Stop Saying This in Sales: The Phrases AI Flagged as Deal-Killers

Somewhere between "just checking in" and "to be honest with you," your sales rep may have just lost a six-figure deal. That's the central finding of a sweeping new analysis from three leading conversation intelligence platforms—Gong, Chorus, and Momentum—which collectively processed more than 2 million recorded sales calls from the past 18 months and used large language models to isolate the phrases most strongly correlated with deal failure.

The results, shared exclusively with JWXperience, paint a surprisingly specific picture. It's not just filler words or obvious blunders that sink deals. In many cases, the phrases that correlate with lost revenue are ones that salespeople have been taught to use—or at least never told to stop using.

At the top of the list: "To be honest with you." Across all three datasets, this phrase appeared in losing calls at a rate 47% higher than in winning ones. Linguists who reviewed the findings say the phrase triggers what they call an "honesty paradox"—by signaling that you're about to be honest, you imply that everything before it may not have been. Buyers pick up on this subconsciously, and trust erodes.

Close behind is "Does that make sense?"—a phrase so common in sales training that many reps use it reflexively at the end of every explanation. The data shows it correlates with a 31% higher likelihood of stalled deals, primarily because it shifts the burden of comprehension onto the buyer and subtly positions the seller as uncertain about their own pitch.

Other flagged phrases include "I think" when used as a hedge before a factual claim ("I think our platform integrates with Salesforce" versus "Our platform integrates with Salesforce"), the word "honestly" used as an adverb more than once per call, and the question "What's your budget?" asked before the seller has established value. That last one saw close rates drop by 22% compared to calls where budget was discussed only after a tailored demo or case study.

Perhaps the most counterintuitive finding involves excessive agreement. Reps who said "absolutely" or "exactly" more than eight times in a 30-minute call closed deals at a rate 18% lower than those who used the words sparingly. Analysts suggest this pattern signals over-eagerness and a lack of critical engagement, which sophisticated buyers read as a red flag.

So what do top performers actually say? The analysis found that high-performing reps share several linguistic patterns. They use specific numbers and timelines early in calls ("We helped Acme reduce churn by 14% in the first quarter"). They ask open-ended, diagnostic questions rather than yes-or-no ones ("Walk me through what happens after a lead comes in from your website"). And they deploy what researchers call "confident hedging"—acknowledging uncertainty without undermining their authority ("I want to confirm the exact integration timeline with our engineering team, but based on similar deployments, you'd be live within three weeks").

The methodology behind the analysis is worth understanding. Each platform used its own proprietary model to transcribe and analyze calls, then a shared LLM layer—built on a fine-tuned version of GPT-4—was used to normalize language patterns across datasets and control for variables like deal size, industry, and sales cycle length. The researchers acknowledge that correlation is not causation; a rep who says "to be honest" may simply lack confidence for other reasons. But the consistency of the patterns across 2 million calls and three independent datasets is, they argue, hard to dismiss.

For sales leaders, the practical takeaway is deceptively simple: record, review, and coach on language. Most conversation intelligence tools now offer real-time prompts that can flag problematic phrasing during a call, but adoption remains uneven. According to a recent Forrester survey, only 34% of enterprise sales teams have deployed conversation intelligence broadly; the rest are still piloting or haven't started.

"The gap between teams that use this data and teams that don't is going to become a chasm," said Marcus Delaney, VP of Revenue Operations at a publicly traded SaaS company that participated in the study. "We've already seen a measurable lift just from coaching reps on three or four phrases. Imagine what happens when every rep has a real-time co-pilot in their ear."

The companies behind the analysis plan to release a public version of their findings later this quarter, along with an open benchmark that any conversation intelligence vendor can use to test their own models against the same dataset. In the meantime, sales managers might want to start with a simple exercise: listen to three of your team's recent calls and count how many times someone says "to be honest." The number may surprise you.