According to The Register, Percona's CEO argues that no existing database is well suited to AI agents. The problem, as described, is that agent workloads are iterative: they involve many back-and-forth steps rather than the straightforward query patterns databases were designed for. That behavior apparently caught vendors off guard.
The Register notes that while the shortcomings are visible, the industry has not settled on what a purpose-built database for AI agents should actually do. The CEO reportedly says the requirements for a clean-slate solution remain unclear, making it hard for vendors to respond with a definitive product.
In short, the source presents a gap: there is broad recognition that current databases are a poor fit, but no agreed specification for the replacement. No other sources were provided, so there is no independent view to compare.