at the university, considering the absolute upheaval and anxiety AI has created in the job market over the past year.

Later, Schmidt reportedly told the graduates, “When someone offers you a seat on the rocket ship, you do not ask which seat, you just get on.” There’s tone deaf, and then there are the AI bros. I’ll say it again: learn to read the room and know when to stop talking. On a side-note, many of you may wonder why I critically analyse AI companies and their conversations. It’s simply because human well-being/safety/job security/sustenance > any tech, any day of the week.

RECKONING OF AI’s FAILURES

Seemingly, the day has come—the day of reckoning for all of AI’s promises. Turns out, Starbucks has put an immediate end to using AI for inventory management. CEO Brian Niccol once believed AI would efficiently reduce product shortages or unoptimised inventory management (humans are mostly portrayed as unoptimised, till they aren’t), but it hasn’t worked out very well. It is being suggested that reasons for axing Automated Counting AI include the artificial intelligence frequently miscounted and mislabeled items, confusing similar milk types, or missing them altogether.

This conclusion was unceremoniously reached just a few months after the journey started. In September last year, Starbucks had implemented Automated Counting across stores in North America. A store worker was supposed to scan store shelves holding milk and other ingredients using a tablet, which collectively pulled data from the camera and the LiDAR scanner to identify, label and list the inventory. Basically, what’s there, what’s running low and what more to order. But if AI cannot distinguish between normal milk and soy milk even when it’s clearly written on a carton or box pack, what hope does the technology really have for being anything greater than a web search assistant or a paid intern for writers shirking effort?

There’s more.

GAP BETWEEN FACTS AND REALITY

HCLTech, a leading global technology company, in its latest Enterprise AI Market Report titled The AI Impact Imperatives, 2026 notes that companies globally are facing an execution gap. There’s a race to scale AI, but also face mounting pressure to deliver results within increasingly compressed timeframes. “AI has moved from being a technology initiative to becoming an enterprise operating reality,” notes Vijay Guntur, CTO and Head of Ecosystems at HCLTech. But he still sees the brighter side. “What leaders are grappling with now is not whether AI can deliver value, but how organisations adapt their structures, decision rights and risk tolerance to keep pace with it”.

There’s a number to this potential scenario. The report estimates that 43% of major AI initiatives by enterprises are likely to fail. That is, as enterprise leaders increasingly find a number of challenges, including hidden constraints across applications, or operating models that were not designed for autonomous, continuously learning systems. HCLTech is very clear that that many organisations are underestimating the degree of cross-functional coordination and decision-making clarity required for these to succeed. AI pilots or programs that are put in place without this handshake between technology in place, teams and workflows, as well as business leaders, are likely to stall. Irrespective of whether you keep pumping in more and more investments.

LIMITS, IS A NEW DEAL

Google sent me an email a few days ago, noting some changes to the Google AI Pro subscription plan I am currently on (it is far too expensive, and I’ve no intention of renewing this—far too expensive). “For the Gemini app, we’re introducing compute-based usage limits that factor in the complexity of your prompt, the features that you use and the length of your chat. Your limit will refresh every five hours until you reach your weekly limit,” is how the email starts off. And then it cheerfully goes on to say, “As an AI Pro subscriber, you’ll enjoy a usage limit four times higher than non-subscribers.”

4 times of what? A bit of hopping around the help pages landed me at this link to check credit usage (bookmark it, I’d suggest). I seem to have 1000 credits at the time of writing this. The big pitch, if you read carefully, is monetary. Buy more credits once you run out. Google says users can buy more credits—2,500 credits for ₹2,450 or 5,000 credits for ₹4,900 or 20,000 credits for ₹19,500. These will remain valid for a year.

Vishal Mathur is Technology Editor for Hindustan Times. When not making sense of technology, he often searches for an elusive analog space in a digital world.