93% of jobs in the USA can be done at least partially by AI, according to a new study, and companies could shift more than $4.5 trillion in labor costs to AI. Researchers studied more than 18,000 tasks across 1,000 jobs to determine where AI could be applied. The upshot: AI capability is growing fast, and it could soon take over an even larger segment of the economy.
“Adoption isn’t there yet,” Cognizant CTO Babak Hodjat told me in a recent TechFirst podcast. “Some sectors are growing faster than others. Also, innovation in AI has been so rapid — new breakthroughs every few months — that this $4.5 trillion potential will likely grow.”
The report is an update of a 2023 study Cognizant, a global IT services and consulting firm with over $20 billion in annual revenue, conducted across the same number of jobs and tasks. The upshot: AI-driven change is coming sooner than expected.
“The surprise was that some of the tasks we were expecting to be automated later on are already being automated,” Hodjat told me. “Some of the breakthroughs, for example in agentic AI and multimodal AI and so forth, are already impacting some of these tasks.”
Software development is one of the most heavily impacted areas, as LLMs have improved in their ability to code. Others include business and financial operations jobs, management roles, office and administrative support roles, and legal analysts.
According to the report, the top six most-impacted jobs by percentage of tasks that can be done by AI are:
- Financial managers: 84% impacted
- Computer and mathematical roles: 67% impacted
- Business and financial operations: 60-68% impacted
- Office and administration support: 60-68% impacted
- Legal occupations: 63% impacted
- Management jobs, including C-suite: 60% impacted
The software development impact has become very obvious over the last few months.
“Pretty much 100% of our code is written by Claude Code + Opus 4.5,” Anthropic lead engineer Boris Cherny said in January (Anthropic makes the Claude LLM). “For me personally it has been 100% for two+ months now, I don’t even make small edits by hand. I shipped 22 PRs yesterday and 27 the day before, each one 100% written by Claude.”
(A PR is a pull request: a request to merge new code into existing codebases.)
The least-impacted jobs include:
- Construction: 12% impacted
- Mechanics: 17% impacted
- Installation and repair: 20% impacted
- Protective services (police, guards): 20-29% impacted
- Personal care roles (childcare, elder care, etc): 20-29% impacted
- Healthcare support roles: 29% impacted
It’s important to note that Cognizant’s exposure score reflects the percentage of a job’s tasks that are AI-automatable. So if a role has 60% exposure, it means roughly 60% of its tasks could theoretically be assisted or automated by current AI capabilities. It does not mean 60% of those jobs disappear, or that 60% of workers in those jobs are replaced.
It may mean, however, that companies will need fewer people doing those jobs, or that people in those roles can do a better job at their careers than before, thanks to AI assistance.
So AI’s reach doesn’t equate directly to job replacement. Hodjat stressed that human context and expertise remain essential.
“Humans are very contextual. We have effective memory and deep expertise in our niches, and creativity. There are fundamental limitations in today’s state of the art that prevent AI from handling everything,” he said.
It is interesting to see that some jobs we might have thought were completely safe – construction jobs like carpenters and framers – are seeing some AI impact.
“In transportation, we saw a rise from 6% exposure in 2023 to 25%,” Hodjat told me. “In construction, from 4% to 12%. Those are big jumps.”
That AI impact in very physical jobs is less about replacing hands-on work and more about augmenting inspection, diagnosis, planning and documentation.
On the flip side, programmers can breathe something of a sigh of relief: while AI can code, it can’t connect everything. Nor, as any vibe coder knows, does it generally get everything right on the first try.
Interestingly, if the value of labor that could be done in the U.S. today is $4.5 trillion, global impact would approach $15 trillion. And that’s at the current state of AI development. But AI, as we know, is not standing still.
Get a full transcript of our conversation here.
