RESEARCH

Where AI is actually starting to kill jobs

AI is already squeezing some job categories more than others. 

On Wednesday, Goldman Sachs published a report showing that AI is starting to press the labor markets of major economies, with industries that have a greater exposure to AI automation seeing a slowdown in job growth over the past four years. The research found that job growth in information and communication services has slowed across all major economies since 2022. 

This includes white collar jobs like call centers, software publishing, management consulting and advertising, all of which have fallen far below historical trends, largely as a result of AI, according to the Goldman Sachs research. 

• For instance, call center employment is down 39% the historical average in the US, 33% down in Canada, and 27% down in Germany. 

• Additionally, entry-level workers are being hit the hardest by AI, especially for those in occupations that are more exposed to AI, such as the ones listed above, the report finds. 

This is coupled with data that shows that AI adoption is spreading widely across these economies. Goldman's research found that adoption rates have increased between 15% and 20% in developed markets, with France, the US, the Netherlands and the UK leading. Adoption is on the rise in emerging markets, too, sitting between 10% and 15%. 

"Overall, our analysis confirms that the conclusions from the US hold globally," wrote Goldman Sachs' Sarah Dong, macro research analyst, and Joseph Briggs, economist, in the report. "AI-related hiring headwinds are clearly visible in official and unofficial employment data, but impacts are limited to a narrow set of industries and workers." 

Goldman's data marks the latest addition to a growing pile of evidence that AI is impacting specific segments of the labor market. However, the broader signals are still largely mixed. Some estimates suggest that companies investing in AI have grown their headcount by as much as 10%, with entry-level hiring rising 12%. Other studies estimate that a large percentage of US work hours can already be automated. Further, some research finds that half of those laid off will be rehired for new jobs, and that AI actually increases the scope of work, rather than narrowing it.

Despite the mixed messaging on the efficacy of replacing humans with AI, layoffs are happening either way, especially for white-collar fields. The US made up 82% of the nearly 160,000 tech layoffs in 2026, and almost half of those were tied to companies restructuring around AI and automation, recent data finds. However, the truth is that AI has become a scapegoat for companies that want to trim headcount. In the long run, this is bound to backfire in two ways. For one, the question of AI ROI is ever-present. While layoffs may provide a temporary buffer that appeases stakeholders, those returns won't last forever. Additionally, blaming layoffs on AI is bound to worsen sentiment around it, both organizationally and publicly, creating resentment among staff that didn't get laid off and creating friction about internal AI deployments. The bottom line is that companies need to tread very carefully in their messaging about AI and its impact on jobs.

TOGETHER WITH CRUSOE

Fine-Tune and Deploy GLM-5.2 Without the Complexity

You need a coding model that remembers the whole task. GLM-5.2 delivers with a 1M-token context window that actually holds onto complex engineering decisions.

But don't just run it—customize it. Crusoe makes it seamless: Fine-tune on your proprietary data without managing a single GPU.

Deploy your private instance in minutes with our Self-Serve platform, or grab your weights to go. Whether you need an agent for legacy refactoring or new feature development, get the performance you demand with zero DevOps headaches.

Your model, your data, your deployment—fully on your terms.

Start Fine-Tuning GLM-5.2 now

CULTURE

How AI lost its most natural early adopters

Younger generations have always been the first to welcome change and new technologies, and AI was no different. But a new report suggests the tide is turning.

Since the Pew Research Center first surveyed Americans' concerns about AI in 2021, negative sentiment has evolved. The 2026 survey, published on Tuesday, found that 52% of Americans say they are more concerned than excited about the increased use of AI, up from 37% in 2021. However, the most notable finding is an unusual pattern among young adults: for the first time, a majority of adults under 30 (55%) say they’re more concerned than excited about AI. 

Those figures place young adults' concerns on par with those across all age groups, including those in their 30s and 40s and those aged 65 and up. That negative sentiment also shows up in job market concerns, which, while cited as one of the top reasons that adults fear AI since 2021, the report found that the youngest adults are now just as likely as those aged 30 to 64 to hold that view. 

Although the young adults' sentiments are on par with those of the other generations, with 71% of adults thinking AI will lead to fewer jobs in the United States over the next two decades, up from 64% last year, it is especially notable since it can impact career choices of the next generation, and as a result, impact the overall job market. After all, concerns about AI jobs begin as early as age 13, when students start thinking about how to align their high school courses with long-term career goals, according to a previous Pew Research report. 

"In general, young people often give us pretty clear indications of where things are going. And people of that generation are very concerned about AI," Dr. Robert Prey, Professor of Digital Culture at Oxford, told The Deep View in a recent interview. "Most technologies are really enthusiastically adopted by teenagers. Parents are worried about smartphones, or are worried about the internet, and teens are usually the first to get into it, and are usually completely sold on these things. With AI, we're seeing something different."

AI pressures are already shifting how job roles are viewed, as Gen Zers are more open to careers in trades than traditional corporate jobs. Specifically, a recent LinkedIn report found that nearly 6 in 10 Gen Zers in France (65%), the US (60%), Germany (57%), and the UK (55%) said technical trades offer more meaning than an office job. This trend has also been documented on social media and in other reports, news stories, and social media.

Even Geoffrey Hinton, Nobel Prize winner widely regarded as the Godfather of AI, publicly advised people in a podcast interview to consider the trades. "Train to be a plumber," he said.

It's interesting to watch how younger generations are changing their tune on AI, because even as the technology keeps advancing, a more fundamental question emerges: how willing are people to actually use it? Historically, demand for cutting-edge technology has been driven by younger, tech-savvy consumers. But what happens when that generation isn't just hesitant, but actively resistant, even opposed, to adopting it? Gen Z has already shown a willingness to push back against workplace norms it disagrees with, popularizing the term "quiet quitting," meaning doing only the tasks required by a job description rather than going above and beyond. Now, despite heavy AI investment and adoption efforts from employers, companies may be facing a harder problem to solve: getting employees in non-AI roles interested in using AI for everyday tasks in the first place.

TOGETHER WITH DESCOPE

The biggest MCP spec update since June 2025 just landed

Every team shipping MCP servers should pay attention to the July 2026 spec revision. Sessions are gone. DCR is deprecated in favor of CIMD. And there are six new authorization SEPs your server is now expected to handle.

This developer breakdown from Descope covers:

• What changed in the transport layer, and why sessions were removed

• The six new authorization SEPs explained

• Enterprise-Managed Authorization (EMA) and what it means for consent sprawl

• A migration checklist to run against your existing MCP server

Read the breakdown before your next MCP deployment.

PRODUCTS

Snowflake targets AI costs with model routing

AI agents can handle complex tasks autonomously, but not every job demands the same horsepower. That's where model routing comes in.

The concept is simple: model routing matches AI models to the complexity of the task, saving users and enterprises money by calling on lighter, cheaper models for most jobs and reserving the more costly ones for when they are needed. Snowflake just became the latest company to unveil its own dynamic model routing solution within Cortex AI Gateway and its flagship AI products, Snowflake CoCo and Snowflake CoWork.

“Enterprise AI is moving toward a world where companies can draw on the best available intelligence for each task without having to manage the model landscape themselves," Baris Gultekin, VP of AI at Snowflake, told The Deep View. 

"Snowflake is building toward that future by helping customers benefit from rapid innovation across the model ecosystem without adding more operational burden. Over time, that flexibility will make it easier for enterprises to adopt AI more broadly, reduce costs, and turn model advancements into sustained business value."

Another major advantage of model routing, as highlighted in the blog post, is that it lets enterprises skip rebuilding agents or apps for different models. In internal evaluations, Snowflake said agents using dynamic model routing with Cortex AI Gateway to build a dbt pipeline saw up to 3x greater token efficiency than a frontier-model-only path, without compromising quality. In another test, engineers completed the same number of pull requests with 25 percent greater token efficiency. 

NVIDIA just released its own routing solution last week: NeMo Switchyard. It is an open-source library for smart routing within popular agent tools, allowing enterprises to build routers to meet their needs and take advantage of the same benefits as above: lower costs that don't compromise performance or require rebuilding apps. 

In a conversation with The Deep View, Kari Briski, VP of generative AI software for enterprise at NVIDIA, also highlighted the performance benefits model routing can offer by rerouting tasks to domain- or task-specific models, in addition to cost-efficient ones. 

"Simple tasks or even niche tasks need to be domain-specific, and so that results in not just faster responses because you're sending to maybe smaller, more efficient models, but then greater token efficiency, and then even in some cases a higher accuracy based on your domain," said Briski.

Underpinning the development of model routing is a broader industry trend towards efficiency and cost reduction, which is imperative as costs continue to skyrocket. A Gartner report published this week predicts that AI inference costs per agentic workflow will increase more than fivefold through 2028. When asked if these rising costs would soften AI demand, Scott Bickley, Advisory Fellow at Info-Tech Research Group, told The Deep View that efficiency would be the focus: "The initial response will be for a focus on efficiency and more intelligent use and consumption via an array of techniques such as fine-tuned models, small LLMs, model routing, caching, distillation, prompt compression, etc., rather than mass abandonment of AI." 

LINKS

• Flock Safety built OS Investigate AI system for plate scans, arrest records

• Anthropic prepares supervoting shares for founders ahead of IPO

• Irregular AI lab report on major AI labs’ hacking draws criticism over gaps

• OpenAI expands ad pilot to 31 European markets as revenue climbs

• Anthropic details Claude experiments in protein design, analytical chemistry

• Z.ai's latest model, GLM-5.3, shows major post-training boosts

• Claude: can now send emails in Gmail and manage files in Google Drive

• Perplexity: Computer now works in email

• MAI-Image-2.6-Preview: landed at #3 in Single Image Edit in the Image Edit Arena

• Replit: OpenAI’s Luna model powers new ‘Free Mode’

• Claude Desktop: now starts ~2x faster than it did a month ago

• Freshworks: Principal AI Solutions Engineer

• Abridge: Software Engineer, Generative AI

• Boom: AI Engineer

• Wispr Flow: ML Engineer

GAMES

A QUICK POLL BEFORE YOU GO