{"id":11173,"date":"2026-07-31T15:35:25","date_gmt":"2026-07-31T06:35:25","guid":{"rendered":"https:\/\/blog.themusio.com\/?p=11173"},"modified":"2026-07-31T15:35:27","modified_gmt":"2026-07-31T06:35:27","slug":"how-one-software-company-stopped-llm-hallucinations-in-sales-training","status":"publish","type":"post","link":"https:\/\/blog.themusio.com\/?p=11173","title":{"rendered":"How One Software Company Stopped LLM Hallucinations in Sales Training"},"content":{"rendered":"\n
The VP of Sales had a problem that shouldn’t have existed.<\/p>\n\n\n\n
His reps were acing AI roleplay training. Empathy scores were high. Objection handling was textbook. The platform’s feedback was consistently positive \u2014 strong rapport, excellent pacing, clear communication.<\/p>\n\n\n\n
And yet, lost-deal reviews kept surfacing the same failure mode. A rep would confidently state a product capability that didn’t exist. An “unlimited” tier that was actually capped. An integration that was still in beta. A pricing condition that had changed three months ago.<\/p>\n\n\n\n
The reps sounded excellent. The product information was wrong.<\/p>\n\n\n\n
The training platform had no way of knowing \u2014 because it was checking for plausibility, not truth.<\/p>\n\n\n\n
This is the story of how one B2B software company \u2014 we’ll call them Vertex Cloud \u2014 fixed that gap.<\/p>\n\n\n\n
The problem: plausible-sounding wrong answers<\/strong><\/h2>\n\n\n\n
Vertex Cloud sells a B2B SaaS platform to mid-market and enterprise customers. Their product is complex: multiple tiers, dozens of features, pricing that varies by seat count and contract length, and integrations with third-party tools that change quarterly.<\/p>\n\n\n\n
Their AI training platform could generate practice conversations from uploaded documents in seconds. Reps practiced with AI customers, got scored on empathy and objection handling, and could repeat scenarios as many times as they wanted.<\/p>\n\n\n\n
But the platform had a fundamental blind spot: it only checked whether a statement sounded<\/em> reasonable.<\/p>\n\n\n\n
When a rep practiced a call and confidently stated that the Enterprise plan included “unlimited API calls” \u2014 when the pricing guide clearly said 10 million per month \u2014 the platform gave full marks. The content of the claim was never evaluated.<\/p>\n\n\n\n
The system was grading confidence, not correctness. And because the platform rewarded fluency, reps were being trained to sound more certain about information that wasn’t always true.<\/p>\n\n\n\n
The cost of plausible wrong answers<\/strong><\/h2>\n\n\n\n
Vertex Cloud tracked the impact for six months before the intervention:<\/p>\n\n\n\n
\n
14%<\/strong> of lost deals involved buyers who discovered inaccurate product information during the evaluation<\/li>\n\n\n\n
9%<\/strong> of support escalations came from customers who were sold capabilities that didn’t exist<\/li>\n\n\n\n
New reps took an average of 10 weeks<\/strong> to reach certification \u2014 and many still failed internal product knowledge assessments after certifying<\/li>\n<\/ul>\n\n\n\n
The VP of Sales put it simply:<\/p>\n\n\n\n
“Our reps were getting better at sounding confident. But they weren’t getting better at being correct. The platform couldn’t tell the difference, and neither could we.”<\/p>\n\n\n\n
The root cause was structural. The training platform’s underlying LLM had never read the company’s actual pricing guide. It only knew what “unlimited API calls” sounded like based on patterns learned from other companies’ products. When a rep said it, the model registered: that sounds plausible<\/em> \u2014 and moved on.<\/p>\n\n\n\n
Plausible isn’t the same as true. But to a general-purpose LLM, they look identical.<\/p>\n\n\n\n
The intervention: replacing plausibility with verification<\/strong><\/h2>\n\n\n\n