WORKPLACE

AI adoption leads companies to hire more, not less

Despite fears of AI replacing workers, companies making the biggest bets on the technology are still hiring.

Companies that invested most heavily in AI grew their headcount by 10% over the past two years, according to a recent study from Ramp and Revelio Labs. The researchers analyzed AI spending and workforce data from more than 21,000 U.S. companies. Entry-level hiring rose by 12% among the heaviest AI adopters.

AI adoption was uneven across industries. Companies seeing the strongest headcount growth were larger, more engineering-intensive, and more likely to be venture-backed, particularly in information, finance and insurance, and professional and technical services. Adoption was far less common in industries such as healthcare, accommodation and food services, and arts and entertainment.

Employment growth was broad across job functions. High-intensity AI adopters saw statistically significant increases in engineering, sales, customer service, finance, and administrative headcount, suggesting that AI usage augments the workforce and increases overall economic activity. 

"If AI lowers the fixed cost of building software, handling administrative work, doing analysis, or improving customer support, the gains can drive outsized growth and unlock new revenue streams that previously required higher fixed costs in the form of new salaries," Ara Kharazian, Ramp's lead economist who worked on the study, wrote in a blog post. 

The findings challenge the narrative that AI adoption inevitably leads to fewer jobs. Headlines about companies citing AI in layoffs, along with warnings from AI leaders about widespread automation and the loss of entry-level jobs, have fueled fears about AI eating the labor market. This study paints a more complicated picture: among companies making the largest investments in AI, hiring continued to grow.

The researchers caution that the findings shouldn't be generalized across the broader economy. Instead, they offer an early look at how heavy AI adopters are changing their hiring patterns.

"We believe [employers] are selecting for a new set of skills, specifically, people who know how to use AI and use it well," Kharazian wrote.

These findings should be taken with a grain of salt. The study focuses primarily on larger, tech-forward companies and doesn't capture how small businesses are using AI or whether companies rely on free AI tools. It also shows a correlation between AI spending and hiring, not that AI directly caused companies to add workers. Even so, the study points to an important distinction: companies seeing the strongest growth don’t just slap AI onto their workflows. They invest heavily in it and integrate it across their organization. While it's still too early to know which jobs AI will ultimately replace, it's important to note that AI literacy is quickly becoming a baseline expectation for workers.

TOGETHER WITH CODER

The Next Big Test Facing AI Coding Agents

By now, we all know the standard watch-outs when it comes to AI agents. Reliability, security, privacy, and the like are all real concerns for any company… but for those of us working in regulated industries (think financial services or government), these “watch outs” can quickly become massive issues – and that’s no good. 

Coder recognizes the scale of this problem, which is why they commissioned this new report from Weave Intelligence to help us address it. Inside, you’ll find a complete breakdown of the requirements for AI coding agents in regulated industries, a practical framework for deploying these agents at scale while still maintaining security and compliance, and how to do it all while avoiding those dreaded operational bottlenecks. 

Download the whole report for free right here. Thanks, Coder!

STARTUPS

Why world models, not LLMs, will deliver physical AGI

The term AGI has been thrown around so much that no one really knows what it means anymore. But an increasing number of experts think the key lies beyond LLMs. 

In this episode of The Deep View Conversations, we sat down with Caroline Ingeborn, COO of Luma, an AI lab dedicated to omnimodal intelligence, to discuss where generalized physical AI could take us. 

Rather than focusing only on video models, Luma takes an omnimodal, or multimodal, approach, creating models that understand text, video, images and audio. This, said Ingeborn, is because humans don't think in one modality. 

These kinds of models have many potential use cases and could even help researchers achieve general intelligence. Luma's primary audience right now is the creative industries, such as entertainment, advertising and marketing. Ingeborn said she sees the technology as enhancing the creative experience rather than replacing creative professionals. 

Topics covered include:

• Where AI fits into creative workflows

• The ethical lines of AI in creativity 

• Luma's mission towards generalized physical intelligence

• Physical AI's potential impact on the labor market

• The different approaches to building world models

• The dangers of centralized power in physical AGI

If you're following the progress of world models, physical AI and robotics, and the use of AI in creative fields, then this episode offers a deeper look at how these technologies are being used today and the transformative impact they could have in the future. 

📺 Watch on YouTube: https://youtu.be/ISueKJW-uqg

🎧 Listen in your favorite podcast player: transistor.fm/shows/tdv/episodes/53-how-to-rethink-agi-through-multimodal-systems-caroline-ingeborn

Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology: tdv.transitor.fm

TOGETHER WITH CONVEX

Code Faster, Easier, And Less Buggy With Convex

You’re probably already familiar with Convex, the backend building platform that helps your AI agents excel – but their latest updates are taking things up a notch.

Convex recently launched plugins that allow you to connect Claude Code, Codex, GitHub Copilot and more directly to your Convex deployment, allowing them to…

• Read data and run functions entirely on their own

• Optimize your project’s performance via production insights

• Improve safety by limiting access to production

There’s even more updates coming in the near future, but for now, see how you can take your agents’ work to the next level with the latest from Convex right here.

RESEARCH

How AI is rewiring scientific discovery at Google

AI's transformative effect on software development is well known. Next up for transformation: scientific research.

Lizzie Dorfman leads science AI at Google Research, where her team develops AI systems to help scientists solve problems across genomics, neuroscience, epidemiology and climate science.

In a conversation with The Deep View at Google I/O in May, Dorfman explained how Google's own researchers have adopted AI agents, why they can now explore hundreds of thousands of scientific ideas instead of just a handful, and why the biggest breakthroughs often come from solving smaller bottlenecks along the way. 

This interview has been edited for brevity and clarity.

Jason Hiner: What's top of mind for you right now?

Lizzie Dorfman: We've done AI and science for a decade... But Gemini was a real sea change. What came out of three years of eating our own dog food is what we call ERA: Empirical Research Assistants. And it's shockingly powerful. But we're at a point where basically 100% of our team, across all these domains, this is how we do our work. We watch them go through a not-terribly-long tunnel from "okay, fine, I've heard you talking about this" to "oh, wow." I heard from someone I considered fairly grumpy and quite serious, who emailed to say he'd just gotten a transformational result he was almost done writing up. And we're mostly pretty grumpy, pessimistic people who say we'll believe it when we see it.

Jason Hiner: How does that actually change the work?

Lizzie Dorfman: For us it ended up being coding agents using a tree-search methodology, where you explore hundreds or thousands of different approaches to solving a problem computationally. That's what gets you extraordinarily creative and performant solutions. We have an epidemiological forecasting example with top-scoring models in the CDC competitions. We generated 200,000 models to evaluate. Most were poor and got shed immediately. On a traditional team, someone might score a couple of ideas. People describe it as: I input some ideas, went to sleep, and woke up with all these cool results. Before, it was serial and iterative... Now it's possible to explore outlandish ideas just to see, because it's trivially more expensive to you, certainly in terms of your intellectual time.

Jason Hiner: So does this reduce the need for human expertise?

Lizzie Dorfman: No. I love this example: solar panels are flat, but what if they weren't?... Someone wrote a paper on curved panels in 2012... We fed it in, reproduced their results, then asked the system to make it better. It did, and voila. But when we looked, it had these photovoltaic pieces that were levitating, not physically connected. It was cheating because you can maximize energy if you don't have to adhere to physics. So we added a loop that checks the solution is physically valid. You can't just shut your eyes, and it's done... Expertise and careful verification are still required.

Jason Hiner: When you take on grand challenges, how do breakthroughs actually come?

Lizzie Dorfman: There's a phase where it's "this is impossible, it's infeasible." ... For brain mapping, we need to bring the computational cost down by multiple orders of magnitude. The breakthrough takes a lot of forms. It's not always something that looks like AlphaFold... It's frequently "oh, wow, another order of magnitude." One big breakthrough is often a series of obstacles overcome. In the last four years, we've had something like 45 papers in Nature and Science. That's results. It's not "hey, blog post, we did a thing."

Read the rest of the interview with Google Research's Lizzie Dorfman.

LINKS

• US in talks with AI labs to create voluntary standards for new models

• Microsoft invests $2.5 billion in embedding AI engineers in customer orgs

• OpenAI proposes giving Trump Administration 5% stake in the company

• Germany's SAP bets on reinventing jobs with AI, not cutting them

• SpaceX has reportedly previewed AI device to investors

• Why AI is spurring talks about Universal Basic Capital

• Weave Robotics launches home robot Isaac 1, deliveries start in fall

• Some high-earners are using AI schools to create custom education for kids

• ZCode: China's Z.ai has launched a coding competitor to Cursor and Claude Code. 

• Notion HTML: Users can now build interactive HTML on their Notion pages using AI. 

• S2.1 Pro: Fish Audio's best voice model is now free for developers. 

• Nemotron-Labs-TwoTower: A diffusion language model from NVIDIA adapted to write tokens in parallel instead of one at a time.

• OpenAI: Agent Post-Training, Context Research

• Fractional AI: Developer Relations Engineer

• Capital One: Applied Researcher I

• Salesforce: Senior Offensive Security Engineer (Red Team)

GAMES

POLL RESULTS

Are you concerned about government intervention in the release of powerful AI models?

Yes (37%)
Somewhat (30%)
No (28%)
Other (5%)