LLM agents struggle to fully discover and deliver enterprise SQL model needs

Can LLM Agents Be Forward-Deployed Engineers? Requirement Discovery and Delivery for Enterprise Text-to-SQL

Computation and LanguageArtificial Intelligence

Summary

LLM coding agents can help automate tasks, but this study tested whether they can act like engineers who figure out exactly what customers need and make deployment decisions. The authors tested these agents on a real enterprise problem involving generating SQL queries from text, where agents had to first learn unclear customer requirements and then deliver a model accordingly. They found that even the best agents missed many key requirements and often failed to meet all acceptance criteria during deployment. The authors also found problems like incomplete requirement discovery, unreliable execution, and poor resource use, suggesting that humans should review the agents’ work and monitor their plans.

What this means in practice

  • For enterprise software developers: Integrate human reviews to verify LLM-discovered requirements when building text-to-SQL systems to improve deployment success rates.
  • For cloud operations teams: Use explicit experiment environment specifications including hardware and budget monitoring to better manage model post-training deployments guided by LLM agents.

Tested on simulated data.

Authors

Marianne Menglin Liu, Rongguang Wang, Zheng Wang, Yewen Wang, Leonid Boytsov, Tao Sheng, Sujith Ravi, Dan Roth

Abstract

LLM coding agents can increasingly automate model post-training, but whether they can act as forward-deployed engineer (FDE) agents, discovering customer requirements and turning them into deployment decisions, remains unclear. We study this on a real enterprise text-to-SQL engagement: a simulated customer, grounded in the real use case's requirements, starts with an ambiguous request. An agent must discover acceptance criteria, data permissions, and hardware constraints in the discovery phase, then plan and execute a post-training recipe in the delivery phase. Across 15 models, the strongest discover over 80% of withheld requirements and open-weight models discover 40-50%; in the delivery phase, only 25% of runs across eight coding-agent configurations pass all acceptance criteria. We identify four bottlenecks in this engagement: (1) incomplete discovery, with 40% of withheld requirements missing or misrepresented; (2) unreliable execution; (3) underused compute budgets and a lack of systematic comparisons across base models and post-training methods; and (4) benchmark accuracy does not guarantee production readiness. We suggest human review of discovered requirements, monitoring of the agent's plan during execution, and an explicit experiment-environment specification covering hardware, budget scope, and production checks.
View PDFOpen arXiv