Application of AI Coaching at Savas Labs

At Savas Labs, we eagerly explore artificial intelligence's (AI) capabilities to enhance our processes and provide valuable insights. Building upon part one of this blog series on AI conversational coaching, we explored ChatGPT’s potential to provide tailored feedback based on transcripts from our internal recorded calls and meetings. Although ChatGPT’s outputs are imperfect, they exceeded our expectations by broadening our perspectives and providing insights into how we could improve communication with our team. Excited by these results, we continued to stretch the possibilities of existing large language models (LLMs) through trials by prompting ChatGPT with specific directions. Below, we’ll describe how Savas Labs has applied LLMs like ChatGPT to learn and improve communication, sales, and management strategies.

Part one of this blog series, “An Exploration of AI Coaching for More Effective Communication,” details our research into the capabilities and limitations of AI coaching, as well as the implementation of Fathom, an AI notetaker, in our virtual meetings at Savas. Fathom has transformed our workflow by automating meeting transcription and summarization, enhancing efficiency and accuracy. Given the impact and power of this tool, we investigated further applications of AI within a virtual call context to enhance our feedback process in real-time, as needed.

We were surprised to gain new and valuable insights solely from transcripts. One of the most promising applications of AI at Savas Labs is evaluating sales calls and one-on-one managerial meetings. By identifying specific areas for improvement aligned with individual goals, we can leverage AI to provide focused feedback. With tailored AI analysis of these call transcripts, we can offer insights that were previously difficult to obtain without involving senior staff in every review.

We have found that ChatGPT can:

• Pinpoint Moments for Improvement: AI can identify key points within sales calls where performance can be improved down to the second. For example, it suggests adjustments to specific phrases to enhance clarity and impact.
• Offer Tailored Feedback: By analyzing conversations from various perspectives—such as those of a seasoned engineer or a senior product designer—the AI provides specialized insights on technical communication, allowing our team to refine their messaging.
• Provide Detailed Recommendations: ChatGPT offers a range of feedback, from high-level observations on communication dynamics to nuanced phrasing suggestions. It can also anticipate potential client questions and recommend strategies for future discussions.
• Devise a Plan for the Future: ChatGPT can generate specific and concise goals for the individual based on areas for improvement noted within the transcript. These can be modified based on the individual’s personal goals.

That said, AI isn’t without its limitations. At times, it may generalize feedback, be overly verbose, overstate specific issues, or misinterpret call dynamics. For instance, the AI might overemphasize speaking time or mistakenly view going over time as a negative when it could indicate productive engagement (or simply just vibing—something AI has yet to pick up on the nuances of). Subjective factors, such as tone, are difficult to determine from a transcript alone.

Beyond sales, we’ve also been exploring AI’s potential in supporting managers with feedback on their one-on-one meetings. ChatGPT evaluates elements such as delivering constructive criticism, staying on topic, and offering affirmations or support to team members. 

While the results have been broadly positive, there have been instances where AI misjudges relational nuances, such as interpreting extended discussions as unproductive rather than indicative of meaningful engagement. Our team members reviewed and commented on ChatGPT's interpretation of their calls with respect to their own. With a fair mixture of spot-on observations and some humorous misinterpretations, these insights have provided managers with valuable reflections on their communication styles and strategies to build positive relationships. 

As we continue to explore AI’s capabilities, we recognize a wealth of opportunities to embed these tools more deeply into our daily workflows—from refining sales and HR training, project manager facilitation of client meetings, to managerial development from one-on-one meetings. While we’re excited about AI’s potential to support different teams at Savas, we remain mindful of the need to refine these technologies and apply them responsibly to maximize their effectiveness. 

One idea to address the limitations we observed within AI’s outputs is to create custom GPTs. For a custom GPT, you can manually input context, prompts, and documentation to help streamline the prompting process and receive more specific and accurate feedback. For example, a custom GPT could include information related to an individual’s role and goals, various personas of the AI (designer, manager, developer, etc.), and possibly even Predictive Index personality types. This may make the process of utilizing AI more efficient and more accurate. 

Our initial experiments with AI have revealed promising possibilities, not just in sales and one-on-one managerial meetings but throughout our entire organization. We’ve seen firsthand how tools like ChatGPT can help pinpoint areas for growth, deliver valuable insights, and streamline otherwise time-intensive processes. While feedback directly from peers is an essential part of our culture and values at Savas, we are excited by opportunities to enhance personal reflection and training by utilizing the AI tools currently available. 

Looking forward, we’re eager to continue exploring AI’s applications in other domains, such as product design and team training. As we push these boundaries, we aim to enhance our work further and deliver even greater value to our clients. Stay tuned as we share more about our journey into the evolving landscape of AI at Savas Labs!