Stop Handing Out AI Tools Like Car Keys. Build Drivers First.

​In this segment of Joseph Ours’ Forbes Technology Council column, Joseph talks about how to structure your organization for AI tools.

Technological change is one of five key drivers expected to create 170 million jobs while also displacing 92 million jobs by 2030. Technology advancements will drive both the “fastest-growing and fastest-declining roles,” according to the World Economic Forum’s 2025 Future of Jobs Report.

The data tells us what may be coming, but not what to do about it. While company leaders fixate on whether AI will take jobs, organizations face a more immediate challenge: They don’t know how to structure work for AI agents. Without this knowledge, companies can’t capture productivity and other gains that justify workforce transformation. They experience displacement without seeing the benefits.

For an example of this already happening, look to the job market. An Oxford Economics report shows that recent college graduates are unemployed at a higher rate, with “signs that entry-level positions are being displaced by artificial intelligence at higher rates.” Organizations are cutting costs through AI without restructuring work to multiply its benefits.

Using productivity tools may give employees a productivity bump, but working in tandem with AI agents will help them deliver far greater gains. To realize these improvements, leaders can’t just hand off AI tools and hope for the best. Employees have to learn to manage these digital colleagues.

AI Agents Need Steering

AI agents shouldn’t be entirely autonomous. They need steering. To conceptualize this, think about a common activity like learning to drive. You don’t start off taking a long-distance trip or even getting behind the wheel.

You’re first a passenger, observing an experienced driver. You continue to learn the rules and technicalities through driver’s education. You get hands-on experience, starting small, like in parking lots and on residential streets, then highways, until you feel comfortable going longer distances. Eventually, driving becomes second nature.

Working with AI agents follows the same pattern. Just as you had to learn what makes a vehicle respond before you could teach someone else to drive, you must learn to operate agents before you can effectively build and manage them. The transition from a knowledge worker to an intellectual worker will require everyone to learn about and manage digital colleagues.

The New Diamond-Shaped Workforce

When you begin deploying agents, you’ll start chipping away at the corners of the workforce pyramid, turning it into more of an obelisk or diamond shape. Entry-level positions contract sharply because agents can handle tasks that previously required entry-level employees. Fewer entry-level jobs and more stable mid-level jobs, where you rely on people’s expertise to solve problems, create the diamond shape.

This means that even if you’re an individual contributor today, you’re going to be a manager of digital colleagues at some point. Similar to bringing on an intern, you will instruct them so they can be successful, and you’ll make it clear that if they run into a problem, they should seek you out rather than trying to solve it independently.

In this model, new roles will emerge, such as AI facilitators who bridge technology, organizational design and change management, and AI experience designers who make agent interactions seamless for end users.

A Four-Level Framework For AI Agent Adoption

Companies need a structured progression for their workforce, starting with foundational competencies everyone must develop. These are capability levels that determine what employees can access and do:

• AI Citizen Role: This is the baseline of basic awareness plus policy compliance. It grants access to productivity tools like ChatGPT Enterprise, Copilot or Gemini. Think of this as the written test before getting behind the wheel.
• Agent Driver Role: Once employees are comfortable with productivity tools, they have to learn to operate agents built by others. Like in our driving example, this is hands-on work in understanding how agents respond to inputs, what context they need and how to steer them effectively.

From this foundation, career paths diverge into specialized competencies:

• Technical Track: Building an agent is not the same as building an effective agent. To do the latter, leaders must transition from an agent builder to an agent architect by gaining knowledge of instructional design principles, information hierarchy, task specificity and context management.
• Leadership Track: Climbing the leadership ladder means thinking more expansively, such as where else to deploy agents to optimize business, how to redesign workflows or what use cases warrant engaging IT for complex builds.

Both tracks require mandatory training and clear decision authority matrices, and you can’t skip levels.

Putting It Into Practice

Start with foundational training for everyone. Before gaining access to productivity tools, every employee needs AI citizen competency. This isn’t just about compliance. It’s about building the baseline understanding that makes everything else possible.

From there, identify internal champions who show aptitude for the specialized tracks. Look for people who are already experimenting with AI tools — who understand the difference between using a chatbot and working with an agent. These are your future agent drivers and builders.

Next, create a governance structure that scales with competency levels. Gaining access to agent-building capabilities should require demonstrated proficiency at the driver level. You wouldn’t hand someone car keys without knowing they can drive, and the same logic applies here.

For leaders, overseeing people who manage digital workers is a new management challenge. If you’re in a management position, you need to understand the cognitive demands your teams face when managing digital colleagues.

The transformation is already in motion. Success will depend on whether you structure change deliberately to capture productivity gains that justify it, or you simply cut costs and wonder why the benefits never materialize.

Be intentional about what you’re building. There’s a difference between building a go-kart and building a Ferrari. Both have wheels and engines, but one is cobbled together from spare parts while the other is engineered for performance. Right now, most organizations are still figuring out which one they need and whether anyone on their team knows how to drive it.

This article was originally published on Forbes.com.

Are you ready to explore how artificial intelligence can fit into your business but aren’t sure where to start? Our AI experts can guide you through the entire process, from planning to implementation. Talk to an expert