Consultant warns legacy bank cores may limit AI and tokenization

Large banks may be spending 70% to 80% of their technology budgets on maintaining existing systems, leaving limited capacity for new products, artificial intelligence and tokenization, according to Dharmesh Mistry, chief executive of Vision 20 20 Consulting. In a Finextra community opinion post, Mistry argued that decades of incremental core banking changes have left many institutions with complex estates that are costly to operate and difficult to modernise.

Mistry described the industry’s accumulated infrastructure as a “Franken-core”, a term he used for the combination of old ledgers, product-specific cores, branch systems, digital channels, compliance tools, fraud platforms, customer relationship management software and API layers that have been joined over time.

He said those systems were usually added for rational reasons, including new channels, regulation and product demands. The cumulative result, in his view, is an operating model in which bespoke integrations connect multiple systems, and institutional knowledge about those links often sits with a small number of employees rather than formal documentation.

That creates organisational risk as well as technical debt, Mistry wrote. Staff who understand how one system connects to another can become critical dependencies when failures occur, and the loss of those employees through retirement, departure or reorganisation can remove knowledge that is not visible in architecture diagrams.

Maintenance burden and AI limits

Mistry cited industry estimates that maintenance of existing bank technology estates can consume between 70% and 80% of IT spending at large banks. On that basis, he said, only a minority of technology budgets remains available for board-level priorities such as product development, new channels, AI, tokenisation and competitive differentiation.

He argued that fragmented data architectures could also limit the usefulness of AI agents and large language models in banking. AI tools rely on access to consistent and reliable data if they are to support autonomous or semi-autonomous decisions. Where customer, account and transaction information is spread across multiple systems and connected through legacy integration code, Mistry said, banks risk deploying AI on top of an environment that cannot provide a single operational view.

The issue is not confined to banks, according to Mistry. He said some vendors selling newer core banking platforms may be recreating similar complexity as they add cloud deployment options, AI modules and tokenisation functionality to systems originally built around more traditional account and ledger structures.

Tokenisation changes the architecture question

Mistry said tokenised deposits and digital assets create requirements that differ from conventional deposit accounting. In his description, a tokenised deposit can carry programmed conditions, settle on a distributed ledger and interact with smart contracts. That implies different data models, settlement processes and interpretations of balances than those used by many traditional cores.

He argued that adding tokenisation features to platforms designed around conventional accounts may produce another layer of complexity rather than a clean replacement for legacy systems.

Mistry also pointed to stablecoin issuers such as Circle and Tether as examples of competitors that do not start from traditional bank core infrastructure. He said large technology platforms could also become relevant if regulatory developments in the US, UK and EU continue in that direction, though he framed that as a potential outcome rather than a current fact.

The post concluded that banks may have to replace, rather than continue to extend, their core banking infrastructure as changes in payment rails, AI and tokenisation place new demands on underlying systems. Finextra labels the post as external author content expressing Mistry’s views.

This story draws on original reporting from Finextra Research.