One gap, six different answers.

The diffusion gap among people who code will certainly shrink, since advances in AI coding capabilities have been rapid and dramatic.

But only about 1% of US workers are software engineers. Even accounting for other jobs where people write code, the fraction of people who code globally is likely a low single-digit percentage. Narrowing the diffusion gap more broadly will have to come from improvements in AI capabilities and interfaces across non-coding domains, simply by the numbers.

Coding will play a role, but it will not be the primary driver of closing the gap.

In fact, requiring coding can reinforce existing disparities. What we’re seeing is a shift toward hybrid models—where domain experts, engineers, and operators collaborate using tools that lower technical barriers. The most meaningful narrowing will come from system-level integration, not just increased coding proficiency.

Coding will help narrow the gap, but it’s not the primary driver.

Critical thinking, curiosity, and strong methodology are what matter most. As AI tools increase productivity, coding enables solutions—but human insight and ingenuity are still needed to guide innovation.

Likely. The more AI is coded into the products and services we already use, the more people will use the AI features of those products and services.

Coding will help narrow the gap by expanding what can be built and who can build, especially as low-coded and AI-assisted development tools mature. But coding alone will not close the diffusion gap. Many workers and small businesses will benefit from AI through applied use cases, workflow redesign, and problem-solving, not through traditional software development.

There is no clear correlation.

AI will benefit specific workers (e.g., enabling entrepreneurship through vibe-coding) and likely hurt others (e.g., call center workers). And the long-term consequences for training early-career workers and/or students remain unclear. Even within the same occupation, AI may differentially impact junior and experienced workers. Several case studies exist across work domains providing evidence to support either group as a beneficiary.

At the moment, it’s very uneven across workers and depends a lot on what tasks they’re using AI for.

Depending on the task, AI may perform the parts of a job that workers previously enjoyed doing themselves, or it may free workers up to make higher-level, creative planning decisions. There’s also research indicating that in some companies, AI is increasing the expectations placed on workers, leading to more work for them. This highlights that productivity benefits in workers’ tasks do not always benefit the workers themselves.

AI usage and worker benefit are closely linked, but impact depends on readiness.

Many emerging workers are using AI, yet gaps in industry-relevant skills remain. Benefits grow when training, upskilling, and support systems align with real workforce needs. Under-networked talent need hands-on experience, practical skills that increase employability, and robust professional networks that promote long-term career advancement in an evolving, AI-driven economy.

There is a strong correlation between AI usage and worker benefit—but only when usage is intentional and supported.

Workers see the most benefit when AI is integrated into their workflows in ways that augment expertise rather than replace it. Through our workforce programs, we’ve seen that pairing AI-enabled tools with hands-on training significantly increases productivity and career mobility. Without training and context, however, usage can be shallow and uneven. The key is structured pathways that enable workers to apply it meaningfully in their roles.

This is a significant concern.

Broad measures of digital skill coverage can mask deeper disparities in access to opportunity, infrastructure, and applied experience. In our work across Colorado and Wyoming, we’ve seen that rural communities and emerging talent pipelines are often underserved—not because of lack of interest, but because of lack of exposure to real-world application environments. Addressing the diffusion gap requires targeted, place-based strategies—connecting communities to workforce programs, testbeds, and industry use cases—rather than assuming baseline digital literacy alone will drive equitable outcomes.

Very concerned.

Averages hide who is being left out, especially rural residents, older workers, low-income learners, people without reliable broadband, and workers in small businesses. AI should be culturally relevant and appropriate to the people and cultures using it. With AI creators centralized in affluent urban areas, can they ensure the tools they create serve people not represented in the creation of those tools? With unequal access to the digital economy worldwide, we still have a long way to go to narrow the AI diffusion gap.

I am extremely concerned that focusing only on digital skills leaves behind vulnerable populations not in the workforce who still rely on technology.

Digital literacy must be inclusive and life-centered—supporting health, finance, and communication—so everyone can safely navigate an increasingly digital world.

It is concerning.

Digital skill coverage at a population or community level doesn’t capture the deeper layers required for meaningful AI use and inclusive access. Many residents still face significant challenges with basic digital skills, cybersecurity, tool awareness, and access to resources. AI adds new dimensions of vulnerability, capability, and confidence that people need time and support to build. Without equity-centered design and community-rooted support, these gaps widen and entire populations are left behind.

Tracking global AI diffusion is essential because adoption patterns directly influence economic competitiveness, national security, and workforce readiness.

AI is a capability embedded in systems, industries, and supply chains. Understanding where and how it is being deployed helps identify gaps, inform policy, and guide investment. From our perspective, it also highlights the importance of regional ecosystems as part of a global network, ensuring that innovation is distributed across communities contributing to real-world solutions.

Measurement is critical. AI is an important new technology that will have lasting economic effects.

Tracking how these effects are distributed helps us understand whether this new technology is closing or widening global inequalities and helps inform global economic policy, investment, and model development. It’s also vital to measure what type of applications AI is most and least useful for so that we can improve the systems and encourage beneficial use cases.

Tracking how AI is diffusing globally is critical to understanding who is gaining access—and who is being left behind.

It helps identify gaps in resources, skills, and infrastructure, while informing more inclusive strategies for education, workforce development, and innovation. By measuring diffusion, we can ensure AI benefits are distributed equitably and intentionally. The rate of traction and adoption alone aren’t enough; equitable participation is essential for meaningful, shared global progress.

Global diffusion patterns show us where economic advantage is growing and where exclusion is taking hold.

They reveal emerging shifts in labor markets, supply chains, and new centers of innovation. For rural regions, global data helps us understand what is possible and where there are different values, barriers, or concerns influencing AI diffusion. These insights help us design locally relevant strategies that reflect our own assets and constraints rather than relying on assumptions shaped by large urban areas.

People—human capital—are our most powerful asset.

Advocates, educators, builders, creators, and culture bearers drive trust and adoption. By investing in people to lead and contextualize AI within their communities, we build inclusive ecosystems that ensure innovation reaches everyone and no one is left behind.

Real world stories, practical use-cases, and word of mouth.

For people to really see the power of AI, they need to see it firsthand, from someone they trust, to solve an actual problem they see value in solving. In rural communities, trusted local organizations, workforce boards, libraries, community colleges, small business support organizations, and employers are often the bridge between awareness and adoption.

The most important tool is the network of trusted local partners and community organizations that already support learning, connection, and workforce development.

People adopt new technologies when they are introduced by someone they trust and when the learning environment feels safe. When we equip these partners with high quality training and time to practice, adoption accelerates across entire communities.

I think we collectively need to learn how to effectively incorporate AI into existing educational institutions.

My belief is that, once we solve problems like student assessment in the age of AI, AI will create opportunities for more students to participate successfully in education. For example, AI chatbots can be used as a tireless always-available and conceptually flexible tutor. Students who were on the margin of participating or succeeding in a field of study (especially a STEM field) will have an easier time keeping up with better-prepared peers with the help of AI.

The two things I am most concerned about are the uneven distribution of benefits from AI and the effect of AI on aggregate human expertise.

I think the biggest mistake could be incorporating AI into our work and education in ways that rob us of the opportunity to develop deep expertise and critical thinking. Even with powerful LLMs, we still need people to make important decisions and have great ideas. AI can provide complementary expertise and skills, but we need to make sure we use it in ways that don’t cause our own abilities to atrophy.

Unregulated AI. The mistake would be assuming the market alone will distribute AI’s benefits fairly, safely, or geographically evenly.

Without intentional investment, safeguards, and community-level capacity building, AI could deepen the same divides that broadband, education, and labor market systems have struggled to close. We are still trying to catch up with the negative impacts of social media. If we play catch up on unregulated AI in 10 years, we may never have the chance to correct our mistakes.

Our biggest mistake would be letting technology outpace our humanity.

AI must never replace kindness, empathy, ethics, and moral judgment. It should free us—not disconnect us—from one another. If we focus only on advancing technology and not the people behind the screens, we risk building a future that accelerates tools while eroding trust, dismantling community, and weakening the human connections that ultimately make our world more just, compassionate, and worth living in.

The next decade’s biggest mistake would be to limit the development and deployment of local LLMs because of rising hardware costs and the lack of a global solution to bring down costs.

I think companies will eventually realize that many forms of AI work well-enough with open-source LLMs run locally. Similarly, many universities today have enterprise AI accounts that exclude AI coding tools because of data privacy concerns; yet it seems increasingly essential that universities train current students to use modern AI tools. Local LLMs run within the university’s networks are again a potential solution here.

One of the biggest mistakes would be treating AI diffusion as a solely a technology skills problem.

If we focus only on teaching people how to use tools, we miss some significant factors that can slow AI diffusion. AI integration changes workflows, roles, and relationships. It raises questions about policy, resources, risk, and benefits. Skills training alone cannot solve these challenges. Diffusion depends on whether the surrounding conditions support meaningful integration.

The biggest mistake would be equating access to AI with impact, and failing to invest in translation.

If we assume that simply providing tools will lead to widespread benefit, we risk widening the gap further. Real progress depends on deployment—integrating AI into systems, infrastructure, and decision-making environments—and ensuring people are prepared to use it. Without sustained investment in workforce development, regional ecosystems, and applied use cases, we could end up with advanced technology that delivers value to only a narrow segment of society.