This article originally appeared on Inc Magazine.

Investing in advanced AI capabilities and complementary infrastructure can steer labor market effects in a positive direction.

Generative AI is affecting the labor market in vast and profound ways and as changes quickly unfold, the challenge is in how to evaluate these effects and hopefully steer them in a positive direction.

Not all AI is created equal–the kind and caliber of systems that businesses adopt will strongly influence labor force outcomes.

Recent technological change has been biased toward automation, with insufficient focus on creating new tasks where labor can be productively employed. In other words, AI systems are primarily replacing some of the tasks previously performed by human workers, pushing down the labor demand, and people lose their jobs. 

Displacement is without question a negative impact of AI on the workforce. Organizations may see short-term gains from reducing headcount and speeding up the completion of some tasks, but if workers lose value and income, the whole economy suffers. On the other hand, if organizations invest in and adopt AI systems that augment human tasks and create completely new ones by generating valuable predictions and insights, the result is increased labor value and demand.

Here’s an example. A human resources professional could use AI to automate the résumé-screening process and generate a narrowed-down list of the most promising candidates. If that were the employee’s primary task, the machine could take their job. 

However, the time they save by no longer scanning resumes and filtering candidates can be used for more personalized communication with candidates and improving the candidate experience, which helps organizations nab the best talent for greater success. At the same time, the AI system could help the recruiter upskill by identifying trends and patterns from the recruitment data would be beneficial for the firm. They could then focus on analyzing these insights to strategize future hiring needs, workforce diversity, and talent management. AI can replace one tedious task, but it also complements and even creates others that it couldn’t do alone. This is something that current AI-driven systems can do, and businesses are seeing the benefits. But take this a few years into the future, and AI systems can be creating entirely new industries and roles.

Right now there’s a real risk that firms will primarily adopt middling AI systems that are productive enough to displace workers, but not productive enough to increase labor demand through other channels. 

AI’s intangibles

It’s easy to imagine how machines can speed up existing tasks–it’s much harder to conceive of ideas and functions that don’t exist yet. This goes for AI firms and the organizations that buy their products. Ambitious technologists and forward-thinking business leaders are needed to build advanced AI systems and invest in possibilities beyond their current frame of reference.

The predictive and creative capabilities of AI that can have a positive impact on the labor market are difficult to explain and understand. To truly increase productivity and the value of labor, AI needs to be accompanied by complementary investments in information technology infrastructure, skills development, and business processes. These investments may involve the accumulation of intangibles such as data, information, and knowledge that are critical inputs for the AI model. Intangibles are hard to protect, imitate and value, and their creation often involves lengthy and uncertain processes of experimentation and learning-by-doing.

But there are tangible reasons to make these complementary investments. For example, if there’s no investment in upskilling, reskilling, and evolving transferable skills, there will be a big risk of labor polarization, in which there are both labor shortages and unemployment.

As technology evolves, it will require new and different skills to implement and use effectively, creating new roles and organizational structures. If there aren’t enough people with those skills, jobs will go unfilled and AI’s impact on productivity and growth will be limited. At the same time, growth will slow in jobs that don’t require higher skills or can’t be replaced with AI, and too many workers who need them.For many business leaders who are eager or curious about adopting AI, the process can feel like stepping into a pitch-black room. But the solution to this isn’t to start slow with simpler automation systems–it’s to embrace the experimentation, looking beyond short-term business milestones to transformative opportunities.