DBAX 8006 Week 11 Integrated Final Paper Example

Will Generative AI Assistants Replace Customer-Support Staff? An Integrated Final Paper Whose Conclusion Claims Nothing Its Own Evidence Sections Left Unestablished

Student Name

Walden University

DBAX 8006: Contemporary Challenges in Business

Instructor Name

Month Day, Year

Will Generative AI Assistants Replace Customer-Support Staff? An Integrated Final Paper Whose Conclusion Claims Nothing Its Own Evidence Sections Left Unestablished

Abstract

Coverage of generative AI often states or implies that AI assistants will replace customer-support staff. This paper examines what peer-reviewed evidence shows. The best field evidence, from one large firm's support operation, shows an AI assistant lifting the hourly count of resolved customer issues by roughly a seventh overall, concentrated among agents who were new or less skilled. Experimental evidence on writing tasks finds similar productivity gains and a narrowing of differences among workers. Exposure studies estimate that many occupations, including customer service, have tasks that such tools could affect. None of this evidence measures headcount changes at adopting firms. The paper concludes that current evidence supports a claim that AI assistants raise support agents' productivity, especially for newer agents, but does not yet support a claim that they replace support staff.

The Problem

Customer support is a natural candidate for automation by generative AI because much of the work involves reading a customer's message and writing a response. Firms are adopting AI assistants that suggest replies to human agents, and some are using chatbots that respond to customers directly. Coverage moves easily from "AI can do parts of this job" to "AI will do this job," and from productivity gains to job losses. This paper asks what the peer-reviewed evidence establishes and where its reach ends.

Evidence on Productivity

The strongest field evidence comes from Brynjolfsson et al. (2025), who studied the staggered rollout of a generative AI assistant to about 5,000 customer-support agents at one firm. Agents with access to the assistant resolved about 14 percent more issues per hour on average. Gains were much larger for novice and lower-skilled agents and small for the most experienced, suggesting that the tool spread the practices of better agents to newer ones. The authors also found improvements in customer sentiment and lower agent attrition.

In a controlled study of professional writing tasks, Noy and Zhang (2023) found that participants given ChatGPT finished faster and produced better work, and that the weakest writers gained the most. Their tasks were not customer support, but they resemble the writing at the heart of support work.

This section supports one claim: generative AI assistants can raise the productivity of support agents, particularly less experienced ones.

Evidence on Exposure

Eloundou et al. (2024) estimated, by rating the tasks in each occupation, that a large share of U.S. workers hold jobs in which a meaningful fraction of tasks could be affected by large language models. Customer service ranks among the more exposed occupations. Exposure, however, means that tasks could be performed faster or differently with the technology; it does not measure whether jobs are lost, since faster work can lead firms to handle more volume, improve service or cut staff.

This section supports a second claim: customer-support work contains many tasks that generative AI can affect.

The Opposing Case

The strongest case for replacement runs as follows. If each agent resolves 14 percent more issues per hour, a firm with fixed demand needs about 12 percent fewer agents to handle the same volume. As chatbots improve, more customer contacts may be resolved without a human at all. Firms facing cost pressure will take the savings. This reasoning is sound as arithmetic and may prove correct. But it rests on an assumption the evidence does not test: that demand for support stays fixed and that firms convert productivity gains into lower headcount rather than into better service, faster response times or expansion. Brynjolfsson et al. (2025) did not report headcount changes at the firm, and their finding of lower attrition suggests at least one route by which staffing could change without layoffs, through slower replacement of departing agents.

Costs and Distribution

Who gains and who bears costs depends on what firms do with the productivity gain. The field evidence suggests newer agents gain most in skill, which could shorten training and change how firms value experience. If firms reduce hiring, the cost falls on future applicants rather than current employees. These effects are plausible, but none has yet been measured.

Comparison With Earlier Automation

The term's earlier work on industrial robots found that firms adopting robots tended to grow employment relative to non-adopters, while their local markets lost jobs (Koch et al., 2021). Whether generative AI follows that pattern in services is unknown; the comparison suggests that firm-level and market-level effects may differ and should not be assumed to match.

What Would Change This Conclusion

Three kinds of evidence would change the conclusion. A firm-level study tracking support headcount before and after AI adoption, with a comparison group, would test the replacement claim directly. Evidence that customer demand for support rises when service gets faster would weaken the arithmetic of replacement, while evidence that it stays flat would strengthen it. And evidence that chatbots now resolve a large and growing share of contacts without any human involvement would shift the question from assisting agents to substituting for them. None of these is available in the peer-reviewed literature reviewed here, and the paper's claim should be revisited as such studies appear.

Conclusion

Each sentence of this conclusion rests on a section above. Generative AI assistants have been shown to raise customer-support agents' productivity, by about 14 percent on average at one large firm and by more for newer agents (Evidence on Productivity). Customer-support work contains many tasks such tools can affect (Evidence on Exposure). Whether firms will use these gains to reduce support staff has not been measured, and the replacement argument depends on assumptions about demand and firm choices that the evidence does not test (The Opposing Case). Effects on firms and on labor markets may differ, as they did with industrial robots (Comparison With Earlier Automation). The evidence therefore supports a claim about productivity, not yet a claim about replacement.

References

Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. Quarterly Journal of Economics, 140(2), 889-942. https://doi.org/10.1093/qje/qjae044

Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2024). GPTs are GPTs: Labor market impact potential of LLMs. Science, 384(6702), 1306-1308. https://doi.org/10.1126/science.adj0998

Koch, M., Manuylov, I., & Smolka, M. (2021). Robots and firms. Economic Journal, 131(638), 2553-2584. https://doi.org/10.1093/ej/ueab009

Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187-192. https://doi.org/10.1126/science.adh2586