After building its own AI marketer and discussing a potential joint venture with Grant Cardone Enterprises, AXL decided to become infrastructure for foundational AI instead.
Every SaaS company seems to be adding an AI assistant. Dmitry Yurchenko believes most of them are solving the wrong problem.
Yurchenko is the co-founder and CEO of AXL, an all-in-one platform for building and operating online education businesses. He founded the company in 2019 with CTO Daniil Musatov. Over the following four years, their team connected websites, courses, CRM, email marketing, payments, booking, webinars, automation and analytics in one platform.
AXL initially responded to generative AI like much of the SaaS industry: by adding AI features inside its existing product. It developed an AI course curator, an AI sales assistant and an AI assistant for its landing page builder. The company then went further and created Mark AI Marketer, a conversational agent that could build websites, send emails and analyze marketing performance.
Mark attracted the attention of Jarrod Glandt, President of Grant Cardone Enterprises, leading to discussions about a potential joint venture. But the experience ultimately convinced AXL that building another proprietary AI agent was the wrong strategy.
Instead, AXL opened its business infrastructure to ChatGPT, Claude and other AI systems through the Model Context Protocol. The company says AXL MCP now exposes more than 1,050 administrative operations across 72 modules, supported by 110 verified recipes for common workflows.
Silicon Valley Journals spoke with Yurchenko about why embedded AI assistants will not save legacy SaaS — and why the next generation of software may be operated almost entirely through conversation.
Q: You say legacy SaaS will not survive simply by adding AI. What exactly is dying?
Dmitry Yurchenko: SaaS infrastructure is not disappearing. What is dying is the traditional way people interact with it.
For two decades, software companies have trained users to work according to the logic of their interfaces. You open a dashboard, find the right section, configure a workflow and move information between different applications.
AI reverses that relationship. Software must now adapt to the user.
People increasingly expect to describe the result they want: create a video, write a program, launch a course, follow up with leads or improve a business. A foundational AI can understand the intention and coordinate the systems needed to execute it.
This reminds me of the transition from software distributed on physical media to cloud SaaS. The underlying capabilities remained, but the delivery model changed completely. Conversational AI will produce a similar shift.
Putting an AI assistant inside a dashboard may make individual tasks faster, but it preserves the old model. Legacy SaaS will survive only if it becomes infrastructure that AI can operate.
Q: Is that what AXL understood when it began adding AI features?
Yurchenko: Not immediately. In 2023, we became convinced that AI would transform the software market, but our first response was similar to everyone else’s.
We created an AI Curator for educational content, an AI Sales Assistant and an AI Assistant inside our landing page builder. Each feature was useful, but when we looked at them strategically, we realized that nothing fundamental had changed. It was still the same SaaS platform with several AI improvements.
So we began building Mark AI Marketer — a universal marketing agent controlled through a single chat. Mark could create and edit websites, prepare and send emails, and work with marketing data based on text instructions.
That felt much closer to the future. But it eventually revealed an even bigger problem.
Q: Mark attracted interest from Grant Cardone Enterprises. What happened during the demonstration?
Yurchenko: AXL joined Embarc Collective, a UBI Global-recognized startup hub in Tampa, in January 2025. In June, we met Jarrod Glandt, President of Grant Cardone Enterprises.
Jarrod did not watch a prepared product presentation. He spoke directly to Mark and asked it to create a website. Mark built it. Jarrod opened the website and registered through it. He then asked Mark to create an email sequence that would be sent according to a schedule.
The entire workflow happened through conversation. Jarrod and his marketing team had not seen anything similar, and we began discussing a potential joint venture.
But Mark was still an MVP built around the 80/20 principle. It covered the relatively small number of marketing actions that solve most everyday tasks for a small business. Jarrod needed much more: sales pipelines, sophisticated automations and access to essentially all of AXL.
In July 2025, we decided to bring Mark deeper inside AXL rather than continue developing it as a separate product.
Q: Why did the potential partnership not move forward?
Yurchenko: From August 2025 through March 2026, we worked on expanding Mark’s access to the platform.
By the time we were ready to demonstrate the broader version, HighLevel had announced its own AI capabilities. Jarrod decided not to continue the partnership discussions. The potential joint venture had been between AXL and his organization; HighLevel was not involved.
It was disappointing at the time. But it forced us to confront a fundamental weakness in our strategy: we were trying to build another AI agent and compete for the user’s attention against ChatGPT, Claude and the other foundational models.
That is an extremely difficult competition for any SaaS company to win.
Q: Was the problem technological or economic?
Yurchenko: Both.
Foundational models are improving faster than an individual SaaS company can realistically match. They are also becoming the place where users already think, work and maintain conversational context.
The economics are equally important. Consumers can access powerful AI products through a predictable monthly subscription. When a SaaS company builds its own agent using enterprise APIs, every token has a metered cost. The company must absorb that cost or pass it on to the customer.
Complex business work requires long context and many actions, so the expense can become significant.
We began asking ourselves: Why should we force customers to pay us for access to intelligence they already have through ChatGPT or Claude? Why should we compete with those systems instead of giving them the infrastructure they need to perform real work?
That question led us to MCP.
Q: How is the MCP approach different from putting an AI assistant inside AXL?
Yurchenko: With MCP, the relationship is reversed.
The user connects AXL to a compatible AI client and signs in with their AXL account. The foundational model provides the intelligence, conversation and understanding of the user. AXL provides the operational infrastructure.
AXL MCP exposes more than 1,050 administrative operations across 72 modules and six business areas: sales and CRM, courses and learning, sites and marketing, communications, automation, and platform analytics.
We have also created 110 verified recipes that help AI agents execute common multi-step workflows correctly.
This is not a chatbot with access to five demonstration features. The AI can work across most of the platform’s operating surface.
Q: What can a business owner ask the AI to do?
Yurchenko: Imagine an expert saying:
“Turn my knowledge into a four-week course. Create the curriculum and lessons, build a landing page, add a lead form, create a $499 offer, configure the checkout, prepare the welcome emails and set up a follow-up automation.”
The AI can translate that request into actions across AXL. It can create the course and lessons, build the page, configure the product, prepare communications, organize CRM records and construct the automation.
