AI Agents Reshape Data Work as SQL Gains Reliability in 6 to 9 Months

3 articles · Updated · InfoWorld · Aug 6

Data teams are being urged to prepare now for AI agents, which are starting to become practical for analytics and pipeline operations after LLMs recently got reliable at writing SQL.
Mechanical data-engineering work is the clearest near-term use case: agents can generate tests from code assumptions, detect schema or API changes, patch simple breakages, and hand humans diagnoses instead of raw failures.
Context remains the main constraint because agents can infer joins, common values and query patterns, but they cannot decide business definitions such as revenue, customer status or fiscal-year boundaries.
Proactive insight generation is still immature, with false positives likely to erode trust, so teams should first build context, automate evals and use infrastructure that can absorb bursty, parallel agent queries.
Latency and isolation will matter more as agents fan out across many database calls; a 10-millisecond engine versus 100 milliseconds can compound into far higher throughput for agent-driven workloads.