When AI Budgets Turn Into Waste: Why Companies Are Pulling the Plug

It’s a story that’s been replaying across boardrooms for the last few months: senior leaders slap a big, shiny AI budget on the table, hoping a wave of clever bots will catapult productivity. In reality, the majority of that money ends up feeding a machine‑driven hamster wheel that spits out endless drafts of half‑baked, error‑riddled documents.

Take Meta, for instance. Reuters reported that the social‑media giant is quietly scaling back its grand‑scale plan to replace human staff with AI agents. The same sentiment is echoing at Microsoft, where an internal memo leaked to the Guardian revealed executive‑vice‑president Ryan Roslansky venting about “low‑quality work slop” that has flooded his LinkedIn feed. "Someone sent me a document that was clearly AI‑generated," Roslansky wrote. "I read it, ran it through my own AI summarizer, and got nothing new. It’s a doom loop."

That doom loop isn’t confined to tech giants. Across industries, employees are burning through pricey AI tokens just to meet the illusion of output, only to deliver content that adds little value. Johnny Taylor Jr., chief of the Society for Human Resource Management, summed it up in a recent interview: "We’re driving the truck and building it at the same time, and that’s a problem." He added that, in hindsight, most firms wish they’d taken a slower, more thoughtful approach to rollout.

What makes the situation even more precarious is the hidden cost of “AI incidents.” A September survey by Ernst & Young found that 36 % of senior AI leaders at companies with over $1 billion in revenue have already faced a catastrophic incident – ranging from data loss to brand damage – that left a dent in the bottom line.

All of this paints a clear picture: the hype‑driven rush to weaponize AI in the enterprise has largely back‑fired, prompting a wave of budget cuts and a more cautious stance on future investments. Even Microsoft, whose own fortunes are tied to selling AI promises, is now re‑examining how much of its own product suite truly delivers measurable gains.

So where does that leave the average office worker? For now, the answer seems to be a return to fundamentals – clear goals, better data hygiene, and perhaps a little less reliance on generative models that churn out “slop.” The hope is that, with a more measured pace, AI can finally move from a flashy expense to a genuine productivity partner.