Habr: Yandex Wordstat data shows search demand for "AI agents" trails bots and services

A column on Habr makes a blunt claim: the market has almost no need for "AI agents" as a standalone product. The author attributes this to where real demand lies — not in the trendy term, but in clear-cut tasks: bots, online stores, payments, and practical automation.

What Wordstat showed

The author's main evidence comes from Yandex Wordstat. According to his data, the query "ии агент" ("AI agent") draws 38 642 impressions per month, but the breakdown is dominated by phrasings like "creating an AI agent," "how to create an AI agent," and "AI agents course." His takeaway: the term mostly interests developers and people looking to break into a new niche, not companies looking for a concrete solution to a business problem.

For comparison, he cites more practical queries. "Бот макс" and "бот max" together account for over 118 thousand searches, alongside mass-market needs like "how to make a website," "how to make an online store," and "how to set up payments." Another telling example is OpenClaw: a single, specific open-source tool that people search for more often than the broad concept of an AI agent. The author's logic is simple: users vote with their searches for a ready-made result, not for an architectural label.

Where the term breaks down

The column separates two meanings of the word "agent." For engineers, it's a perfectly normal technical term: a system with a planner, a set of tools, memory, and a decision-making loop. But in marketing, the author writes, the concept quickly turned into the promise of a universal "magic thing" that will supposedly sort out any process on its own. That shift is where the gap opens between expectations and what can actually be sold to a business.

"Supply and demand live in parallel universes."

From this gap, the author draws several practical rules for the market:

• Users don't buy an "agent" — they buy a bot, a calculation, a report, or an online store.
• The name of the technology barely affects demand if the end benefit isn't clear.
• Even a popular tool loses out if launching it requires Docker, API keys, and a VPN.
• The term can be useful inside a team, but it works poorly as an external product promise.

Why pilots stall

The harshest part of the piece covers enterprise deployments. The author describes a familiar scenario: companies are already budgeting for "AI agents," assembling teams, and launching tenders, yet they often can't clearly say where exactly such a tool would fit. Many of the tasks now being packaged as agentic have long been handled by a combination of CRM, webhooks, SQL templates, or an ordinary chatbot.

The result is that businesses aren't being sold a solution to a problem, but a new label slapped on top of the same old process chaos. Anything that can be automated with a regular pipeline doesn't get better just because it's wrapped in an LLM.

The article then draws a parallel with previous hype waves — blockchain, big data, and dot-com. In the author's estimate, most AI pilots at Russian companies never make it to production, and for the same reasons: there's no success metric, no process owner, the demo works only on test data, and the task itself is framed as "implement AI." As before, the projects that survive are the ones that address a narrow, practical pain point and quickly show economic impact.

What it means

The article's point isn't that agentic systems are useless, but that the market rarely buys technology in its pure form. If AI agents do go mainstream, users won't see an "agent" — they'll see a clear service: a booking bot, a cost-estimate check, an online store builder, or automated calculations. For startups and enterprise teams, the takeaway is the same: sell a measurable result, not a trendy term.