What if you could simulate a difficult interview or conversation before it ever happened?
No, this isn’t a review of “The Rehearsal,” the HBO Max show where people can train for an uncomfortable conversation by simulating it beforehand. It’s the ambition of speech analysis startup Yoodli, which unveiled new AI-powered tools that help users simulate dialogue and analyze speech patterns.
The new tech uses generative AI to produce real-time follow-up questions based on a user’s responses to questions for a mock conversation. Yoodli also introduced a personal speech coach that can be used during live discourse, providing immediate contextual feedback and analysis.
The release of the new tools come as hundreds of thousands of tech workers are re-entering the workforce amid widespread layoffs, Yoodli co-founder Varun Puri told GeekWire. The goal is to help people land their next job by helping them ace their next job interviews.
Puri gave GeekWire a demo last week. Here’s how it works:
- Users select the questions they want to practice for the mock interview simulator. They can choose from a range of topics and levels of complexity, with the tool adjusting the depth of questions based on answers the user gives.
- After answering the initial prompt, the tool spits out a follow-up question. The mock interview can last up to a minute, allotting time for a total of three questions.
- Once the interview is over, Yoodli provides analytics on word usage and delivery. It also offers suggestions for how to rephrase certain answers.
During the demo, Puri did a mock job interview for a position at GeekWire. Puri acted nervous, helping to demonstrate Yoodli’s ability to provide feedback.
The tool found that Puri used the filler word “um” 10 times. His pacing was a bit fast, speaking at 180 words-per-minute versus the recommended 170. Yoodli also rephrased Puri’s answer about his experience in journalism. Here’s his original response, then the rephrased version produced by Yoodli’s AI:
Puri’s original answer: “Yeah, that’s a great question. You know, I was like the head of my school, uh, newspaper. Um, I also wrote blogs for GPT, and now I’m talking really, really quickly because I also, in the past had my own little website and my website went viral and it was picked up by CNN and BBC and 250,000 media outlets. And I’m just so overqualified.”
Yoodli’s rephrased answer: “That’s an excellent question. I served as the head editor of my school’s newspaper and contributed to blogs for GPT. Additionally, I owned a personal website which became viral and was featured in major news outlets including CNN and BBC, gaining momentum through 250,000 other media resources. Overall, I believe that with my extensive background, I am highly qualified for this.”
Yoodli’s private speech coach only captures the voice of the user and is designed to give real-time contextual feedback. It can tell a user if they are speaking too fast, using too many filler words, or dominating a conversation by speaking for a majority of the call.
“It’s like your buddy who will poke you under the table,” Puri said.
Yoodli’s new features are currently being tested by students at the University of Washington Foster School of Business to practice for job interviews. They are also being rolled out more broadly through Yoodli’s partnership with Toastmasters, which has more than 300,000 members worldwide.
The startup’s platform uses natural language processing to provide users with a transcript and analysis on use of filler words, non-inclusive language, pacing, body language, among other actionable insights. This data is private to the user but can be sent to friends or coaches who could provide human insights. Here’s an example of Yoodli analyzing a speech by Elon Musk in 2014.
There are a number of competitors pitching similar AI-powered speech coach training platforms. Seattle-based startup Read AI recently launched a platform called “Speaker Coach” with similar offerings; there’s also Philadelphia-based startup Orai. While there are a number of tools that offer AI-generated transcripts and other similar features, Puri said Yoodli is the first of its kind to focus solely on speech and communication.
Asked what happens if the Yoodli platform provides inaccurate feedback or misguided questions, Puri said that the platform will learn from user behavior and respond. He said the AI is not always going to be completely accurate, but that the company did “a lot of prompt engineering to keep it within limits.”
Yoodli was founded in 2021 at the Allen Institute for Artificial Intelligence (AI2) incubator in Seattle by Puri and entrepreneur Esha Joshi. Puri is a former product manager with X Development, the R&D organization of Google parent Alphabet, and Joshi was a product manager at Apple. They won Young Entrepreneur of the Year honors at last year’s GeekWire Awards.
The startup has raised more than $7 million from Madrona Venture Group, Cercano Capital, AI2, and others.
