The business world is buzzing about the release of ChatGPT4 Omni (GPT4o), and what it means for tools that use GenAI. We spoke with Ariel Hitron, CEO of Second Nature, to find out what he predicts for the GenAI world, how it will enhance Second Nature, and the extra benefits it will drive for Second Nature users. 

What can we expect to see in the GenAI market?

There’s currently a fierce war going on between different vendors, and the market is moving incredibly fast. Each company wants to have the most cutting-edge large language model (LLM), which is the engine that powers generative AI solutions. 

But the window within which an LLM has an edge is very short. For example, OpenAI spends billions of dollars training a model, and just a few months after they release it, a competitor – usually Meta – copies it and makes it open source. 

This erodes the value of the model extremely quickly. As a result, vendors can’t expect their new LLM to deliver ongoing value for the user or revenue for the vendor. There’s some talk in the market that big cloud providers will buy LLMs and offer them as part of their overall cloud offering, so customers won’t really be choosing an LLM for themselves. 

That’s why the real value lies in the applicative, or application, layer. This is an interface which sits on top of the LLM, and interacts with both the LLM and the user. When the user inputs a prompt to their solution, the application layer interprets it and sends the instructions to the LLM, and then returns the response from the LLM to the user. Second Nature is an example of an application layer. 

 

How does competition in the LLM market benefit Second Nature?

Because Second Nature is in the applicative layer, not the LLM itself, we (and therefore our customers) gain an advantage from the competition among LLMs. As LLMs get stronger, so does Second Nature. 

Before LLMs took off, we built Second Nature’s conversational AI processes ourselves. When the ChatGPT3 beta arrived, we were part of early beta testing and started embedding it into our processes. Ever since then, we’ve had a growing relationship with them and other LLMs, looking for the best ways to serve the business needs. 

For enterprises, the LLM is of limited benefit unless it’s grounded in your enterprise data, kept secure and private according to your data privacy standards, and is part of your enterprise workflow and systems, for example single sign-on. That’s why pure LLMs don’t compete with Second Nature, because we have the data needed to refine the model and all the enterprise bells and whistles which aren’t really the focus of the LLM companies. 

We all saw the launch video of a flawless conversation between the OpenAI product professional and ChatGPT4o, including visual cues. What makes Second Nature a better solution for customer-facing training?

In theory, any organization could buy a premium version of ChatGPT4o for every employee, and they could all use the app to hold practice conversations with AI. This would be sufficient for an individual who wants to practice their phone skills or public speaking skills. But it wouldn’t deliver effective training experiences for a company, for two main reasons. 

First of all, you have no control over the kind of conversation that your employees have. If you’re trying to roll out new messaging across your entire thousands-strong sales force, you want them to be practicing on a consistent simulation that is grounded in your brand content. You want to coach, train, and certify each rep in cold calls, demo calls, your sales deck, objection handling, etc. Without Second Nature in the application layer, their conversations won’t be relevant to your use case. 

Second, you need visibility into your employees’ training experiences. How will you know who has completed the training, how well they performed in their practice calls, what kind of progress they are making, and whether they’ve understood the message you wanted to convey? Will you know if the way that you explained the new messaging was unsuccessful and your employees are confused? It’s vital to be able to see who is practicing, what kind of feedback they are receiving, and how effectively they are improving. 

Basically, any GenAI conversational tool needs to be integrated into your workflows and grounded in your data, and that’s not something you can get off the shelf from an LLM. You need something like Second Nature in the application layer if you want to create relevant training sessions and gain visibility into the user experience.

How is Second Nature using ChatGPT4o to enhance the user experience?

First, let me point out that (as of the date of this interview) the full GPT4o functionalities aren’t yet available as a stable API. But as soon as it is, we will implement them for our customers to have a superior user experience. We’re committed to delivering the bleeding edge of innovative technology, paired with enterprise reliability and scalability.

One of the main benefits will be an improved conversation experience. There’ll be less latency and faster responses from the AI avatar, and you’ll be able to interrupt and talk over them, enabling more realistic voice conversations. Your role play partner will have a more expressive voice and more nuances of tone, and we’ll support more languages. We’ve already added Spanish, German, and Portuguese, and other languages are on the way. 

Using ChatGPTo, our training avatars will be able to respond to your video input as well as your audio. In practice, that means that they’ll say something like “You’re looking tired today” as well as “you sound perky this morning.” It can also read your screen shares, so it can respond to slides that you present as part of a demo, for example. We’re also working on a speaking video avatar, although that requires a lot of compute power which means a higher price point, so we’re evaluating how crucial this is to our customers. 

We’re also using GPT4o to elevate the analysis and feedback that we provide. For example, users will receive feedback on their facial expressions at various parts of the conversation. The analytical engine will use facial expressions as another data point that indicates where the user is hesitant or uncertain, which makes the feedback more valuable.

What is your vision for Second Nature with ChatGPT4o?

My goal is for Second Nature to become a sales conversation copilot that supports you every step of the way. 

You’ll be able to use it before a sales call to practice the conversation that you’re about to have, with a training session that’s grounded in your data. For example, you’ve already spoken a few times with representatives from a particular company, but now you’re about to talk to the CFO who is a decision-maker. You’ll receive a role play that draws on all the information you’ve collected about that company during all your previous conversations. 

Then during the call, the platform will give you real-time prompts that remind you about points you need to cover and suggest areas that you should emphasize. Finally, after the call you’ll see detailed feedback that analyzes your performance in that call and delivers actionable advice for improvements in the future. 

Any final words?

We’re excited to incorporate ChatGPTo into our solutions, and will roll it out to our customers as soon as we feel it’s enterprise ready.