A new survey shows that 57% of enterprises have experienced AI agents delivering wrong answers with complete confidence in the past six months, with the root cause traced to missing or inconsistent business context. The problem stems from how most enterprises choose their retrieval systems — prioritizing ease of use and simplicity over accuracy — which means flaws only become apparent after the system is already in production.
Summaries like this, in your inbox every morning.
Sign up free →What happened
A VB Pulse survey of 101 enterprises with more than 100 employees found that 57% traced a confident but wrong AI agent answer to missing or inconsistent business context in the past six months, and 31% said it happened more than once.
Why it matters
Most enterprises rely on document retrieval as their primary way to feed business context to AI agents — 38% use this as their default approach — yet they choose retrieval systems based on ease of ingestion and operational simplicity rather than retrieval accuracy. This mismatch means accuracy problems often only surface after the system is already live and causing costly mistakes.
What to watch
The survey reveals a structural gap in how enterprises evaluate AI agent infrastructure. Prioritizing operational ease over accuracy in the selection phase appears to leave enterprises vulnerable to the kind of confidently-delivered misinformation that damages trust in AI deployment.
The survey captures a critical gap in how enterprises deploy AI agents: while the underlying language models may be performing correctly, the context layer — the business information fed to those models — is often broken or incomplete. This disconnect emerges from a process problem. Enterprises select retrieval systems that are easy to set up and operate, not systems optimized for accuracy. Document retrieval has become the default context mechanism for a large plurality of enterprises (38%), yet it is evaluated on the wrong criteria at purchase time, meaning issues only surface in production when an agent confidently delivers a wrong answer based on stale metrics or documents that were never retrieved.
The fact that 31% of affected enterprises report the problem happening more than once suggests this is not a one-off edge case but a recurring operational reality. The survey points to a known fix, though the article body does not complete the explanation of what that fix entails.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
