Accenture and Wharton students tackle the real cost of running AI at scale

A question many companies cannot yet answer

Most public discussion of AI centers on upside: more revenue, greater efficiency and innovation. According to the Wharton AI & Analytics Initiative (WAIAI), a key piece is still unclear for many companies: what AI costs at scale. As initiatives leave the pilot phase, many organizations are beginning to see AI as a significant and growing cost center, and the economics of compute and data are becoming a topic in boardrooms and finance teams.

That was the question Accenture brought to the Wharton AI & Analytics Accelerator, an experiential learning program that pairs students with real companies to solve business challenges using data and advanced analytics.

Selen Karaca-Griffin, Americas Research Lead at Accenture, described the gap as obvious and urgent. "Everyone is talking about the upside of AI," she said. "What's missing is a rigorous accounting of what actually hits the expense line." She added that implementation, human oversight, compute and data costs are showing up in budget conversations, and that "most companies don't have a framework for thinking about them."

What the student team examined

Accenture asked the students to study the full economics of implementing and operating AI systems. The project focused on quantifying two things: the initial implementation costs and the ongoing operational costs of AI at scale.

The stated goal was to help organizations understand trade-offs, plan investments and make better-informed decisions about where and how to deploy AI. WAIAI says this reflects a wider shift in how companies approach AI in operations, from experimentation to execution, and from excitement to accountability.

Weekly collaboration instead of a handoff

The Accelerator is built around close collaboration between students and industry partners, who work as one problem-solving team. Accenture did not hand over the problem and return after eight weeks, according to the report. Karaca-Griffin and her colleagues met weekly with the student team, shared client perspectives and refined the problem as the work went on.

"We had a continuous dialog to share knowledge and address challenges as they appeared," she said. "Those sessions felt more like working sessions than check-ins." WAIAI says this kept the work grounded in questions Accenture teams hear from executives across industries.

Karaca-Griffin also praised the students' fresh perspective. In her words, they "ask sharp questions, push you to articulate your thinking more precisely, and bring a different rigor."

From classroom to client conversations

By the end of the eight weeks, the team had produced work Accenture could build on in practice. "We gave them a very complex problem, and they did really well," Karaca-Griffin said. She noted that Accenture will validate and build on the results, but that the work is already something "we are discussing with clients."

WAIAI stresses that Accelerator students do not work with simulated data or hypothetical scenarios, but develop solutions with direct real-world applications.

Accenture describes its participation as part of a long-term partnership with Wharton. "This wasn't a one-off experiment," Karaca-Griffin said. The project has laid groundwork for future tools, including the potential development of a scalable AI cost calculator that could help organizations repeatedly assess and plan AI investments.

What it means for companies

The source does not publish the students' findings or any cost figures. What it does show is that a major consulting firm sees a lack of cost frameworks as a real problem for clients. If the proposed calculator is built, it would give organizations a repeatable way to estimate and plan AI spending. For now, it remains a potential development, not an announced product.