AI is changing work. Are local leaders ready?

At a bar in Pittsburgh’s Strip District this week, robotics organizers were worrying about a dystopian near future in which wealthy technologists command armies of poor gig workers through targeted AI agents. Hundreds of miles away, an entrepreneur support group in Louisiana was optimistic about how AI could superpower local founders.

To this, New Mexico economic developers met at the Albuquerque conference just shrugged.

There are so many conflicting predictions that most workforce development programs don’t even try to adjust their strategies.

He told me that there are so many conflicting forecasts for the next 12 to 18 months that most workforce development programs aren’t even trying to adjust their strategies.

A year ago, Mark Zuckerberg said half of his Silicon-era company’s software code would be written by AI, but now Microsoft says about 30% of its software code is already written by AI. Last month, Microsoft’s head of AI predicted that within the next 18 months, all computer tasks will be better done by AI. Recent analysis suggests that in an era of high interest rates and geopolitical turmoil, a third of jobs will be at risk by 2030.

A New Mexico leader said this to me over late afternoon coffee. “Battle the hatch.”

Amid all this noise, what are the signals for local leaders to follow?

What local leaders need to know about artificial intelligence

The term “artificial intelligence” makes many researchers and computer scientists nervous because of its imprecision. AI basically refers to computer programs that mimic human cognition. But there are many ways to do this.

Thirty years ago last month, IBM DeepBlue became the first mainstream example of “artificial intelligence.”

The computer program performed millions of calculations, ranked chess moves based on human-coded rules, and became the first to defeat Russian grandmaster Garry Kasparov. However, this turned out to be useful only in highly controlled environments with predefined rules and clear outcomes, such as chess.

Successive breakthroughs, such as advances in autonomous improvement in “machine learning” and deep learning (which basically means ML with large amounts of data), have led to today’s AI craze. Then, in 2017, an influential paper by Google researchers identified a more efficient way to connect large amounts of unrelated information.

This led to today’s “large-scale language models” that predict the most likely response to a prompt by running probabilities on a given dataset. Although they can produce amazing results, they ultimately rely on training and imitation of the source. One influential paper called them “probabilistic parrots.” Today’s LLMs are best known for their incredible scale, being trained on virtually all available digitized human-generated content and performing calculations at the speed of light, typically via cables between devices and data centers.

LLM is just one example of so-called artificial intelligence. Artificial intelligence itself is a branch of computer science.

It remains to be seen whether this is the path to lasting breakthroughs with self-learning systems, or simply the most popular tool at the moment. In any case, we will know more in the coming year, and whether this moment is seen as the dawn of a new era or a momentary, exciting flash in a field that is decades old will only be fully understood in the coming years.

In any case, this happens fast enough that we survive until we can verify our claims and face the consequences.

Even really big changes take time to penetrate.

For advice, I turned to Mr. Kumar Garg. He is now the head of Renaissance Philanthropies, a flashy D.C.-based global funder focused on science and technology. Like other alumni who embraced the technocratic idealism of the Obama administration, he is sharp, optimistic, and wears stylish glasses.

“AI is a pretty positive story, and the early chapters are generally positive,” Garg says. “We’ve been able to build AI models that can predict the structure of proteins, something that previously took graduate students years to do. Now it’s being done at scale.”

He added: “The real concern is where it’s going.”

To make sense of the onslaught of AI predictions and hot takes, I rely on a 2-by-2 square. One axis is the seriousness of AI as a breakthrough. Is the impact no greater than other seismic paradigms we have experienced (such as the introduction of the Internet), or is this self-learning path something different (more like the use of fire)? In his last book, pollster Nate Silver introduced the Technology Wealth Scale. Credit cards are 7, electricity is 8, wheels are 9. Another axis is whether these technologies bring us more or less benefits.

The top right is a very positive story about an unprecedented breakthrough that will end scarcity. The bottom right is apocalyptic, with these technologies threatening humanity. The bottom left is the most important. AI is an overhyped benefit meant to enrich the already wealthy. cards on the table. Personally, it’s closer to the top left. Something similar to the Internet, which was really disruptive, but brought us more benefits over time.

The totality of credible and honest debate leaves us at an impasse. I pity the local leaders who get yelled at by overconfident voters in each of these quadrants.

“‘Trust us’ doesn’t work,” Garg says. “We have to build tools and show how we are mitigating the downside.”

The main concern across stances on AI is around jobs. Our culture wraps health and wellness in socio-economic measures. So the question is whether this kind of emerging technology creates more jobs than it destroys. Garg reminds local leaders that no matter how quickly these projected improvements occur, they have time to adapt and learn.

“Even really big changes take a huge amount of time to move forward across society,” Garg says. “They have second- and third-order effects.”

What the Internet can teach us: Adopt fast, destroy slowly

We’ve been here before. Did the Internet destroy or create jobs? The academic answer is both. People make mistakes in timing.

Home Internet access more than doubled between 1997 and 2000. By the end of that period, most large companies were already online. Recruitment happened quickly. Job losses came later.

• Travel agencies declined by 28% between 2000 and 2003 due to the proliferation of online bookings.
• Between 2007 and 2010 during the Great Recession, newspaper employment fell by 26%.
• With the rise of streaming, video rentals declined by more than 60% between 2008 and 2011.

The pattern is consistent. Implementation of technology came first. Job displacement then occurred, often several years later, and was usually accelerated by economic downturns. The Internet didn’t instantly kill the industry. It made them weak. The recession has done its job.

The AI ​​seems to be following the same script. Adoption is rapidly increasing. Generative AI tools are spreading throughout the workplace. However, measurable labor market transfers remain limited. This combination of rapid adoption and slow destruction may seem contradictory. it’s not. Technology first changes tasks, then companies, and then, often under external pressure, changes entire industries.

“That doesn’t mean everything will always go well,” Garg says. “We need to invest in managing the positives while considering the negatives.”

There is one notable difference. The Internet expanded during economic expansions and reshaped jobs during recessions. AI comes with higher interest rates, tighter capital, and more cautious hiring. That may mean the impact will be less in mass layoffs and more in hiring freezes, changed roles and changing expectations of what workers should do.

Some of it may turn out to be true in narrow areas. But history suggests caution. Technology advances rapidly. Labor markets rarely change.

“Work is actually very complex,” Garg says. “Humans are helping mediate between the technology and what the company is trying to do.”

What local leaders lack

Many states in the United States typically create a list of “high priority” occupations that determine what workforce programs are funded. A review of recent lists shows little adaptation to the status quo, major changes amidst the chaos, and cowardice overlooked.

• Virginia’s current dashboard allows filtering by region and education level, but does not include measures of AI exposure or skill variability.
• Claims adjusters, buyers and purchasing agents, and bookkeepers remain prominent in Pennsylvania, and these are precisely the routine, structured roles that researchers at the Bureau of Labor Statistics now say are susceptible to AI-related impacts. Gloat research shows that despite the number of AI-proficient workers increasing from 1 million to 7 million between 2023 and 2025, rural areas have few or no software-related jobs listed.
• Indiana appears to be an exception, as it overhauled its methodology last year and incorporated metrics such as “change in skills” and “likelihood of changing jobs.”

Most systems still rely on long-term forecasts built on pre-ChatGPT assumptions. This framework is increasingly disconnected from the pace of change its constituents are experiencing. This sign does not mean that AI will immediately replace workers. It’s that we’re getting into a familiar pattern. This means rapid adoption followed by slow and uneven labor market impacts, shaped not only by the technology itself but also by economic conditions and policy choices.

The question is not whether AI will reshape jobs. It’s already happening. The question is whether local leaders are ready to shape how that change happens.

“This is not something we sit on the sidelines about,” Garg said. “We can shape where technology goes.”

This is the foundation of Technical.ly’s editorial perspective. Our AI Ethics Policy is based on the old editorial principle that technology is neither good, bad, nor neutral. In other words, tools have no emotions. The risks of AI are less about what the tools can do and more about who is using them.

When I spoke at a conference in New Mexico last week, one of the attendees, an AI skeptic, asked me exactly this question. “What makes us so optimistic that we can use this class of technology for good?”

I regretted how naive I sounded, but I told him what I believed. Many of us would like to use these tools to do more good than bad. So, most importantly, help us responsibly use whatever tools are available to us to face the real issues facing our communities.

---