Behind the Curtain: Slow, hard AI

Sweet's hands-on experience matches recent studies by MIT, McKinsey and others, showing that successful AI use for large organizations remains elusive — the "productivity paradox." The viral MIT paper found 95% of the 52 organizations studied "are getting zero return" on their investment in generative AI.

• "AI, at the enterprise level, is hard," Sweet told us Monday in an interview from Toronto, where she was visiting clients. "I am talking to CEOs almost every day. Their frustration is mostly about: How do I move my organization fast enough?"
• "They recognize it's less about the technology, and more about the willingness to truly reinvent the work, the workforce."

How this affects you: An Accenture survey of 3,000 global executives, shared first with us, found 85% of the C-suites plan to increase AI spending this year — and expect big results. One likely casualty will be jobs, for three interrelated reasons:

1. Companies plan to invest heavily in training employees to use AI — but know that a good number won't make the transition fast or successfully enough.
2. You can't actually implement AI at scale without cleaning up your complicated, outdated processes. That means doing less, often with fewer people.
3. AI will do a lot of jobs now done by humans. Companies cut when companies can cut.

The big picture: Sweet's findings match what we hear from CEOs every day. But Sweet, like many others, believes that after a short-term hit, new jobs and businesses will be born.

• Accenture has retooled its entire business to help businesses make this transition — in part by helping them build a "digital twin." Sweet believes that soon, companies will have so much data in AI systems, including the internal methods of doing their work, that they'll be able to run constant simulations in their digital twin to vastly improve performance.

To do so, companies need more than AI-proficient staff. They need leaders, often different from the ones existing today, who can figure out how to pull AI levers across the company to truly move the needle meaningfully.

• "I tell CEOs a cross-functional steering committee is not a strategy," said Sweet, who was a law-firm partner before working her way to the top of Accenture, which has nearly 800,000 employees in 120+ countries, and annual revenues of $65 billion. "There's a real methodology to what you need to do."

• "If you take nothing else away from this discussion," she added, "you will never scale as long as AI [deployments] are separate projects where learners are not held accountable for delivering the outcomes. That made sense in the early days of using gen AI and agentic AI. ... Companies have been slow to move from that, to holding leaders accountable for now using the technology and delivering on those outcomes."

The long view: A New Yorker column points out that slow adoption of transformative technology is a phenomenon that goes back to the steam engine. At first, new technologies may actually "reduce productivity growth because they are so disruptive and difficult to incorporate into existing ways of doing things," the column says. "Only later do productivity gains show through — a pattern known as the 'J curve.'"

• Despite the short-term hurdles, Sweet said she's overall very positive about AI: "I get to talk to the CEOs who are going to use AI to do amazing things, whether it's in medicine or materials or energy efficiency. ... So I think that there's going to be a lot that people don't see today."

"It's going to be pretty interesting and fun," Sweet told us as she headed off, "to see how this can change the dynamic for those who really are willing to embrace the idea of reinventing everything they do."

• Go deeper: Why you should be AI-obsessed.