NEWARK, DE — Artificial intelligence tools increasingly used by buyers to research software products are delivering incorrect pricing and feature information in nearly two-thirds of cases, according to a new study that raises fresh concerns about lost deals and distorted purchasing decisions.
Kodec AI said its analysis found that AI platforms returned inaccurate or misleading answers in 62 percent of simulated buyer queries involving B2B software products. The firm tested more than 200 query cycles across Series B and later-stage SaaS companies in the technology and financial services sectors.
The study examined how leading AI search and discovery platforms handled common buyer questions, including pricing details, feature comparisons, and product capabilities. In many cases, the platforms bypassed official company websites and documentation, instead relying on outdated third-party sources, reseller listings, or competitor-authored content.
Kodec labeled the phenomenon the “Rogue Sales Rep” problem, describing AI models that confidently present incorrect information without a verified source of truth. According to the report, the failures fell into three main categories.
In some cases, AI tools quoted discontinued free tiers or legacy pricing, setting buyer expectations at zero cost before sales teams ever engaged. In others, models conflated pricing by pulling figures from cloud marketplace resellers rather than official vendor pages. The study also found instances where AI systems attributed features to the wrong products after sourcing information from competitor-written “Best Alternatives” articles.
“These are revenue leaks, not minor glitches,” a Kodec AI spokesperson said, noting that a single misquoted enterprise price can derail a deal before it reaches a sales conversation.
The findings point to a broader technical gap as AI becomes more central to how buyers evaluate vendors. Kodec said many companies lack machine-readable, authoritative data structures that AI systems can reliably reference. Without them, models default to scraping unstructured web content, increasing the risk of errors.
The issue is becoming more acute as the web shifts toward what Kodec describes as an “agentic” phase, where AI tools act on behalf of users by comparing vendors, narrowing shortlists, or even initiating procurement steps. In that environment, incorrect or unverified information can exclude vendors entirely without human review.
To address the problem, Kodec AI said it offers search infrastructure designed to help enterprises audit how they appear across AI platforms and implement governed data structures that models can treat as authoritative sources.
As AI-driven research becomes a first stop for enterprise buyers, the study suggests that inaccurate machine-generated answers are no longer just a marketing concern, but a direct threat to revenue and competitive positioning.
For the latest news on everything happening in Chester County and the surrounding area, be sure to follow MyChesCo on Google News and MSN.