Reviews

I recognize I may not be addressing the question directly, but even if large language models (LLMs) like GitHub Copilot are not yet at the level required for complete autonomy, isn't it reasonable to anticipate that, over time, their ability to learn patterns from our codebase will evolve to a point where human intervention becomes significantly less frequent? With this in mind, I believe the focus should be on preparing for that future—gradually adopting copilots and leveraging empirical data to guide their implementation and integration into workflows.

From discussion with peers in banking and financial services, the 'taker' model is popular to 'dip their toes' into the Gen AI solutions. Often run as pilots or from existing partnerships with say Microsoft, this provides a low risk entry point into Gen AI.