An Exploration of AI Coaching for More Effective Communication

Artificial intelligence, a field of research dating back to the 1950s, has for many years impacted the way we work and interact. With OpenAI's release of ChatGPT in November 2022 and image generation tools like Dall-e, the average person's ability to create novel content has skyrocketed, showering the industry with attention and investment that many have long speculated would come. OpenAI recently raised the largest round of venture capital ever! You can’t avoid hearing about AI in nearly any context; it has become commonplace in the global lexicon. Some argue that AI is overhyped; but our friends at Y Combinator, like OpenAI, also in our backyard here in San Francisco, share their perspective as experts in the center of the innovation and investment. It's worth a listen (or at least a summary from ChatGPT if you feed in the transcript). In the last couple of years, we're seeing AI, simultaneously a generations-old field, while still quite nascent in many ways, deliver substantial impact to the way we work. We expect the pace of adoption of new tools to accelerate considerably faster in the years, even months to come. We’re optimistic about continually enhancing our creativity and productivity, and are eager to bring our clients and partners along with us on the journey.

Here at Savas Labs, one of the larger adoptions AI has had is the transcribing, and most importantly, processing of our recorded meetings. Before the pandemic occurred, we had been recording our videos for years. One tool we’ve rolled out across the team is Fathom, a Y Combinator company. Fathom is an AI agent that joins your meetings, records them, and provides a transcript after the call with a concise summary. The comprehensive user interface makes it easy to reference and link to sections of conversations. This tool has done away with our note-taking; it does a better job than we ever did, and we can immediately share transcripts, as well as takeaways from a meeting. This saves many hours/week and performs better than even our best notetakers used to. Win-win.

We now try to record most every meeting. Given this treasure trove of every kind of conversation we have: sales calls, manager/mentor one-on-ones, client presentations, you name it, we figured there might be some value in analyzing conversations. Via Labs, our R&D space, our founder Chris asked us to take a deep dive into the tooling and possibility of an AI assistant to provide coaching and feedback to improve the efficacy of our meetings and communication. As a company committed to learning, growth, and improvement, we were excited to see what the cutting-edge tools can provide. 

At Savas, we have been exploring leveraging existing AI tools to improve our internal processes in a context-appropriate way. We asked: What capabilities does artificial intelligence have in evaluating digital communication and providing insights to our team that we’ll actually use? 

To find an answer, we took the specific approach of evaluating a virtual sales meeting. According to marketing researchers at the University of Tennessee, 68% of Business-to-Business (B2B) buyers prefer communicating online with a salesperson because of the convenience and ease of access to information. Several AI tools have recently been released to improve sales and communication, making this a relevant use case for our experimentation. 

We know that large language models (LLMs) like ChatGPT can evaluate a call transcript and provide constructive feedback while pointing to direct evidence within the call. However, one emerging growth area is LLMs' ability to assess verbal and non-verbal cues during video calls. For example, AI can detect the words used within a vocal delivery, talking time, speaking points, and sentiment analysis based on facial expressions. Although there are limited existing examples of real-time coaching in the market currently, there is promising potential in AI’s ability to provide valuable feedback on communication patterns, talking points to practice, and anticipating client questions. 

Interested to hear the perspectives of our client strategists at Savas, I turned to Zakk Rhine and Cody Davis to gain insight into their experience with successful sales calls. Client relationships begin with us seeking alignment between our values, services, and the client's needs, ensuring we determine essential project details. In our current workflow at Savas, Fathom, our AI notetaker has been an excellent aid for reflection and follow-up on meaningful conversations. But what other tools could be helpful?

I asked Zakk and Cody what they believe would provide the most value to them in improving their sales calls, and one key theme stood out: technical expertise and communication. Zakk and Cody turn to our design and development teams to ensure they effectively answer technical questions from clients and articulate complex concepts. “When you have more knowledge about the specifics, it builds confidence on the other side," Zakk shared. Numerous AI tools are emerging to address the pain point of interpersonal communication and fluidity in sales; however, few existing tools focus on sharpening knowledge and articulating specific technical concepts. This idea of tailored coaching and training based on individualistic needs broadened our view of potential AI applications that could be used across different teams at Savas, such as management training and new hires.

Although the idea of an individualized AI coach is promising, our team prioritizes being mindful of the potential for bias in any algorithm—or AI-based decision-making. Approaching AI tools with an awareness of the limitations of these models is crucial in making responsible decisions on how to utilize the results. Even with the necessary ethical considerations, the potential of using existing LLMs to identify areas for improvement in our internal and external communication is something both exciting and potentially quite valuable in our internal practices. The market gaps in technical training and real-time coaching offer a unique opportunity for further exploration into how a prospective AI tool could be designed to streamline feedback from virtual calls.