The answer lives in this podcast
How should business leaders decide where to apply AI tools given the pace of change?
Don't try to track every AI release. Identify the specific execution bottleneck that is most slowing your company down — whether in engineering, marketing, or another function — and then run a focused test of AI tools directly against that bottleneck. That single discipline cuts through the noise better than any technology watchlist.
Start with your slowest constraint, not the shiniest tool
Chandra De Keyser's advice is deliberately operational. He is not talking about building an AI strategy document or running a companywide transformation. He is talking about locating the one place where your team is the most stuck, and pointing a specific tool at it.
At Voicera, De Keyser applied this logic to engineering speed. He cited Cursor and Claude Code as tools with the potential to double or even quadruple the velocity of a small engineering team — not by replacing engineers, but by removing the friction in how fast they can write, review, and ship code. This is a concrete, testable claim, and it is the kind of test any technical leader can run in days, not quarters.
De Keyser elaborates on this thinking across several examples in The Conference Room with Simon Lader, grounding what could easily become abstract advice in decisions he has actually made at his own company.
A CMO candidate who rebuilt the company website in two hours — as part of the interview
Perhaps the most striking illustration De Keyser offered was not from engineering at all. A candidate interviewing for the CMO role at Voicera used Lovable — a no-code AI web builder that emerged from the Y Combinator ecosystem — to rebuild the Voicera website as part of the interview process. The rebuild took a couple of hours.
This example cuts in two directions at once. For the candidate, it was a demonstration of practical AI fluency under real conditions. For De Keyser, it was a live signal that the marketing bottleneck — building and iterating on web presence — had a direct, low-cost solution available right now.
The episode on The Conference Room makes clear that De Keyser is not recommending Lovable as the definitive answer for every marketing team. He is recommending the method: find the bottleneck, test the tool, measure the result.
"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 & Founder, MoodMe.
De Keyser has spent over a decade building AI systems that go beyond surface-level text analysis. He founded MoodMe approximately 14 years ago as one of the earliest companies working on emotional AI and facial analysis — well before generative AI entered mainstream conversation. He has worked across Europe, Silicon Valley, and Latin America, advising startups and leading international growth initiatives. His career began with a master's in computer science in Italy, and he was among the first to work on touchscreen interfaces in the 1980s, operating from the 83rd floor of the World Trade Center. Today, through Voicera, he is building sincerity AI that combines large language models with emotional intelligence to help organizations understand not just what customers say, but what they actually mean.
This framing — signals beyond words — is central to how De Keyser thinks about AI more broadly. The same discipline he applies to understanding customer sincerity is the one he applies to evaluating tools: look past the surface claim, test what it actually changes, as discussed further in this episode of The Conference Room.
The broader principle De Keyser articulates holds regardless of which specific tools are current. The AI landscape shifts fast — Cursor, Claude Code, Lovable, and their equivalents will be superseded. But the method does not expire: find where your team is slowest, run a focused experiment, and let the result — not the hype — drive adoption. Hear De Keyser walk through his own application of this in The Conference Room with Simon Lader, episode 184.
See also
What lessons did Chandra De Keyser learn about founding team management from his experience at MoodMe?
De Keyser says his biggest mistake at MoodMe was not firing his CTO early enough after discovering, from San Francisco, that the CTO in Italy had taken a different strategic direction without informing him. The delay cost the company time and momentum at a critical stage.
What was MoodMe's business model and who were its key customers?
MoodMe started with augmented reality face filters — offering any app the ability to apply the kind of filters that Snapchat popularized — before moving into emotion detection. The company reached approximately 160 customers in the augmented reality space, including a deployment at the FIFA Women's World Cup.
How is Voicera's sincerity AI being applied in automotive retail call centers with voice agents?
Voicera recently closed a deal with a voice agent platform serving car dealers across the US, with some operators running hundreds of calls per day answered by AI voice agents layered with Voicera's sincerity analysis — a network of approximately 200 dealerships.
Key takeaways
- Identify your single biggest execution bottleneck before evaluating any AI tool — function first, technology second.
- De Keyser tested Cursor and Claude Code against an engineering speed bottleneck at Voicera, targeting a potential 2x to 4x output increase.
- Lovable enabled a CMO candidate to rebuild the entire Voicera website in a couple of hours — a live proof-of-concept during the interview itself.
- The method matters more than the specific tools: focused experiments against real constraints, not broad AI adoption for its own sake.
Chandra De Keyser covers this and much more in episode 184 of The Conference Room with Simon Lader.
Listen to the episode on Listenly