The Future of AI-Powered Training: Conversational Coaching

I recently sat down to complete a mandatory online security training at work. You’ve probably had to do something similar: a 45-minute online quiz with lots of little examples, occasional pop-ups to make sure you’re still paying attention (whoops) and multiple choice questions at the end to test what you’ve learned.

Today’s learning and development (L&D) methods are pretty one-directional, lacking both proven effectiveness and scalability. Either you rely on relatively boring software to teach a skill, or already over-stretched managers have to spend their limited time training employees. What if there was a far better way to leverage software to support learning? What if you could learn and prove mastery through conversation?

OpenAI’s upcoming release of the GPT-4o audio modality may hold exactly these exciting opportunities for conversational learning and coaching. Prior to GPT-4o, you could use “Voice Mode” to talk to ChatGPT, but the latency was noticeably slower than regular human response time in conversation. GPT-3.5 had an average latency of 2.8 seconds, while GPT-4’s latency averaged 5.4 seconds.

OpenAI’s announcement claims that GPT-4o’s voice modality “can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in a conversation… GPT-4o is [also] especially better at vision and audio understanding compared to existing models.”

The audio modality hasn’t been released publicly yet, and its exact capabilities remain to be seen, but both the short-term and future possibilities are enormous. As OpenAI CEO Sam Altman put it in a recent blog, “Talking to a computer has never felt really natural for me; now it does.”

To revisit my security-training anecdote, what might that training look like powered by this new technology? Instead of annoying pop-ups and multiple-choice options, which are all quite passive, what if I were asked to converse with an AI and explain what phishing is and how to prevent it, receiving feedback to improve my skills in the process?

The phrase “conversational AI” is already taken (it’s a term of art for chatbots in a customer service setting). So, let’s call this new AI-powered paradigm “conversational coaching.” We see a lot of potential for this new model, and history is instructive on why.

Active vs. Passive Learning: A Capsule History of Conversational Coaching

You’ve probably come across the idea that “you only really know that you understand something if you can teach it to someone else.” Socrates certainly thought so when he quizzed Meno about the nature of virtue. Conversational learning and oral assessment were core to classical education for good reason. In the 19th century, they were still held in high esteem for their thoroughness.

As one Oxford scholar wrote, students examined orally “have been thus tried more completely than could be done by printed papers; for a man’s answers suggest continually further questions; you can at once probe his weak points; and, where you find him strong, you can give him an opportunity of doing himself justice, by bringing him out especially on those very points.”

Only in the 20th century did written exams swing into vogue, with their vibe of objectivity and their permanence as a written record. But the artifacts of conversational learning are still seen in higher education today in everything from university discussion seminars to PhD oral exams and the Oxbridge system of tutorial-based learning.

The value of oral assessment and learning is also well supported by current research. Compared with written exams, oral assessments offer a rich, if not richer, understanding of student learning. They also give instructors the chance to probe student responses in detail. Oral exams give students flexibility in elaborating their thinking and developing their capacities in communication. Oral examination allows instructors to evaluate more intangible qualities beyond straight knowledge that are otherwise difficult to test for. When it comes to capturing mastery of a topic, oral evaluation has big advantages over far more passive methods like multiple-choice.

Conversational Coaching: Potential Applications Across Business Workflows

Consider the situations in the business world where employees need to quickly master new skills and demonstrate expertise.

For instance, in sales, ramping up on a new product that requires subject-matter expertise in a technical domain can be challenging. Practicing with an AI coach could help salespeople learn about the new offering at their own pace, improve their answers to tricky customer questions, and ultimately, close more sales.

Support teams might practice a new technical support issue area with an AI tool, which would then notify their manager when they’d mastered the new subject matter. This would also benefit onboarding new employees.

The potential use cases for conversational AI in HR are similarly transformative. Consider the range of difficult conversations HR needs to prepare for, from compensation discussions to managing conflicts and layoffs. Conversational coaching could help HR teams and managers practice sensitive employee conversations and equip employees with better negotiation skills.

In my own experience at Battery Ventures, I am very nervous about “throwing new hires into the fire” for one-on-one conversations with CEOs. If a CEO asks about which portfolio companies we have based in France, or at what stage we invested in a certain business, will they answer correctly? An AI tool that helps gauge readiness could expedite this process with greater confidence.

Recently, I received presentation training to support my public speaking work. While I had a great coach, this kind of support is difficult to deliver at scale. Not every employee has access to personalized coaching. Conversational coaching can democratize access to personalized support and unlock a lot of untapped talent.

Conversational Coaching: Challenges on the Road to the Future

Conversational coaching faces challenges. One clear question is whether people will be willing to talk voice-to-voice with their computers. Explaining out loud how phishing attempts work to my computer does seem a bit odd.

Interestingly, studies show younger people’s preference for AI over humans for therapy. That is not a use case where I’d naturally think people would prefer machines! For sensitive topics, or areas where we worry we have stupid questions, AI might feel more approachable.

Privacy and security are critical considerations. Ensuring users that their data is secure will be paramount. Will employees feel open asking “dumb questions” to AI when their manager might get a report on what was discussed?

Ultimately, this may be the most prosocial use case for AI. Conversational coaching is not about automating away headcount; it is about democratizing access to personal support, helping people flourish in their jobs and beyond. That is a future we can all support and one I am excited to invest in.

The information contained herein is based solely on the opinion of Brandon Gleklen and nothing should be construed as investment advice. This material is provided for informational purposes, and it is not, and may not be relied on in any manner as, legal, tax or investment advice or as an offer to sell or a solicitation of an offer to buy an interest in any fund or investment vehicle managed by Battery Ventures or any other Battery entity. The views expressed here are solely those of the author.

The information above may contain projections or other forward-looking statements regarding future events or expectations. Predictions, opinions and other information discussed in this publication are subject to change continually and without notice of any kind and may no longer be true after the date indicated. Battery Ventures assumes no duty to and does not undertake to update forward-looking statements.