A first-line AI agent: when a bot actually pays for itself, and when it stays a toy

A first-line AI agent pays for itself where managers are drowning in a stream of repetitive first-contact inquiries — and almost nowhere else. Gartner forecasts that over 40% of AI agent projects will be scrapped by the end of 2027 because of rising costs and unclear business value. The fate of any particular bot is decided before launch: is there a real flow of inquiries, has the cost of the routine been calculated, where do the conversations end up.

Below is a breakdown with numbers: how an agent differs from a button bot, why the speed of the first reply is worth real money, how to calculate payback before implementation, what the Klarna story teaches, and which signs tell you the bot will stay a toy.

How is a first-line AI agent different from a button bot?

A button bot walks the customer through a rigid menu and breaks on the first open-ended question. An AI agent understands a question asked in the customer's own words, answers from the company's knowledge base and steers the conversation toward its goal — a qualified lead in the CRM. The customer feels the difference immediately: instead of "press 1 for opening hours", they get an answer to what they actually asked — about prices, timelines and terms.

Both setups work around the clock, but differently: at night a button bot collects phone numbers, while an agent answers questions and qualifies. A request left at two in the morning gets its first substantive answer right away, without waiting for morning.

The first line is only one type: what other kinds of AI agents for business exist and how they are built is covered in a separate article.

It is also worth knowing the limits in advance. Complex negotiations, haggling and non-standard situations stay with people; the agent's job is to take the mass of first-contact inquiries off their hands so managers can sell, and to hand the conversation over to a human in time, along with its history.

Why is the speed of the first reply worth money?

Because a lead goes cold within minutes. The classic lead management study by MIT and InsideSales.com showed that contacting a lead within the first 5 minutes rather than 30 makes you 100 times more likely to reach them and 21 times more likely to qualify them. An AI agent wins exactly that window: it answers within seconds, at two in the morning and on weekends, when the live sales team is physically unavailable.

The other half of the problem is how companies themselves respond. A Harvard Business Review audit covered 2,241 US companies: the average response time to an online request was 42 hours, and 23% of companies never responded at all. Those who responded within an hour qualified leads nearly 7 times more often than the rest.

The studies are old — 2007 and 2011 — and since then their conclusions have only grown sharper: a customer who writes to you in a messenger is writing to two of your competitors at the same time. The one who answers first, and to the point, wins.

Where an AI agent pays for itself: five signs

The economics work out where there is a high volume of first-contact inquiries with repeating questions: real estate, services, retail, e-commerce. If managers answer the same twenty questions day after day while warm customers wait in the same queue, the agent returns the investment within months. Check yourself against the signs below — the more matches, the faster the payback.

• Managers spend a noticeable share of the day on "just a quick question" — questions already answered on the website and in the price list.
• The first response time is measured in hours, and some leads go to competitors in the meantime.
• At night and on weekends inquiries pile up unanswered until Monday morning.
• Incoming requests are not qualified: warm and cold leads are handled the same way, in queue order.
• Managers enter customer data into the CRM by hand — or do not enter it at all.

Our project with a property developer was built on exactly these signs: a stream of inquiries from people who want to "figure things out first", and managers drowning in first-contact work. A first-line bot explains, answers typical questions and warms the customer up, and puts qualified deals into the CRM — managers work only with warm leads.

If you are still deciding what to automate first, start with the overall map — we put it together in “Business automation: where to start”.

How do you calculate the payback of an AI agent before implementing it?

Treat it as an ordinary investment. The cost of handling first contacts manually follows a formula: the number of first-contact inquiries per month × average handling time × the employee's hourly rate + an estimate of the leads lost to slow responses. Compare that sum with the cost of implementation and support, and you get the payback period. A project without a clear payback period is better left unstarted, however fashionable AI may be.

An example of the arithmetic:

1. Count the flow: say, 900 first-contact inquiries a month, 700 of them typical questions about prices, timelines and terms.
2. Estimate the time: at 10 minutes per inquiry, that is about 117 hours of manager time a month.
3. Multiply by the rate: at 800 rubles an hour, that is around 93,000 rubles a month spent purely on answers that are already in the price list.
4. Add the lost leads: how many requests go cold during hours of waiting — in most calculations this is the largest line item.

The market gives an upper reference point: according to a study by SberAnalytics and Sber Business Soft (November 2025, 559 respondents), AI assistants on the first line of support and in call centers cut labor costs by up to 30–40%, and 39% of Russian companies already use AI agents and assistants.

We do this calculation during the first consultation: we go through the flow of inquiries, find where leads and hours are being lost, and set the goal in money terms. Every custom development project of ours starts there.

What does the Klarna case teach?

That an agent wins on scale and loses on quality if you leave it without people. In its first month, the AI assistant of the fintech company Klarna handled 2.3 million conversations and took on two thirds of support chats — a workload equivalent to 700 agents. A little over a year later, the company brought people back onto the line.

The first-month numbers were impressive: resolution time fell from 11 minutes to under 2, repeat inquiries dropped by 25%, and the estimated effect was $40 million on 2024 profit. Then complaints about formulaic answers in complex cases piled up, and in May 2025 Klarna started hiring live agents again.

"Investing in the quality of human support is the way forward for us" — Sebastian Siemiatkowski, CEO of Klarna, from a Bloomberg interview, May 2025.

The lesson for a smaller business is the same as for a fintech giant: the setup that works is hybrid. The agent closes the mass of typical inquiries, a human picks up the complex and sensitive ones, and the handover between them happens without the customer starting over, with the full conversation history. Projects that promise to "replace support entirely" and projects that "add a bot for show" fail with equal regularity.

When does an AI agent stay a toy?

When it is installed for fashion's sake — with no flow of inquiries, no knowledge base and no goal in money. Gartner forecasts that over 40% of agentic AI projects will be scrapped by the end of 2027 because of rising costs, unclear business value and weak risk control. The failure scenarios repeat so consistently that you can spot them before the start.

• There is no flow. At fifty inquiries a month the agent saves a few hours, and the payback calculation from the section above does not add up.
• "Agent washing". Gartner estimates that out of thousands of "AI agent" vendors, roughly 130 have genuine agentic capabilities; the rest are renamed button bots and scripts.
• The knowledge base has no owner. Three months in, the agent is answering from an outdated price list, and customers notice before the owner does.
• Conversations never reach the CRM. The bot had a chat, the history stayed in the messenger — time is saved, revenue is not.
• The goal is stated as "implement AI", with no metric. There is no way to prove the effect, and the project is the first to be cut when budgets shrink.

Every one of these is cured at the design stage — before money is spent on implementation.

Why is an agent without a CRM money down the drain?

Because a conversation without a deal is worth nothing. An agent that talks to a customer and leaves the history in the messenger has saved the manager time — and lost the sale: nobody will see the request, nobody will set its status, no reminder will fire. What pays for itself is the whole chain, where every link has its role.

1. The agent runs the conversation, answers from the knowledge base and collects the data: what is needed, when, what budget.
2. A qualified inquiry automatically becomes a deal in the CRM — with the fields filled in and the full conversation history.
3. The manager picks up where the bot left off, with no "the customer wanted something".
4. The funnel is visible end to end: from the first message in a messenger to payment.

This is also where the answer hides to the question of why so many companies find that "the CRM doesn't work": managers do not fill it in by hand. The agent removes that problem by design — the data reaches the system without a human. Why a CRM stalls and how to fix it is a separate piece, and how to tie the bot, the CRM and accounting into a single loop is covered in the article on integrations.

What do you do about hallucinations and reputational risk?

Restrict the agent to an approved knowledge base and build escalation to a human. An agent plugged into a language model "as is" will sooner or later invent a price or a condition for a customer. A working setup is built differently: answers come only from price lists, internal rules and company descriptions, and for a question outside the base the agent honestly says it will check and hands the conversation to a human.

• Every conversation is logged: you can see where the agent coped and where the knowledge base needs filling in.
• Critical scenarios — complaints, legal questions, price negotiation — are routed to people immediately, with no attempts to "close" them with a bot.
• The knowledge base grows from the logs: every unanswered question becomes a new entry in it.

Tone is part of reputation too: according to Zendesk CX Trends 2025 (a survey of about 10,000 consumers and CX professionals across 22 countries), 64% of consumers are more willing to trust AI agents that communicate in a friendly, human way.

A separate point for companies in Russia is data. Customer personal data ends up in the knowledge base and the conversation logs, so the whole setup has to live on servers in Russia and comply with Federal Law 152-FZ, Russia's personal data protection law. We put that in the contract: hosting in Russia, access segregation, sensitive data never leaving the client's perimeter.

How to launch an AI agent: six steps

Launch takes weeks, and it rests on one resource — your knowledge base. Connecting the agent to Telegram or WhatsApp is technically fast; the long part is assembling and verifying the price lists, terms and standard answers it will use when talking to customers. After that comes a staged rollout, starting with a share of the traffic.

1. Assemble the knowledge base: price lists, terms, internal rules, typical questions and answers. Everything managers repeat to customers every day.
2. Connect the channels your customers actually write in: Telegram, WhatsApp, VK, a website widget, Avito — Russia's largest classifieds marketplace.
3. Link the agent to the CRM: a qualified inquiry immediately becomes a deal with its history attached.
4. Define the escalation rules: which topics and triggers go to a human right away.
5. Launch on part of the traffic and read the logs: where the agent coped, where it got it wrong, what is missing from the base.
6. Keep adding to the base and widening the coverage: with every log review the agent answers more accurately.

Steps 1 and 5 eat up the most calendar time and are the most often underestimated. A knowledge base without an owner goes stale within a quarter — assign someone responsible before launch, not after the first complaint.

In-house or with a contractor?

Look at the cost of a mistake and the depth of the integrations. A prototype agent can be put together in an evening these days, even without a programmer — at that level it is useful to experiment yourself to understand what is possible. The line is crossed where customer data, payments and the CRM link appear: beyond it you need engineering with contractual responsibility.

This "tool versus engineering" line applies to any AI in business, not just first-line agents — we went through it in detail in the article on vibe coding. The logic is the same: a tool gives you speed of experiment, engineering gives you reliability in production.

There is a middle path as well: a standard first-line AI agent as an off-the-shelf solution from the catalog — a fast start with adaptation to your process, and when the business runs into its limits, the setup is extended with development.

If the payback signs from this piece match your situation, tell us about your task: we will put a money figure on your manual first-contact work, design the "agent → CRM → dashboards" chain and see it through to a result under contract, with data kept in Russia and support after handover. The team has 300+ automation and AI projects behind it.

Sources

• MIT / InsideSales.com: Lead Response Management Study, 2007 — the effect of response speed on contact and lead qualification
• Harvard Business Review: The Short Life of Online Sales Leads, March 2011
• Klarna: press release on the AI assistant — 2.3 million conversations and two thirds of support chats in the first month, February 2024
• Bloomberg: Klarna brings live agents back into support, May 2025
• Gartner: over 40% of agentic AI projects will be scrapped by the end of 2027, June 2025
• CNews: a SberAnalytics and Sber Business Soft study on automation and AI in Russian companies, January 2026
• Zendesk: CX Trends 2025 — consumer trust in AI agents