Published: May 12th, 2025 by Dr. Souvick Ghosh and Charles Costa

Customer‑service delivery remains one of the most intricate and cost‑intensive pillars of enterprise operations, owing to high employee turnover, continual training demands, and substantial investments in knowledge‑management infrastructures that sustain frontline agents. While many have pondered the effects of automation on these professionals, the meteoric rise of foundation models such as ChatGPT, Claude, and Gemini has propelled the discussion from speculative theory to operational reality.

Figure 1. Human-AI Collaborative Customer Service (generated using ChatGPT o3)

Drawing on our collaboration with a major global e‑commerce marketplace, we recently examined how enterprises weave generative AI into the twin cornerstones of modern customer support—chatbots and the Agent Desktop. The latter is the command center for every interaction: it surfaces a customer’s history, reveals technical anomalies, and aggregates internal notes, allowing agents to triage issues with efficiency and empathy. Yet even with these affordances, locating the precise policy, synthesizing it for the customer, and logging an accurate resolution can still feel like threading a needle.

Because our goal is to equip practitioners with a clear, research‑grounded map of this evolving terrain, we asked two guiding questions:

1. Which tactics enable multinational organizations to introduce generative AI at scale into their customer service operations without fracturing the service experience across regions and languages?
2. How might AI systems be architected so that they augment and fit the needs of human customer service agents?

Our findings, detailed in a published iConference paper (Costa & Ghosh, 2025) and also a journal article currently under review at Data & Policy, offer three takeaways:

1. Human–AI collaboration, not substitution. Chatbots perform admirably for repetitive “how‑to” queries, but complex, empathetic problem‑solving still requires a clear escalation path to human service agents. Generative models excel when they summarize prior conversations and provide relevant knowledge snippets for the human agent.
2. Human judgment for mission‑critical calls: Regulatory frameworks such as the Digital Services Act (The EU’s Digital Services Act, 2022) explicitly restrict fully automated decisions in high‑stakes contexts and while evaluating potential violations. Therefore, supervisory review still remains indispensable.
3. Smarter localization at lower cost: Generative AI dramatically reduces content localization costs. However, human expertise is essential to capture subtle linguistic and cultural nuances and domain‑specific jargon that AI may miss.

Looking ahead, generative AI is poised to shift customer‑service work from retrieval toward relationship management. By weaving large language models into existing agent desktop workflows and respecting regulatory guardrails, enterprises can cultivate symbiosis rather than substitution, delivering faster resolutions and deeper human connections.

Definitions:

Generative AI: Machine‑learning models that create new content (text, images, or code) by learning statistical patterns from vast corpora.

Foundation model: A large, pre‑trained network (e.g., ChatGPT, Claude, Gemini) that can be adapted to many downstream tasks.

Agent desktop: The unified interface through which service agents view customer history, internal notes, diagnostics, and AI‑generated suggestions.

Digital Services Act (DSA): EU regulation (in force for all platforms since Feb 2024) that limits fully automated, high‑risk decisions and mandates human oversight.

References:

Costa, C., & Ghosh, S. (2025). Empowering customer service with generative AI: enhancing agent performance while navigating challenges. Information Research, 30 (iConf (2025)), 150–158. https://doi.org/10.47989/ir30iConf47566

Costa, C. & Ghosh, S. (2025). AI and Automation: Harmonizing Machine Efficiency and Human Expertise in Crafting Customer Service Content. At Data & Policy. [Full Paper] [Submitted]

The EU’s Digital Services Act. (2022, October 27). https://commission.europa.eu/strategy-and-policy/priorities-2019-2024/europe-fit-digital-age/digital-services-act_en