This is not a client case but a study on open data. The “before” numbers come from market research and company publications (sources below). The department team and timeline come from our offer. The “after” values are the metrics we write into the pilot and department contract. The economics were calculated with the site’s calculator for a department at $4,999.
| Industry | Online retail, marketplaces |
| Company size | Model: 3 support operators, a website and two marketplaces |
| Systems | CRM, delivery service, marketplace seller accounts, website chat, social media |
| Department team | Support, marketplace reviews and questions, chat sales, owner dashboard |
| Department timeline | 4–6 weeks |
| Plan and price | Turnkey department, from $4,999; the pilot from $1,499 counts toward it |
The path from an agent to a department
- Pilot: support. The agent answers typical questions about orders, delivery, and returns. Metrics: requests without a reply and the share of questions handled without an operator.
- Department, weeks 1–2. Marketplace reviews and questions within an hour; sales in the website chat and social media: product choice, size, stock, and checkout.
- Department, weeks 3–6. An owner dashboard: orders, returns, ads, and requests by channel every morning. Training for operators.
- Support. Monitoring, fixes, and a metrics report. The pilot counts toward the department price.
Who works in the department
- AI support operator. Order status, delivery, returns, payment: around the clock, with complex cases passed to an operator. Every fifth request to online stores gets no reply (Callibri, 2018)
- Reviews and questions agent. Answers marketplace reviews and questions by the store’s rules. 75% of buyers read reviews, and about 40% expect the seller to reply within a day (Napoleon Reviews, 2024)
- Chat sales agent. Picks products, answers about size and stock, and takes the buyer to checkout in social media and on the website. Stores ignore a quarter of social media messages (Callibri, 2018)
- Owner dashboard. Orders, returns, ad spend, and requests by channel by 8:00. One in four employees spends more than 6 hours a week on reports (hh.ru, 2026)
What happens without an agent: research data
| Requests to online stores that get no reply | every fifth |
| Social media messages that stores ignore | a quarter |
| Buyers who read reviews before buying | 75% |
| Buyers who expect a seller reply within a day | about 40% |
What we fix in the pilot contract
| Requests without a reply | 0: the department answers around the clock |
| Time to answer a marketplace review or question | 1 hour |
| Report for the owner | by 8:00 every day |
Economics by the site calculator
| Department price | from $4,999 |
| Support | from $480 a month |
| Monthly saving by the calculator | $1,776 |
| Department payback by the calculator | about 3 months |
Calculation assumptions: 4 employees on routine work, salary $720 a month, 60% of the work goes to the agents. Salary with taxes ×1.3, support already deducted, payback calculated for a department at $4,999.
Why a department and not one agent
By 2027, a third of agentic AI deployments will combine agents with different skills, and among large Russian companies 23% already run multi-agent projects. For a business, a department in this case study means three things: one customer and one history for all agents, round-the-clock work with no new hires, and one report for the manager.
What stays with people
The department takes the flow and the routine, not the people. Decisions on prices and discounts, negotiations, difficult customers, and disputed cases stay with your staff: each agent hands them to a person together with the history. The manager gives agents tasks the same way as people and sees the result in the morning report.
Sources and fact-checking
- Callibri, 2018: 537,485 requests to 233 online stores; every fifth request got no reply, and a quarter of social media messages were ignored (callibri.ru)
- New Retail, 18 October 2024: a Napoleon Reviews survey; 75% read reviews, and about 40% expect a reply within a day (new-retail.ru)
- Kommersant, 2026, hh.ru data: one in four employees spends more than 6 hours a week on reports (kommersant.ru)
- Gartner, 26 August 2025: by 2027, a third of agentic AI deployments will combine agents with different skills for complex tasks (gartner.com)
- CNews, 26 August 2026: a study by Jet Infosystems and Smart Ranking: 23% of large companies run multi-agent projects (biz.cnews.ru)
Found a mistake? Tell us