Avito procures voice bots and intelligent agents amid investments in neural networks

Avito has begun seeking suppliers of ready-made voice robots and licenses for intelligent agents. The procurement shows that the company wants to do more than simply test generative AI in a laboratory—it aims to deploy it in real customer processes right now.

What Avito is Buying

Based on the procurement description, the company needs two classes of solutions. The first consists of ready-made voice robots that can be quickly connected to incoming and outgoing calls. The second involves licenses to use intelligent agents built on large language models. This is an important detail: we're talking not just about classical voice menus or scripted bots, but about more flexible systems that can understand context, maintain dialogue, and respond to non-standard questions without a rigidly predefined decision tree.

For a large classifieds platform, this is a logical next step. Avito handles a massive flow of communications between sellers, buyers, customer support, and internal teams. In such an environment, automating calls and text scenarios delivers quick operational gains. Ready-made solutions allow the company not to wait while building the entire technology from scratch internally, but to start with already working products and test where they deliver the best economics, response quality, and processing speed.

Why the Company Needs This

The main practical benefit of such procurement is freeing people from repetitive tasks. If a voice robot and an agent based on a language model handle routine inquiries, employees can shift to more complex cases: disputed transactions, escalations, anti-fraud measures, and non-standard user requests. For a platform at Avito's scale, even a small reduction in time per contact translates into significant savings over the long run. Plus, the company gets more predictable service during peak load hours.

The most obvious use cases are:

• initial call reception and classification
• answers to typical questions about listings, delivery, and statuses
• routing the customer to the appropriate specialist
• capturing conversation structure for subsequent quality analysis
• automatic summary of dialogue for CRM or internal systems

But it's not just about reducing call center load. Intelligent agents become an interface to data and business processes. If they're connected to internal services, they can do more than just talk to the customer—they can actually execute actions: check listing status, suggest platform rules, flag an issue, create a ticket, or suggest next steps. For users, this looks like faster and more complete service; for the company, it's a way to reduce losses at each support stage.

Testing Ground for Its Own Model

The most interesting layer of this story involves the company's own technological strategy. Experts suggest that purchased solutions can be used not only as an applied tool but also as a testing ground for developing its own sovereign architecture. In other words, Avito can observe which scenarios external agents handle best, where they fail, which dialogues require additional training, and which business functions have the highest demand within the platform. This is no longer a demo—it's gathering battle-tested experience under real load.

This is especially important given the multibillion-dollar investments in its own neural network development. When a company first puts ready-made agents into production, it gets not abstract hypotheses but live data: actual calls, actual customer objections, actual failure points, and actual performance metrics. This approach helps it determine requirements for its own model faster, understand which components are worth building independently and which are more cost-effective to purchase from the market for the time being.

In essence, the procurement can become an intermediate stage between experiments and a vertically integrated AI platform. First, the company buys ready-made tools, then accumulates operational experience with them, and afterward transfers the best practices to its own stack. For a major player, this is a pragmatic path: don't wait for perfect internal technology, but instead automate processes in parallel and accumulate data for the next iterations. Likely, such solutions will first launch on limited scenarios, then gradually expand their scope of use.

What This Means

The market is converging on a model where large digital platforms simultaneously buy external AI agents and build their own. For Avito, this is a way to accelerate customer service automation today and reduce dependence on third-party technologies tomorrow. For the industry as a whole, it's a signal: AI employees are no longer viewed as a showcase for innovation but as an operational layer of business, measured by quality, speed, economics, and level of autonomy in practice.