In-house AI builds face rising maintenance costs

By Milan Rojan

A recent study has warned that regulated financial institutions could face significant long-term costs from building and maintaining artificial intelligence (AI) and regulatory technology infrastructure internally.

The warning has been outlined in new research which has examined the changing economics of developing software as generative AI tools have accelerated prototyping and development.

The research has argued that while tools such as GitHub Copilot and Claude Code have reduced the time required to develop regulatory intelligence prototypes, the costs associated with running, securing and updating those systems have remained significant.

The research has highlighted architecture management, security requirements and regulatory changes as ongoing challenges for firms operating in highly regulated markets. The analysis has noted that up to 80% of software costs have occurred after launch, increasing the potential maintenance burden of internally developed systems.

Neil Wands, Chief Operating Officer of FinregE, has said: “The strategic question for the board is no longer whether a capability can be built in-house. The real question is whether owning the lifelong maintenance of a utility is the most efficient use of a firm’s most expensive resource: its human capital.”

The research has also pointed to the pace of regulatory change, with firms having to monitor more than 200 changes a day across multiple jurisdictions, including developments linked to the EU AI Act and Digital Operational Resilience Act (DORA).

The company has said its Regulatory Operating System (ROS) has been designed to support this model through global regulatory horizon scanning, digital rulebooks and its AI Regulatory Insights Generator (AI RIG).

Wands has added that institutions that have distinguished between proprietary capabilities and essential infrastructure could have been better positioned to allocate engineering resources towards differentiated products and customer experiences.

The findings have highlighted a broader shift in enterprise AI strategies, as financial institutions have increasingly evaluated the total cost and governance requirements of owning AI infrastructure beyond initial development.