Rapid advances in artificial intelligence are more likely to change how work is performed than eliminate large numbers of jobs, according to a new TD Economics report that says the conditions required for widespread AI-driven displacement have yet to materialize.
The report, by Senior Economists Rannella Billy-Ochieng’ and Thomas Feltmate, says a large-scale disruption would require AI to perform workplace tasks reliably and autonomously, be economically viable for businesses and be adopted rapidly across firms. TD Economics says those three conditions remain difficult to achieve simultaneously.
The report does outline a more severe scenario in which those barriers are overcome. Under two scenarios examined by TD Economics, the U.S. unemployment rate could rise by 0.7 to 1.4 percentage points above its baseline by the early 2030s as AI adoption accelerates and workers displaced by technology struggle to find new employment.
However, TD Economics describes that outcome as a risk scenario rather than its base case, saying current evidence points to a more gradual adjustment in which AI is integrated into existing workflows and changes the nature of jobs rather than eliminating them outright.
Technical capability is only one factor
The report says AI already has the potential to affect a substantial share of workplace tasks. Research covering more than 18,000 workplace tasks indicates that roughly half of all U.S. jobs have at least one-quarter of their tasks exposed to AI technologies.
TD Economics cautions that exposure should not be confused with automation. A task being technically capable of being performed by AI does not mean it can be completed autonomously or that an employer will choose to automate it. Jobs typically combine multiple tasks involving different skills, judgment and expertise.
The report says fully autonomous performance remains uncommon. One study cited by TD Economics estimates that less than two per cent of jobs contain more than half of their tasks that could be fully automated using AI combined with existing software tools.
Software development is an area where AI capabilities have advanced significantly, partly because computer-generated outputs can be evaluated objectively and feedback can be provided quickly. But TD Economics says that progress does not necessarily translate across the broader economy, where many jobs require tacit knowledge, interpersonal interaction and complex judgment.
Economics could limit adoption
Even when AI can technically perform a task, businesses still have to determine whether replacing or substantially altering human labour makes economic sense.
TD Economics cites research showing that while 36 per cent of occupations contained at least one technically exposed task in a study of computer-vision applications, only eight per cent contained a task that appeared economically viable to automate at scale. The report says implementation costs, oversight and operational risks can narrow the economic case for automation considerably.
Businesses also face costs related to compliance, privacy, cybersecurity, legal requirements and changes to existing workflows. The report says AI usage costs can vary substantially depending on the complexity of the task, with routine activities generally requiring fewer resources than tasks involving extended reasoning, planning or complex problem-solving.
TD Economics also points to uncertainty around the costs faced by AI providers, including computing infrastructure, data transfer, model training and hardware required to operate systems at scale. While some costs have declined, the report says rising costs for high-bandwidth memory and some advanced semiconductor components have created additional pressures within the AI supply chain.
Adoption is growing but remains uneven
The pace at which businesses adopt AI will also determine its effect on employment, according to the report.
Around one-fifth of businesses recently reported using AI technologies, based on the Business Trends and Outlook Survey cited by TD Economics. Adoption remains uneven across industries and company sizes, with larger organizations leading deployment efforts.
At the worker level, nearly half of workers surveyed reported having used AI at least once, while 13 per cent said they use it daily. TD Economics says the gap suggests that awareness and experimentation are spreading more quickly than deeply embedded, everyday use.
The report argues that this gradual adoption gives businesses, workers and policymakers time to adjust. Companies can retrain employees and develop future talent, while policymakers can consider measures to support employment and workers can build skills for an AI-enabled workplace.
Early effects are appearing in exposed occupations
Although TD Economics finds little evidence of broad AI-driven disruption in aggregate employment, it says there are signs of adjustment in occupations that are particularly exposed to the technology.
Employment growth in highly exposed occupations has slowed materially in recent years, with the effect more pronounced among younger workers. The report also says unemployment among the most highly exposed workers has been marginally higher than the overall rate.
The adjustment appears to be occurring primarily through hiring rather than widespread layoffs. TD Economics cites Indeed Hiring Lab data showing that occupations with a larger share of skills that generative AI can transform have experienced more pronounced declines in hiring.
The report also says businesses using AI are more commonly responding by retraining workers rather than eliminating positions. Data from Federal Reserve Bank of New York regional business surveys cited by TD Economics show retraining was reported more frequently than reductions in employment among firms using AI in 2026.
Severe disruption remains possible
TD Economics says a more serious employment shock could occur if AI becomes capable of reliably performing a broad range of tasks, businesses find the technology financially feasible and transparent, and adoption spreads quickly enough to outpace the labour market's normal adjustment mechanisms.
Its moderate disruption scenario assumes AI adoption accelerates and lifts annual productivity by 0.5 percentage points above the baseline by 2031. A more severe scenario assumes annual productivity rises by one percentage point above the baseline by that year, with the faster transition creating greater potential for worker displacement.
The report says the effects would not necessarily be evenly distributed across the economy. Industries with high concentrations of AI-exposed occupations could experience greater displacement, while other sectors could benefit from productivity gains, higher incomes, increased consumer demand and new employment opportunities.
TD Economics also warns that a recession occurring at the same time as rapid AI adoption could intensify employment losses by weakening confidence, demand and hiring.
The report concludes that the three conditions required for an extreme AI-driven employment shock are not currently aligned. As a result, it says the future of work is more likely to involve gradual change, giving businesses, workers and policymakers time to prepare for the technology's broader effects.