# From Chatbots to Strategic Assets: Making AI Assistants Work For You

> A NachoTuesday interview — the C-Suite Insights series from NachoNacho, the B2B SaaS and AI
> marketplace. In conversation with Hunter Zhao, Founder, GPT-Trainer.

- **Page:** https://nachonacho.com/nachotuesday/video/From-Chatbots-to-Strategic-Assets-Making-AI-Assistants-Work-For-You
- **Guest:** Hunter Zhao, Founder, GPT-Trainer
- **Guest on LinkedIn:** https://www.linkedin.com/in/hunter-zhao-1b47a329/
- **Recorded:** 2025-07-08
- **Watch:** https://www.youtube.com/watch?v=FOYYJ87rYrs

## Summary

This webinar features Hunter from GPT-Trainer, an AI agent platform that helps businesses automate support, qualify leads, and conduct document analysis. The platform offers an intuitive chatbot training workflow for creating and deploying RAG-based multi-agent systems that can be integrated into websites, Slack, SMS, WhatsApp, Facebook Messenger, Instagram, and CRMs like HubSpot and Zendesk. GPT-Trainer also plans to roll out an AI voice feature for inbound and outbound calls, opening up new possibilities for sales outreach and qualification. The company emphasizes providing custom solutions for specific use cases, especially for mid-market to enterprise clients who prioritize data security and deep integrations with existing business systems. Key Takeaways: Versatile AI Agent Platform : GPT-Trainer offers a cutting-edge platform for automating support, qualifying leads, and analyzing documents, with flexible deployment options across various communication channels and CRM integrations. Focus on Customization and Security : The company provides dedicated hosting and custom implementations for clients, ensuring data security and deep integration with existing business systems, which is crucial for larger enterprises. AI Voice Capabilities : A new AI voice feature is being rolled out to enable automated inbound and outbound calls, allowing for direct automation of phone support, lead qualification, and call scheduling. Addressing Industry-Specific Needs : GPT-Trainer has successfully built AI modules for niche use cases, such as a medical device regulatory compliance assessment company, demonstrating their ability to handle domain-specific and complex challenges. Simplifying AI Adoption for SMBs : For small to medium businesses without in-house AI expertise, GPT-Trainer offers a no-code builder that allows for the creation of simple chatbots in under 10 minutes, with technical assistance available for more advanced workflow automations and data processing. Impact on Productivity and Cost Savings : Adopting AI agents can significantly improve operational efficiency, make existing teams more productive, and reduce labor costs by automating routine tasks. Future of AI Interactions: The future of AI will involve multimodal LLMs and standardized protocols, leading to more powerful, versatile, and natural-sounding AI agents that will become indistinguishable from human interactions and raise the bar for customer service. AI and Job Displacement : AI is not expected to replace people but rather to replace those who do not learn to work with AI, similar to past technological revolutions. Learning AI tools will increase an individual's value in the market.

## Full transcript

Host: Welcome, everybody, to another NachoTuesday! Today, I have Hunter here from GPT-Trainer. Without further ado, we'd love for you to introduce your company and tell us what you do. Hunter: Absolutely, Andy. Thank you for having me on the show. GPT-Trainer is a cutting-edge AI agent platform that empowers businesses to automate support, qualify leads, and conduct document analysis. Out of the box, it features a very intuitive chatbot training workflow that lets you create and deploy RAG-based multi-agent systems. You can integrate these on your website, in Slack, as a standalone link, or as part of a login-protected library. You can also connect it with SMS, WhatsApp, Facebook Messenger, or Instagram. We also have built-in native integrations with popular CRMs like HubSpot and Zendesk. If a direct integration isn't supported, we offer a general-purpose API that makes it super easy to connect with other business systems. We also have dedicated support within Zapier, Make, and Workato, which are quite popular nowadays. One really cool thing we're working on right now, Andy, is that in a couple of weeks, we're going to be rolling out AI voice. This will let you connect our AI agents to phone numbers you purchase directly inside the platform, enabling both inbound and outbound calls with this new feature. So, that's going to open up a brand new dimension of possibilities within GPT-Trainer. Host: Very exciting! So, that's coming out in a couple of weeks? Hunter: That's right, that's right. Before the end of July, we'll have the first iteration. We're going with a phased rollout. The first iteration will probably have inbound support and maybe single-number outbound call support. But we're going to evolve it to support number sets for outbound calls so you don't get rate-limited by telecom agencies. This way, you can have 10 or 20 numbers that rotate, making no more than six outbound calls a day per number. This ensures we stay fully compliant while still implementing automation for sales outreach and qualification. The Journey to GPT-Trainer Host: Very cool. What inspired you to go down this path and start GPT-Trainer? Tell us more about your background, and is this your first venture, or have you done more in the past? Hunter: This isn't our first venture, to be honest. We've been an AI company for about seven or eight years, back when it was known as natural language processing. Large language models weren't as prominent then. When we first started, we were working on a template-based report generation engine. Obviously, you start with a template, unlike ChatGPT, where you start from scratch or with a single prompt. When ChatGPT came out, we immediately started integrating that technology. We shifted from being focused on academia and research to commercial sector use cases. We were one of the first to develop production-level retrieval-augmented generation (RAG) workflows, which we initially called "context-aware generation," but I guess that didn't really catch on. That was the initial creation of GPT-Trainer. As the platform continued to evolve, we noticed that the complexity for setting up advanced AI agents became difficult to manage through a purely no-code approach. Honestly, there are a lot of no-code platforms out there today you can try. So, we started moving toward a service-led business model. We offer managed hosting, custom implementation, and even AI literacy workshops for strategic partners looking to get the most out of this whole AI revolution. So, we provide both software and service. Addressing Gaps in Generative AI & Customer Success Host: Very nice. What were some of the gaps you saw in how people currently use GenAI? In many ways, it's too broad for a lot of people, but the more custom and tailored you can make it for a specific solution, the better, right? Hunter: It really depends on the use case, Andy. For mid-market to enterprise clients, one concern that always comes up is data security and data safety. For these clients, we provide dedicated hosting. This is because most of our technology is in-house, except for large language models, as we don't develop LLMs directly. So, we put our technology on a dedicated server that we co-manage with the client for a fixed license fee. This is something larger companies really like because they have ownership over their data, they see exactly where it's going, and we can implement deep integrations with their existing business systems instead of having them rely on a public-facing API. That's been pretty helpful. There are plenty of other advantages, but that's the main one. Host: Very nice. Any other specific examples of companies you've helped and what you did for them? Hunter: Of course. If you're looking for example companies, the one I'm always happy to share is Ryden.ai. They are a medical device regulatory compliance assessment company. The typical workflow for a medical device company bringing a product to market involves building and testing the product, documenting all specifications, and then applying for FDA approval. Companies like Ryden provide advice on how best to get that approval. Conventionally, domain experts and consultants review these documents and assess them against regulatory standards and laws. Ryden is taking a creative approach. They're the first company to build an AI-powered assessment system that takes documents from medical device companies and uses AI to evaluate whether they conform to these laws, identifying any gaps along the way. This allows medical device companies to get very fast feedback and iterate to ensure they're well-positioned for FDA approval. We came in to build their entire AI module. In fact, you don't see it in the app because it's fully white-labeled, but GPT-Trainer actually serves as their backend. Our technology runs the entire AI module. We're essentially their core tech. The advantage here is that we're experts in this technology. So, as new models and more powerful LLMs emerged, and as additional capabilities were requested, we continued to improve without them having to maintain an in-house team of AI experts to monitor the market and build things out. They're saving a lot of money and greatly accelerating their product development timeline. That's something really cool we did. I can give you other examples if you're interested. Host: Yes, I'd love to hear about other organizations that might be listening today who could find value in this. It sounds like GenAI isn't for everybody. Sometimes people need really custom solutions, and it depends on a case-by-case basis. GenAI can't be all things to all people. That's really where you guys come in and help people define an exact solution that could genuinely help their business and product. Hunter: Absolutely. The use case with Ryden that I mentioned earlier is very domain-specific and quite niche. However, out-of-the-box, GPT-Trainer most frequently serves organizations looking to automate support. This is one of the biggest use cases for GenAI since it came out. For support automation, you can put together a chatbot on your website, trained on your own data, within 10 minutes of coming onto the platform. That's extremely simple and works right away. This is what a lot of our current clients use. If you need to pull data from your database to personalize the interaction so the AI knows who it's talking with, that's entirely possible. We have built-in webhooks, function calling, and many automations that are ready to go. We just have to configure it. Again, it won't take weeks to develop and deploy, but at best, days. This is something we also help with. It's every business person's dream: under budget and on time. We've never had a project that extended beyond the initial estimated timeline. It's really thanks to our very powerful proprietary platform here at GPT-Trainer. Onboarding and Addressing SMB Barriers Host: Very cool. So, you guys really make it possible for even non-technical teams to deploy this type of technology. Can you perhaps walk us through that process for a team, from start to finish, to get this technology onboarded onto their platform? Hunter: If you're a small or even medium business and you don't have any in-house AI expertise, GPT-Trainer has a no-code builder. It's so simple that you technically don't even need to drag and drop, unlike those flowchart-based builders you might have seen. GPT-Trainer is all natural language and UI playthroughs. You don't even need to configure workflows. Agent instructions are given in English or whatever language you prefer. It takes less than 10 minutes to build a simple chatbot. For more advanced workflow automations, you'll probably need some technical assistance. We can provide tools, on-demand access to external data, or even MCP server integration. You can set up webhooks to trigger events, automate data through https://www.google.com/url?sa=E&source=gmail&q=Make.com, Zapier, or directly via native integrations. You can also connect to and manipulate the agents directly through our API. Our team can help with anything technical. As I mentioned earlier, we're a service-based company, meaning we have a forward-deployed engineering team. So, if customizations are needed, if you need a custom prompt built, or if you need data processing or cleanup before you upload it for training, we can assist with it—of course, on a commission basis. Generally, our price point is a lot cheaper compared to bigger shops like Accenture or Deloitte. We are definitely professional help. Host: What's the most complex situation you've come across in terms of a client and their problem that they needed your help with? It sounds like perhaps the medical company was a good use case there. Hunter: A lot of people come to us from the small to medium side, and they might not be looking to invest in a full-fledged SaaS app hosted on the cloud in a private server—that's a full build-out. We'll take it on, of course, but more commonly, people are looking for workflow automations. Instead of hiring a new resource, can we make the existing team more productive? Can we use AI to automate some of the daily operations? That's where custom build-outs in Make, Zapier, and Workato come in. GPT-Trainer has native integrations or modules within these platforms. So, you can start a workflow within, say, https://www.google.com/url?sa=E&source=gmail&q=Make.com, then add a module, search for GPT-Trainer, and it will show up. It's extremely intuitive to use. You can either build it yourself or have us build things out for you. We're quite familiar with all three of these platforms. If needed, we can use them to connect to your systems, or if you don't have any, we'll build out an entire workflow. It's a lot cheaper and faster than developing an entire SaaS from scratch. AI Versatility, Voice, and Future Trends Host: Very nice. What are some of the biggest barriers for SMBs and smaller teams trying to bring AI into their organization, and how does GPT-Trainer address them? It sounds like you offer a "white glove" service to help them integrate, and you have a no-code approach that makes it very easy for them to integrate, starting with very simple applications. Is there anything else you'd like to address for SMBs who might be on the fence about adopting AI? Hunter: I'd say GPT-Trainer is designed to be extremely versatile. Its versatility means there are many use cases I haven't covered here that people are already using. If you're looking to do lead generation, lead qualification, marketing, or content production, all these things can be done within GPT-Trainer. If you connect to https://www.google.com/url?sa=E&source=gmail&q=Make.com, Zapier, or Workato, you could even do automated email alerts or run a campaign for outbound hyper-personalized newsletters, for example. You can even write a new newsletter for every user in your list if you have separate profile information about each. That's all extremely powerful, and we can support it. With the AI voice project nearing its rollout, I'd say it opens up a brand new dimension of what GPT-Trainer can do. AI voice would enable use cases like direct automation of inbound calls for support, and if necessary, we can do human escalation depending on certain triggers or intent-based monitoring that indicates a user needs to talk to a human agent. For outbound calls, we have background agents constantly monitoring the transcript as the conversation goes on. We can extract data in real time from these transcripts, even while the call is ongoing. Depending on what data is extracted, we can forward it to a CRM and populate it directly. If you're doing real outbound calls, as I mentioned earlier, you can set up number lists—number sets. If you upload, say, a thousand different phone numbers you want to call, the number set you use will essentially rotate the numbers for compliance purposes to reach out to each of those users. So, you essentially have an AI SDR that's much cheaper than hiring a full-time resource for it, and it will sound no different from a real human. Host: Very cool. Any reservations you've seen from business owners about adopting AI? Do a lot of people, for instance, fear some of the quality control issues with having AI on the front lines and not being able to control it? Hunter: That's a perfectly valid concern, Andy. I think there's still a stigma against AI interactions when it comes to customer support and other use cases where AI is rapidly popularizing. It certainly comes down to quality. While AI agents are versatile, if you don't configure them correctly, they will struggle—it's almost like hiring a new employee and not teaching them how to do the job. AI will be inaccurate or just won't know how to answer a question if you don't upload the right training data. The immediate remedy here is to work with experts in the AI domain who truly know what they're doing. Be very clear with your requirements, and then come up with a proof of concept or a baseline product you can start using right away. The optimization and iteration will happen on the go. You could pilot a project to only 10% of your web traffic or 10% of your lead list, whatever you want to connect to, and then scale up as things become more stable and as you gain more confidence. On the flip side, despite the resistance to using AI, you have companies that are already using it. Even Amazon, when you go on Amazon.com, you're talking to Rufus, their AI assistant. The largest companies are already actively using AI agents for operational improvements, and they're pretty good. I've used them for refunds; I've used plenty of AI customer service, and they do a good job in my opinion because engineers and experts spend time optimizing that performance. Nothing's going to work perfectly out of the box. This is where we can help. We are the experts you can trust, and we'll partner with you to build out and customize these solutions, iterating over time. Host: I think you touched on a great point there about why generic large language models aren't really a good fit for every business. The more specific and custom you can get to your use case, the better. So, the more your AI chatbot can actually chat with your customers and understand more about them, it's going to create a more personalized, natural feel compared to someone who might just use a ChatGPT plugin. And we all know how those go. Hunter: Yeah, for sure. Other advantages to using GPT-Trainer, which is LLM vendor-agnostic, is that we have enterprise agreements in place to govern the data usage relationship. Whatever data we send through our platform to these LLM providers won't be used for training or fine-tuning purposes. That's a plus because if you're using it for sensitive needs within your company or if you're exposing sensitive data to the AI, you want to make sure it's not being leaked. So, whatever data you use to train the model on GPT-Trainer will be safe. There's an actual legal basis to protect your data from being utilized by others in training. Host: Can you give an example of where you've done head-to-head comparisons between a generic AI model and one of your chatbots? How did the performance differ? Hunter: It's less about performance here, Andy; it's straight up one or zero. Can this use case be supported through something like ChatGPT versus a custom-made GPT-Trainer bot? I'll give you an example. We partner with a few 911 dispatch centers. Just hearing that name, you know for sure that accuracy is paramount. One of these known dispatch centers is Saginaw County, Michigan Dispatch. Their team used our platform to build several kits. GPT-Trainer enables dispatchers to quickly query policies and procedures, emphasizing the idea that you don't need to know everything, just where to find the answer. Our platform effectively meets that need. It's not possible at all with a generic AI assistant because it's just not trained on that information. You can't talk to ChatGPT and know anything specific about Saginaw County 911 dispatch operations; that information isn't there. It's one or zero: you have it or you don't. The data we use to train GPT-Trainer agents is dispatch-center specific. The Executive Director there actually said they've seen significant benefits with dispatchers using the application to quickly access HR-related information. It reduces the time to get information, which previously took several days. Seasoned dispatchers can benefit by retrieving procedures for high-risk, low-frequency events while they're on call. New dispatchers find value in identifying precise dispatch procedures they might not yet perform. Usually, these dispatch centers have very high churn rates due to the nature of the business—a lot of burnout, high requirements, but low staffing. So, anything to make onboarding easier and make dispatchers' lives easier will help reduce churn and benefit their operations in the long run. That's something we're quite proud of. Host: That's a good point. You don't know what you don't know, and especially for AI, the more personalized it can get, the better for each particular use case. What kind of demand trends are you seeing for AI personalization across different industries? Hunter: I'm definitely seeing a lot of interest in AI voice and automating phone support. I remember seeing a statistic, I can't recall the exact number, but somewhere between 70% and 80% of ongoing support today is still done via phone calls. Web chat is nice, emails are nice, but at the end of the day, the phone is still a very dominant channel for support. So, automating phone support, or doing lead qualification over the phone, call scheduling, call booking over the phone—these are all very real, applicable use cases. This holds true basically across all the industries we've interacted with. Going back to the 911 dispatch example, even being able to identify non-emergencies from the huge volume of calls that come through is a big deal. I was told that, when talking with the New York City Police Department, they're getting 25,000 calls on average every day, and a significant portion are non-emergency. So, if an AI can help filter that, you're getting resources to people who need it faster, rather than having a human process through and waste time. Host: Yeah. People calling in saying, "Oh, my cat is missing," or something. Hunter: Exactly. That's exactly what it is. So they're like, "That's not an emergency!" "It's an emergency to me!" There you go. That's exactly why we're spending a lot of time developing AI voice. I said it'll come out in three to four weeks, but we've been building this for several months now, and we're just starting the announcement because it's about to be ready. The initial MVP will be capable of inbound handling, but we will probably expand it to outbound in another two weeks after that, with the ultimate vision of the system being able to do automatic follow-ups, calendar integrations, phone number rotations, etc. We're partnered with a company that's essentially a competitor of Twilio called White. I know the CEO personally, so we share a good relationship and are working together to build out this tech stack. As a platform, we do support white labeling, and we fully intend for our agency partners who are currently white-labeling our platform to eventually get access to these AI voice features as well. Custom Model Training and Job Outlook Host: Very interesting. Where do you see custom model training heading in the next one to two years? Are we moving towards hyper-specialized AI assistants for every business? Hunter: I don't think hyper-specialized AI agents are going to be the big thing. I think multimodal LLMs (LLMs trained with text, voice, and image data simultaneously) are going to be the holy grail. This, alongside the popularization of MCP standards, is going to give rise to much more powerful and versatile AIs. The advantage of having multimodal LLMs is that you save the latency you currently incur from using separate text-to-speech and speech-to-text modules. You're going to have real-time conversations that sound extremely natural and responsive, basically indistinguishable from you and me talking now. The way they are at the moment, they don't support RAG, and the cost is prohibitive, but I imagine that in a year or two or three, this is going to be much more widespread. For MCP, I mentioned earlier, MCP is an abbreviation for Model Context Protocol, developed by Anthropic, the providers of the Claude LLM series. It allows developers to connect language models to external databases and tools in a very standard way. So, it'll greatly accelerate the development of custom AI agents that integrate with multiple third-party apps and services. These two types in combination, I think, are going to create very fast, very smart, and versatile AI agents in a couple of days. Host: Very cool. How do you think that's going to affect customers and their expectations for AI-powered interactions? Now people can kind of tell, but it sounds like it's going in a direction where people won't even be able to tell the difference between a human and a robot. Hunter: Yeah, there's still definitely a heavy stigma when people don't like interacting with AI. It might be a generational thing, but as models become more powerful, the gap between human-led service versus AI ones is going to get smaller. Not to mention that as industries and countries turn to be more service-led, people costs—labor costs—are going to go up, and that's going to make AI much more appealing. Not to say that AI is going to replace people per se, but it's going to make people much more productive in the long run. On the flip side, the AI's advantage here is 24/7 on-demand availability. This is something you simply cannot have, at least affordably, for human-powered teams. It might even raise the bar for what customer service should be as this becomes more widespread. Host: Oh, I definitely think so. Every business, large and small, is going to have AI agents as part of their core IT framework. What do you think in terms of job displacement in five years from now? How do you think jobs are just going to reallocate, or do you see potentially higher unemployment rates? Hunter: I agree with the popular saying that AI is not going to replace people; AI is going to replace the people that don't know how to work with AI. Think about the time of the dot-com boom, when personal or home computers became popular, like Windows 95 with the Office suite. That was the enterprise standard, replacing much older, more legacy technology. Did it replace some people? I'd say the people who weren't learning how to use these new productivity tools, yes, they did get replaced. But the ones who became more competent with these tools became more efficient; their value increased. If you know how to use productivity tools, then you're actually more in demand in the market. So, yes, there's going to be a shift in distribution, and it's going to change the structure of the economy, but is it going to replace entire sectors? No way. That's definitely not going to happen. Just like every technological revolution that came before, AI is no different. I think all you have to do to protect yourself is learn AI. Host: Smart words from a guy from MIT! So for everybody listening out there, definitely learn these AI tools. You should definitely check out custom GPT or GPT-Trainer as well for custom GPT models. Where can they learn more about your company, Hunter? Hunter: You can go to our website at https://www.google.com/url?sa=E&source=gmail&q=gpt-trainer.com. Again, that's https://www.google.com/url?sa=E&source=gmail&q=gpt-trainer.com. You can sign up for an account there for free; it's free to use and try out. We also have a free account plan. If you're looking for something a little more advanced, if you need to see a demo, book a call. We'll be there to guide you through. Host: And these guys can definitely come up with really custom solutions for you. If you don't know what to do with AI yet, their team, with their service side of the business, can actually help you come up with the idea, formulate it, and assist with the implementation. That's one thing I was really attracted to with what you guys offer. Hunter: As I mentioned earlier, AI voice is just around the corner, so definitely make an account and stay tuned. Even if you aren't a big fan of text-based widgets or text-based bots, AI voice is going to blow your mind. This is going to be a very significant, perhaps the biggest, upgrade we've had since the product's initial rollout, and it's coming in three weeks. So, very exciting stuff! Host: All right, Hunter. Well, everybody who was listening today, visit https://www.google.com/url?sa=E&source=gmail&q=gpt-trainer.com, or you can pick it up today in the NachoNacho B2B SaaS and AI marketplace. We look forward to doing more with your business. We love the product and the service you're working on. Thank you for coming on today, Hunter. Hunter: Thank you, Andy, for hosting me. Thank you.

---

All NachoTuesday interviews: https://nachonacho.com/nachotuesday
About NachoNacho: https://nachonacho.com/llms.txt