New AI Roles in the Workforce

The American workforce is reorganizing around AI faster than most curricular planning cycles can absorb. The U.S. Census Bureau Business Trends and Outlook Survey reports the share of firms using AI rising from 3.7 percent in September 2023 to 5.4 percent by February 2024, a 46 percent increase in five months. Lightcast finds that the average U.S. job has seen one-third of its skills change between 2021 and 2024, and that the top quartile of jobs has seen 75 percent of skills change in three years. Brynjolfsson, Chandar, and Chen (2025) report that early-career workers aged 22 to 25 in AI-exposed occupations have seen a 16 percent relative employment decline, with software developers in that age band down nearly 20 percent from their late 2022 peak. The labor market is moving. The question for higher education is whether curriculum is moving with it.

This paper is a reference document for that work. It catalogs the new job categories that have emerged across the enterprise AI economy, organized into three tiers: management roles that govern AI strategy, risk, and adoption; business roles that apply AI within functional domains such as finance, HR, customer success, and analytics; and technical roles that build, deploy, and operate the underlying systems. The taxonomy spans roughly forty roles in current calibration, including the Chief AI Officer, Head of Applied AI, AI Ethicist, AI Risk and Governance Specialist, AI Product Manager, AI Compliance Analyst, AI HR Business Partner, AI Architect, Model Validator, AI Redteam Engineer, Prompt Engineer, and a long list of others. For each role, the paper provides a description of the function, an embedded role footprint diagram on the Enterprise Data and AI Value Architecture, a capability profile, and the undergraduate and graduate coursework that would prepare a graduate to enter or develop into the role.

The architectural anchor matters. Throughout the paper, every role is mapped onto the same reference model, the Enterprise Data and AI Value Architecture, so that curriculum committees can see which stages of the value chain a role actually owns. The Chief AI Officer footprint covers governance and oversight across the top band. The AI Ethicist footprint clusters around use case approval, risk classification, and human-in-the-loop. The Model Validator footprint covers evaluation and output validation. The AI Customer Success Manager footprint covers workflow integration and adoption. Mapping the roles to a shared architecture forces a discipline that role-by-role descriptions usually skip: it makes the gaps visible, it surfaces overlaps and confusions between role titles that sound similar, and it gives curriculum designers a basis for deciding which existing courses cover which footprint and where the curriculum needs to extend.

The pedagogical implication for each tier is different. Management AI roles need an integrated AI strategy and governance track in MBA and executive education programs. Every MBA graduating after 2027 should be able to read the Enterprise Data and AI Value Architecture, articulate the difference between BI dashboards, predictive ML models, and generative AI, evaluate vendor claims, and make capital allocation decisions across an AI portfolio with risk and value framings. Business AI roles need AI integrated into every functional concentration in undergraduate business programs, not added as a separate degree. A finance program that treats AI as a single elective is no longer responsive when AI tools are now embedded in financial modeling, audit, fraud detection, and reporting. Technical AI roles need computer science programs that take a deliberate position on AI specialization, preserving the foundational curriculum while building substantive AI tracks that go deeper than a single course can carry.

Three observations cut across the tiers. First, AI fluency cannot live only in computer science departments. McKinsey Global Institute (2025) data show that 80 percent of skills used in the workplace today are still needed for activities that cannot be automated, and 57 percent of hours could in theory be performed by AI agents. The interface between human judgment and AI capability is where the value is created and where most curriculum is currently silent. Second, undergraduate education needs to do two things at once. It needs to preserve the foundational disciplines that produce durable thinking, including mathematics, statistics, writing, ethics, systems thinking, and organizational behavior. It also needs to layer AI-specific coursework on top so that a graduate enters the workforce able to work with AI rather than around it. Third, graduate programs need to differentiate. An MBA with an AI concentration, a Master of Science in Business Analytics, a Master of Science in Data Science with a generative AI track, a Master of Science in AI Engineering, and a Master of Public Affairs with an AI Policy concentration are not interchangeable. They serve different roles in the taxonomy, and the curriculum decisions need to be made with that distinction in view.

The roles described in the paper are advisory examples. They are not job descriptions to be copied verbatim. They are calibration points to help departments and curriculum committees see what employer demand looks like and where existing courses come closest to meeting it. The expectation is not that every institution will offer every track. The expectation is that curriculum decisions will be made with line of sight into the actual taxonomy of roles the labor market is hiring against, and with a defensible answer to the question of which roles the institution is preparing graduates to enter and which it is consciously choosing not to address.