As AI Takes On More Work, Which Skills Matter Most?


AI is becoming increasingly capable of the analysis, modeling, and problem-solving that have long defined knowledge work. For businesses, that raises a challenge beyond deciding which tasks to automate: how to develop people who can direct, evaluate, and take responsibility for work that AI produces. 

That was a central theme of the recent AI Horizons webinar hosted by Kartik Hosanagar, faculty director of Wharton Human-AI Research. He was joined by Michael Roberts, Wharton’s vice dean of the MBA program and professor of finance, and Chris Callison-Burch, the Raj and Neera Singh Professor of Artificial Intelligence in the Department of Computer and Information Science.

Their conversation explored what AI’s growing capabilities mean for workforce development, entry-level careers, and the skills businesses will need next. 

A Strong First Draft Still Needs Judgment 

Hosanagar opened with research he coauthored with Callison-Burch and others, examining AI’s performance on 615 open-ended questions drawn from 238 business school cases across 18 disciplines. Unlike tests of factual recall, these questions required reasoning through ambiguous business problems. 

The models scored roughly 87% when awarded partial credit. When answers had to meet the full requirements to receive credit, however, even the best models fell short on about half the questions. 

The distinction matters for businesses. A polished recommendation can capture much of what is needed while still missing something consequential. Producing a persuasive answer and producing an answer ready to act on are different standards. Roberts also emphasized a limitation: business cases supply carefully curated context. In an actual organization, identifying the relevant information is often part of the work itself. 

For managers, the findings suggest that evaluating AI requires looking beyond how convincing an answer sounds to whether it addresses the full problem. 

Knowing the Business Becomes More Valuable 

As AI takes on more production work, Hosanagar suggested that human effort shifts in two directions: upstream, toward deciding which questions to ask, and downstream, toward checking the answers. 

Both depend on knowledge of the business. An AI system may generate a technically sound analysis without understanding the relationships, competing priorities, or circumstances that determine whether a recommendation will work. 

“I think our superpower in some sense is our own knowledge and context that’s not in the model,” Roberts said. 

That changes what proficiency looks like. Employees need to explain the problem clearly, provide relevant context, challenge assumptions, and judge whether the result fits the situation. Familiarity with AI tools alone is insufficient. The implication for hiring and training is to develop business expertise alongside AI proficiency. The most useful employees will connect what the technology can produce with what the organization actually needs. 

Automating Junior Work Creates a Training Challenge 

Many experienced professionals learned to exercise judgment by doing the underlying work repeatedly. Building financial models, writing code, and revising analyses helped them recognize mistakes and eventually supervise others. 

AI complicates that progression. If new employees immediately begin reviewing work generated by a system, how do they develop the expertise to assess it? Callison-Burch compared the problem to reading an answer at the back of a textbook. An explanation may be easy to follow without giving the reader the understanding that comes from working through the problem independently. 

The speakers described changes in education, including assessments that require students to orally explain their reasoning, evaluate AI output, or demonstrate understanding without relying on the tool. 

Businesses face a parallel challenge. As routine assignments are automated, employers will need to be more deliberate about how junior employees practice, receive feedback, and build judgment. Faster output does not, by itself, mean that learning is taking place. 

Productivity Gains Create Choices About Growth 

Callison-Burch offered business leaders a thought experiment: if AI made every employee twice as productive, would the organization reduce its workforce or find ways to grow? The scenario was hypothetical, but the management choice is immediate. Leaders can consider how additional capacity might support new products, better services, or work that previously lacked resources. 

They also need to consider where people entering the workforce will fit. Removing junior tasks without creating new development opportunities risks weakening the pipeline of future experts and managers. Roberts noted that the change is about the types of jobs as well as their number. Some familiar entry-level roles may give way to different ones. 

The discussion also surfaced emerging possibilities. Hosanagar described hiring for a go-to-market engineering role that combines commercial knowledge with the ability to build systems using AI. Callison-Burch described MBA students creating prototypes without a software engineering background. 

Adaptability Requires More Than Learning a Tool 

Roberts identified a gap between executives who know their businesses need to change and younger employees who are comfortable with AI but may lack the business knowledge to guide that change. Closing that gap requires people who can combine domain expertise, technical proficiency, and the ability to rethink workflows. It also requires communication and influence: colleagues need to understand and support changes in how work gets done. 

Asked which skills would remain valuable, Callison-Burch emphasized teamwork, clearly defining a project, and adapting to feedback. Roberts pointed to continuous learning and the ability to avoid “change fatigue.” Hosanagar highlighted asking the right question. 

For businesses, the challenge is to build those capabilities while adopting the technology. As AI handles more of the work, organizations still need people who can define the objective, recognize what is missing, and help others move forward. 

We will be happy to hear your thoughts

Leave a reply

Som2ny Network
Logo
Register New Account
Compare items
  • Total (0)
Compare
0
Shopping cart