AI Agents Will Do More, But Humans Aren't Obsolete

Upwork CEO Says AI Won't Eliminate Human Work

Due to the government shutdown, the latest jobs figures are unavailable. It is noted, however, that the Trump Administration is reportedly dissatisfied with the U.S. Bureau of Labor Statistics and is seeking new leadership, following the withdrawal of E.J. Antoni’s nomination.

In the absence of official data, corporate leaders can look to private sources for indicators of a softening labor market, including data on how AI is influencing competition and employment.

A newly released annual pulse survey from Dayforce of approximately 7,000 workers across six countries reveals that 71% of workers have not received AI training in the past year, despite 63% acknowledging the importance of developing such skills.

The competition for AI talent is also evident on freelance platforms. While companies like Upwork and Fiverr have been negatively affected as generative AI displaces simpler contract work, Upwork’s CEO Hayden Brown reports that AI is simultaneously helping the platform attract larger enterprise clients.

Research she cites indicates that 63% of executives feel they lack adequate in-house talent, a factor that may explain a more than 50% increase in searches for skills like prompt engineering on Upwork last quarter. This raises several considerations:

Defining 'AI Talent': Brown states there are 250,000 AI experts on her platform, with 80,000 in the U.S. She describes a growing demand for an "AI generalist profile", professionals valued for creativity, problem-solving, judgment, and an ability to adapt quickly to technological changes.

As AI is applied to diverse functions, both technical and functional expertise are highly sought after, with platform ratings serving as a key validator of skill authenticity.

The Future of Gig Work: With the unemployment rate for Gen Z more than double the national average, freelance work is often a necessity rather than a choice.

While gig work carries risks such as a lack of benefits, burnout, and isolation, Brown emphasizes the strong desire for flexibility among younger generations, who prefer not to be tied to a single employer.

She advocates for a portfolio career approach, allowing workers to diversify income streams and maximize their return on investment.

Leadership Perspectives on AI: Leaders who view contractors merely as a cost-saving replacement for full-time employees are likely to see AI through a similar, narrow lens.

In contrast, leaders who perceive a flexible talent pool as a strategic asset for building a dynamic organization are positioned to fare better. Brown contends that even as AI agents take on more tasks, "there will be plenty of work for humans."

Walmart Seeks More Developers to Build AI Agents That Will Automate Their Work

A senior technology executive at Walmart has stated that the retail corporation plans to continue hiring elite engineering talent, even as it develops more AI agents designed to automate aspects of its work.

This focus aligns with CEO Doug McMillon’s recent comment that “AI is going to change literally every job.” McMillon acknowledged that while some roles and tasks at Walmart will be eliminated, others will be created.

The company is actively recruiting for new developer positions, such as an "agent developer" role created last month to deploy agents that automate complex workflows.

A company representative added that Walmart will continue to invest in upskilling, recruiting top-tier engineers, and redeploying staff to areas requiring uniquely human skills.

This hiring strategy persists despite a contraction in the wider IT job market, where economic uncertainty has made employers more selective, often favoring candidates with AI expertise. Many of Walmart's open engineering positions specifically request AI-related skills.

Company executives have projected that Walmart's global workforce will remain roughly the same size over the next three years, though the composition of jobs will shift significantly. This shift includes a greater reliance on AI agents, which are already boosting productivity for software engineers.

For example, an AI "super agent" named Wibey unifies over 200 other AI agents built by Walmart's developers, allowing them to share and utilize each other's creations.

These AI tools are delivering tangible results. One set of agents automates compliance with accessibility requirements, automatically identifying 60% of related software bugs and fixing 95% of them, leading to an eightfold productivity improvement.

AI is also used to help engineers modernize legacy code, reducing the time previously spent consulting senior colleagues.

While over 95% of Walmart's engineers use AI coding assistants, the company anticipates a future where AI agents can collaborate with each other. This advancement will introduce new management challenges, such as overseeing AI work performed overnight.

Karnati stressed that humans will remain essential in the process, as engineers are now discovering more complex problems that require human intervention, and the company is not at a stage where AI agents can operate entirely without oversight.

With Only 78 Examples, New Technique Creates Sophisticated AI Software Agents

A new study reveals that you don't need massive datasets to train powerful, autonomous AI agents. Research from Shanghai Jiao Tong University and the GAIR Lab shows that by focusing on data quality over quantity, you can achieve superior results with a tiny fraction of the data.

Their framework, called LIMI (Less Is More for Intelligent Agency), proves that true machine autonomy comes from strategically curating high-quality examples of agentic behavior, not from amassing enormous datasets.

In experiments, they trained models using a meticulously curated set of just 78 examples. These models then significantly outperformed others trained on thousands of examples in key industry benchmarks.

This means if you're working in an enterprise where data is scarce or expensive to collect, you now have a more practical path to building effective AI agents.

The Challenge of Building Agents That Actually Work

Think of "agency" as an AI's ability to function autonomously to actively discover problems, form hypotheses, and execute solutions by engaging with its environment and tools. In short, these are AIs that don't just think; they work.

The common assumption has been that to build this kind of intelligence, you need a lot of data, leading to complex and expensive training pipelines. But what if you could achieve more with less?

How the LIMI Framework Works for You

The LIMI framework shows you how to cultivate sophisticated AI agents using minimal, high-quality demonstrations. The key is in the curation process. Each demonstration consists of:

• A Query: A natural language request from a user, like a software development task.

• A Trajectory: The complete step-by-step process the AI uses to solve the query, including its internal reasoning, tool use (like running code), and observations from the environment. This trajectory captures the entire problem-solving arc, including mistakes and course corrections, so the model learns from the full process, not just the final success.

To build their dataset, the researchers started with real-world queries from professionals, then used an advanced AI to generate more. A team of experts then vetted and selected only the best 78 examples. They then used a human-AI collaborative process to complete these 78 tasks, capturing every step of the interaction to create rich, complex training trajectories.

What This Means for You in Practice

When the team tested their model trained on just 78 examples, it dramatically outperformed much larger base models on benchmarks designed for agentic skills, tool use, and coding.

This discovery fundamentally changes how you can develop autonomous AI systems. Instead of undertaking massive, costly data collection projects, you can leverage your in-house experts to create small, high-quality datasets tailored to your specific needs.

This lowers the barrier to entry and allows you to build custom AI agents that provide a real competitive edge on the workflows that matter most to your business.

Stay with us. We drop insights, hacks, and tips to keep you ahead. No fluff. Just real ways to sharpen your edge.

What’s next? Break limits. Experiment. See how AI changes the game.

Till next time - keep chasing big ideas.

Thank you for reading