WaiWai CEO and corporate AI course instructor at WowUni, Mik Weisman, discusses why Russia needs an AI agent marketplace, how it will facilitate AI adoption in domestic companies, and what business problems it will help solve. Market experts also share their views: Pavel Khmelinsky, CTO of the commercial division at GK Samolet, Dina Popova, Managing Partner of i-Free Group, Pavel Rykov, founder of rpa.icu, Boris Tseitlin, researcher and machine learning specialist, Ivan Yunitsky, co-founder of VOIC LAB, and Andrey Grabarnik, Head of genAI products at Just AI.
At WowUni, we train corporations on AI adoption across various domains — finance, sales, recruitment, marketing, logistics. From our observations, due to the novelty of solutions and the absence of a unified benchmark, companies often don't understand what exactly to implement. Meanwhile, as TAdviser has noted, agentic AI has become one of the most discussed directions in the corporate sphere. The penetration of agentic AI coding tools Lovable and Cursor in American corporations grew from 50% to 82% in just six months (December to May), according to Business Insider.
For example, GitHub Copilot Reviewer, Cursor BugBot Agents, CodeRabbit, as well as promising tools like Graphite, Cline, and Greptile give companies a serious competitive advantage by accelerating digital product development. In Russia, they are used with restrictions due to their foreign origin, especially in the corporate segment.
The current agentic AI landscape
In our experience, Russian companies face typical problems: fragmented tools, difficulty integrating solutions into established infrastructure, and a lack of transparent metrics for evaluating AI effectiveness. Local specifics compound this: the need for language models to understand Russian business context and strict compliance with personal data protection legislation.
Russia is rightfully renowned for its talented developers. They are already creating many effective AI agents. The emergence of a unified national platform for hosting them — a marketplace — will catalyze explosive growth, similar to what the mobile app industry experienced after the App Store launch. The abundance of AI solutions will quickly translate into quality. At the same time, the marketplace must meet strict requirements for security, privacy, and intellectual property protection.
Key objectives of the AI agent platform
The current situation with AI agents resembles the era of mobile apps before the App Store. Back then, the market was fragmented, and apps were difficult to deploy. Today the same problem exists: an abundance of tools, a whole zoo of solutions, but no unified platform. The marketplace should become a place where you can assemble a team of agents for specific business tasks or personal needs.
A unified platform modeled after mobile app stores will create a competitive and transparent environment. Costs and real value will become clear. Today, the price spread for AI implementation is enormous: from 12 million to 200 million rubles. This causes confusion in corporations: the need to adopt technology is obvious, but deciding which agent is needed and how to integrate it into existing infrastructure is extremely difficult.
Tools and efficiency: developing and evaluating AI agents
The AI agent platform is designed as a centralized guarantor of quality and efficiency measurement.
Both methods have already proven effective. For example, I created an "agent that judges agents" at the ETH Global hackathon in Brussels, and this solution won first place. This way, not only LLMs but also the agents themselves can be evaluated, creating a kind of compass for business.
We are now entering the era of hyper-specific agents. Creating an AI agent for a particular industry requires more than just the ability to code: deep industry expertise is also essential. An AI agent developer for crop harvesting should spend considerable time in fields somewhere in Rostov before advising agricultural firms on how to run their business. The marketplace will allow specialists with expertise in narrow domains to create and easily publish AI agents, monetizing their unique knowledge.
The platform's success depends on two stakeholders: businesses that need measurable results, and developers who need ease and convenience.
For example, at Alfa Bank and GK Samolet, we trained teams to create their own AI agents. Corporations raised the question — how to create a single entry point, create and use AI agents securely. And most importantly, what benchmarks to introduce?
The marketplace should allow running agents in a secure isolated environment within the organization's perimeter. Agent development requires an SDK for both engineers and domain specialists without programming skills. The SDK can also automatically select the appropriate model for a given task, including custom models and improved open source models fine-tuned on Russian texts.
The platform needs to integrate both global and local MCPs — meaning agents and software from email to Jira can be integrated.
Localization as a competitive advantage
When we deployed a sales agent for Ontico conference tickets, the efficiency metric was conversion improvement. It grew 2.21x. For the Kodix development studio, which sold a complex B2B product, the key indicators — contacted leads and scheduled meetings with potential clients — were 2,970 and 12 respectively. We also developed an AI agent for recruiting rare-specialty professionals at a major bank. Key metrics included not only the number of positions filled per month — 27, but also the time saved by HR managers on the recruitment process — 48%.
The developer's task is to handle the complex work of adapting LLMs to local specifics: from communication style to strict compliance with Federal Law 152 (personal data protection).
A good example of creating a successful product without directly competing with leaders is the experience of French LLM Mistral. This not-the-most-powerful model found its niche by fully complying with European regulatory requirements (GDPR), allowing data to be stored within the EU, and being trained on French texts to better understand local context.
What the marketplace should be
For an AI marketplace to become a sought-after and effective tool, it must meet several key requirements.
As a result, the AI agent market will become more transparent and accessible. For businesses, this means quick selection of proven solutions with clear pricing and efficiency metrics, and for developers — the ability to focus on creating specialized agents without spending resources on sales and integration. This effect can trigger mass AI adoption in companies and elevate the domestic market to a new level of maturity.