Model validation of a generative-artificial-intelligence-based avatar for customer support in banking

Jochen Gerhard, Martin Gombert, Björn Henrich and James Smith

Journal of Operational Risk

Abstract: This study presents an innovative validation approach for a generative-artificial-intelligence-based avatar named Ava, designed to handle customer inquiries in the banking sector. Classical model validation frameworks fall short when applied to generative artificial intelligence models due to the opaque, black-box nature of such models. This paper introduces a systematic framework of guardrails that emphasize trustworthiness, including rigorous testing, real-time monitoring, scenario assessment and effective governance. Our validation approach ensures Ava adheres to principles of human oversight, fairness, transparency and reliability, which are vital for safe use in financial services. Using Commerzbank’s Ava as a case study, the framework provides insights into achieving trustworthy artificial intelligence in highly regulated sectors, balancing innovation and compliance.

References: Add references at CitEc
Citations:

Downloads: (external link)
https://www.risk.net/node/7963290 (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:rsk:journ3:7963290

Access Statistics for this article

More articles in Journal of Operational Risk from Journal of Operational Risk
Bibliographic data for series maintained by Thomas Paine ().