How AI can provide conversational coaching
A hypothesis. Prepare for a call with AI, get real-time feedback, then reflect and grow.
See what it became, Striver at Matters
Reformatted from its original design.
In today's fast-paced, data-driven world, artificial intelligence (AI) is revolutionizing our approach to virtual communication. At Savas, a remote-first company, we are keenly familiar with the workflow of meeting and collaborating online, and, like many other companies in a post-pandemic, AI-chattering world, we are eager to understand how we can harness the power of AI as this technology rapidly expands.
Our research into conversational coaching sparked internal discussion about the future of AI and virtual feedback, and we were inspired to explore innovative approaches to digital communication and training. After spending the last few months researching and testing the capabilities of existing Large Language Models (LLMs) to immediately provide team feedback from our virtual meeting transcripts, we considered how AI could help prepare for important calls and presentations. We envisioned an AI assistant that could coach team members effectively in real-time and provide thoughtful reflection points afterward, helping them be better informed for next time.
This technology could be used in a wide range of professional scenarios: candidate interview preparation, design presentations, client scope conversations, and internal one-on-ones for managers and their direct reports. The more people use this tool, the more team members across distinctive roles could become readily equipped to handle their unique challenges, build stronger relationships, and feel more confident in their communication.
Preparing for an important meeting
Presentations are as much about delivery as they are about content. AI can be the ultimate coach, analyzing video and audio to help team members prepare for a meeting and evaluating a presenter's performance afterward. Before the presentation or important meeting, users could privately practice with an AI assistant, refining their approach and key discussion points before the real deal.
Transcript: "So, this moodboard is meant to kind of, you know, capture a natural, sustainable vibe with earthy tones and give a feeling of calmness and connection to the environment."
Insight: Clarify using more concise language: "This moodboard captures a natural, sustainable vibe through earthy tones, evoking calmness and a strong connection to the environment."
Suggestions: Rehearse your talking points ahead of time and aim to reduce filler words to explain design choices clearly and confidently. Before the call, review the client's mission statement and specific goals and prepare to reference how each design element supports those objectives.
Time-stamped performance overview
By combining video and audio, the AI assistant could assess comfort levels with specific topics, identify strengths and weaknesses, and note changes in word usage or signs of rambling. Visual cues on a timeline, synchronized with the video, could point out specific moments needing improvement, such as a lapse in attention or an off-topic digression. For each flagged point, the system could offer tips to refine delivery and body language, ultimately promoting confidence.
Live feedback: a sidekick on the call
During a call, immediate feedback on performance poses a risk of distraction or making the speaker sound overly scripted. The AI assistant could provide a method for live feedback that would enhance performance and allow people to be more confident and prepared. This idea has two parts: preparation and priming before a call, and display and delivery timing of this data during a real-time conversation, via phone call or virtual meeting.
5 minutes remaining in your call. Tip: Bring up budget constraints and timeline limitations.
Answer suggestion: Address different CMS capabilities and what the client should anticipate during the handoff of the product.
Data-driven real-time suggestions
Harnessing data from an AI practice call, previous client calls, or meetings with similar clients, the AI assistant could leverage this information in real-time conversations. Flexible to individual preferences, this technology could adapt to a user's needs and goals, whether that is cues via mobile notifications or insights as speaker notes present during a call. Embracing uncertainty, AI has the profound potential to be a highly personalized sidekick to enhance communication.
AI pattern recognition
Imagine a system that not only tracks questions but also identifies recurring themes across similar conversations with different people, showing us the connective tissue we cannot see. By consistently feeding AI the transcripts from calls, patterns emerge, including those we often miss in real-time conversations. For instance, what if the same type of question from potential clients comes up in 10 different ways?
The AI assistant could detect and group these variations, predicting client behavior based on historical data. The insights, or highlighted themes, are linked to specific client quotes from call transcripts.
The client said: "We just can't spend what we don't have. So, you know, that's certainly an important possibility to consider... we can budget it for the next year."
Insight: The client frequently emphasizes their budget limitations and the importance of minimizing unnecessary costs.
Private conversation cues
Primed with this information, a customizable AI assistant could be present in the call, sharing private and selective cues via the in-message chat that remind you to tune in, how to answer specific questions, and if you missed any important or relevant topics based on the time left in the call. This subtle yet powerful tool could help the individual gain a greater sense of self-awareness and confidence going into a meeting.
[Private] AI Assistant. Key notes for conversation: Budget conversation: aim to comfortably discuss budget expectations earlier in the conversation to ensure financial alignment. Streamline documentation protocol: share process for sending NDAs and necessary documents post-meeting.
[Private] AI Assistant. Reminder: 5 minutes remaining. Wrap up any questions related to design and development handoff process.
Post meeting: reflecting and devising a plan for the future
Not all feedback should be received or delivered the same way. Focusing an AI-powered tool on the art of communicating and receiving feedback through the lens of various personas would provide an array of perspectives that might otherwise go unnoticed. At Savas, this tool could be built upon real conversations with developers, designers, and managers, tailored to make suggestions accordingly, based on whatever perspective an individual needs.
Custom feedback on call (as a Senior Software Engineer, Senior Product Designer, or Design Team Manager)
22:39, emphasis on accessibility: Ian could strengthen his point by sharing concrete examples of how accessible design has improved usability in similar projects. "In a recent project for a university website, we found that accessible color contrast and larger font sizes improved user engagement by 15% among older audiences."
Various personas
Existing tools like Otter, Fireflies, and Fathom (our favorite) leverage abilities to query existing transcripts through various templates. We envision leveraging similar functionality to allow users to choose how they want to receive the AI assistant's feedback.
People on our team are interested in hearing perspectives outside their expertise to better understand how to approach a particular problem and improve for the future. For instance, our client strategists expressed wanting to sharpen their ability to articulate technical concepts related to development to clients. The AI assistant could paraphrase the script with this context in mind and suggest how to clarify or refine key points based on the user's role.
The bigger picture: how these tools work together
The convergence of these three solutions presents an opportunity to completely transform client interactions, team feedback, professional development, and training. By embracing AI to identify patterns, understand perspectives, and deliver tailored coaching, businesses could create a more strategic and confident workforce.
- For client strategists: Predicting client questions and framing responses based on historical wins and losses could be game-changing, creating a proactive approach rather than a reactive one.
- For teams and individuals: Tailoring feedback through the lens of various personas makes each critique actionable and relevant, promoting clarity and understanding.
- For presenters: Personalizing coaching tools ensures that improvements are specific and confidence-focused, leading to a better overall presentation experience.
At the rate at which AI integration is becoming commonplace, our realities are seamlessly becoming intertwined with the technology surrounding us. Although these ideas are speculative and come with their own challenges and risks, the strategic use of AI for feedback and preparation could ultimately support individuals in their personal and professional development. Instead of fearing the unknown, we are excited by imagining the possibilities of using these tools for the greater good, upholding our mission of maximizing positive impact.
Read the series
We wrote up the research in two posts: An Exploration of AI Coaching for More Effective Communication and Application of AI Coaching at Savas Labs.
This research continued at Matters, our sister product studio. Formerly Labs, now Matters.
