AI reaches 82% accuracy in database management but needs human oversight
Updated
Updated · InfoWorld · Apr 30
AI reaches 82% accuracy in database management but needs human oversight
8 articles · Updated · InfoWorld · Apr 30
BIRD benchmark data puts top text-to-SQL models at nearly 82% execution accuracy, versus about 93% for human database engineers.
The report says AI already speeds routine SQL writing, optimisation and support tasks, but still struggles with complex "last mile" database problems without human review and refinement.
Rising customer expectations for faster fixes and performance are pushing internal teams and service providers to adopt AI, while full automation remains too risky for critical database systems.
With AI agents accidentally deleting millions of records, is the push for database automation introducing unacceptable risks?
As AI handles routine tasks, what new strategic skills will define the indispensable human database expert of tomorrow?
How will the new White House AI policy change corporate liability when AI agents inevitably cause system failures?