In large sales organizations, managers have no bandwidth to watch every sales call — so they rely on transcripts and LLM summaries that capture what was said, but not how it was said. Voicera analyzes video and audio signals at scale to surface deals where a customer's words signal enthusiasm but their behavioral cues signal insincerity — turning bloated revenue forecasts into accurate ones.
When a sales team has thousands of representatives each conducting 20 to 30 calls per week, reviewing video footage of every conversation is simply impossible. Sales managers default to written transcripts or AI-generated summaries — tools like Gong or Fathom — to keep up.
Those tools are good at capturing content. They are blind to tone, hesitation, micro-expressions, and the dozens of non-verbal signals that experienced salespeople read instinctively in a room. The result: a deal marked "high confidence" because the customer said the right words, when the underlying signals point in the opposite direction.
This is the precise gap Voicera fills. Its platform processes the video and audio layers of sales calls — at the scale of entire enterprise teams — and flags deals where something doesn't add up. Hear Chandra De Keyser explain the full mechanism on Listenly.
De Keyser uses a concrete example: a customer says the words every sales rep wants to hear — "I like your product and I have the budget." The transcript looks great. The LLM summary is positive. The deal gets logged as likely to close.
But the video and audio signals tell a different story. The customer's tone is flat, eye contact is absent, the pacing is off. Voicera's model catches these signals and flags the deal as at risk — before it contaminates the revenue forecast with false optimism.
The downstream impact is significant: revenue forecasts reflect reality rather than the over-optimistic picture that emerges when organizations trust words alone.
"What was said is part of the story. How it was said is sometimes as important, if not more important. So that's where we add tremendous value, because we can unherd these signals."
Chandra De Keyser — Co-founder, Voicera.
De Keyser has spent his career building technology that reads humans objectively at scale. He founded MoodMe approximately 14 years ago — one of the earliest companies in emotional AI and facial analysis, predating the generative AI wave — before pivoting to Voicera's sincerity AI platform. He began with a master's in computer science in Italy, then joined a startup in the 1980s working on an early touchscreen device from the 83rd floor of the World Trade Center. His work has spanned Europe, Silicon Valley and Latin America, moving from augmented reality deployments (including a face-filter campaign at the FIFA Women's World Cup) to enterprise-grade sincerity detection for sales teams.
Because there were no existing sincerity data sets, Voicera built its own proprietary data set by design, deliberately including a diverse mix of signals to train the model from the ground up.
Generative AI is a broad category that produces content — mainly text through LLMs like ChatGPT, Claude, or Grok — as well as images and video. Sincerity AI goes further by analyzing not just what is said but how it is said, combining large language models with emotional intelligence to detect behavioral signals beneath the words.
As of the time of the interview, Tracefuse works with 700 brands and has removed over 16,000 reviews. Shane Barker attributes this growth primarily to the platform's focused approach to review management.
This answer comes from episode 184 of The Conference Room with Simon Lader. The full conversation with Chandra De Keyser covers sincerity AI, emotional intelligence at scale, and what revenue teams are missing by trusting transcripts alone.
Listen to the episode on Listenly