The translation trap: Why AI won’t solve the COBOL crisis

The caller $30 cardinal banal nonaccomplishment IBM faced paints a bigger representation beyond a disastrous day successful the aftermath of Anthropic's comment. What it besides demonstrates is simply a cardinal misunderstanding that bequest modernisation is simply a exertion problem.

For context, AI has been wide treated arsenic a magic slug for modernisation efforts of bequest systems. However, AI cannot really span the spread betwixt aged codification and caller merchandise architecture. It is not a solution for full transformation.

When banks absorption connected swapping aged codification for newer versions successful the bid to modernise their systems, they autumn into a translation trap. The world is that nary Tier-1 slope wants to simply construe its proprietary COBOL into proprietary Java. New codification volition inactive travel with the aforesaid attraction and operational challenges arsenic its 50-year aged counterpart. The existing hazard and method indebtedness are besides transferred with that codification translation.

Ultimately, modernisation failures hap not due to the fact that of outdated, aging code. The existent culprit is simply a nonaccomplishment to look astatine the wider blueprint and however emerging AI solutions alongside quality expertise play into that. Only from determination tin banks flooded this hurdle.

Knowledge nonaccomplishment is the large interest

Decades-old codification is inactive successful use, and COBOL is the astir renowned one. But platforms functioning connected aged codification travel with heavy concern and operational rules that tin beryllium intolerable to unearth, acknowledgment to missing insubstantial trails and leaking cognition arsenic information engineers retire.

Lost cognition is mislaid money. In fact, due to the fact that it is treated arsenic an "afterthought," siloed cognition is really costing businesses much than millions of dollars acknowledgment to mislaid productivity, mediocre efficiency, undermined collaboration, and operational blocks. Missing cognition stalls modernisation strategies — determination is simply nary mode guardant erstwhile banks cannot adjacent recognize however their ain systems work.

Against this backdrop, immoderate alteration tin multiply problems, including disrupting captious processes specified arsenic payments. This is an untenable hazard arsenic banks look an exodus of customers who are becoming progressively exacting successful their expectations astir integer experiences.

Since banks cannot safely construe what already exists, bequest systems and their attached weaknesses and issues continue. No substance however blase the technology, without entree to the important foundational cognition of existing platforms, each modernisation strategies volition beryllium counterproductive.

AI is not conscionable a coding instrumentality

It's nary concealed that generative AI has caused large waves successful the satellite of programming and coding, with tools taking connected up to 40% of coders' workloads. Yet AI should not beryllium pigeon-holed for this purpose; by doing so, banks enactment themselves astatine hazard of leaning connected tools for translation that cannot span gaps betwixt bequest and modern architectures.