Question 1

How much does an AI agent for business cost?

Accepted Answer

It depends on the task. Chatbots start from RUB 15,000, AI assistants from RUB 25,000, AI agents from RUB 35,000, and custom development from RUB 100,000. Current prices are on the Services and Prices page. We confirm the final estimate once we have reviewed your task.

Question 2

How long does AI-agent implementation take?

Accepted Answer

You get an implementation plan within 48 hours of our first meeting: the workflow we would automate, the delivery stages, the integrations, and a preliminary estimate. Implementation then takes up to 14 days from the moment the plan is agreed and we have your materials and test access. We split delivery into stages with weekly demos, so you see progress long before the final launch. If priorities or scope change, we revise the dates in advance and agree on an updated plan.

Question 3

What does your team need from us to start?

Accepted Answer

We start with the project goal, an owner on your side, a description of the current process, examples of real requests, and your knowledge material: policies, catalogues, scripts, documents, or links. Before integration, you provide test access to the CRM, website, and other systems, plus security rules and the categories of personal data involved. If the process has no written specification yet, we can document it during the audit.

Question 4

Can you integrate an AI agent with a CRM, website, Telegram, and WhatsApp Business API?

Accepted Answer

We connect CRMs, website widgets and forms, Telegram Bot API, WhatsApp Business API, email, telephony, calendars, payment services, and internal systems. For a specialised service, we review its API or webhooks and design the data exchange. Before estimating the work, we check platform limits, available API methods, account permissions, and third-party provider fees. This review exposes restrictions before they affect the launch.

Question 5

How does the AI agent handle personal data?

Accepted Answer

We design the architecture around the data categories and your internal policies. The system sends the minimum necessary information to the model, applies role-based access, protects data in transit and at rest, logs actions, and separates environments. We document the model provider, storage location, deletion schedule, and access list in the technical specification and contract. For Russian personal-data projects, we align the design with Federal Law No. 152-FZ; we can also sign an NDA and apply data masking.

Question 6

Can the AI agent make mistakes, and how do you reduce the risk?

Accepted Answer

AI can produce an inaccurate answer, so we connect a verified knowledge base, restrict permitted actions, add validation, and define confidence thresholds. Critical operations follow deterministic rules or require employee approval. We test the system against real scenarios before release, then monitor conversations and refine weak spots after launch. When context is insufficient, the AI agent asks a clarifying question or escalates the conversation.

Question 7

When does the AI agent hand a conversation to a manager?

Accepted Answer

Project rules trigger the handoff when the customer requests it, confidence is low, a conflict arises, a non-standard discount is needed, a payment fails, the topic is sensitive, or the knowledge base lacks an answer. The manager receives the conversation history, contact details, collected parameters, and escalation reason in the CRM or team messenger. The customer can continue without repeating the full request.

Question 8

What happens after launch?

Accepted Answer

From project acceptance and launch, we provide 45 calendar days of support at no extra charge: we monitor stability, fix errors in the agreed logic, review real conversations, refine agreed AI scenarios, update the existing knowledge base, and help your team use the interface. Using the first weeks' data, we retrain the solution within the agreed architecture when answer quality or conversation handoff requires it. At the end of this period, we prepare recommendations for further development and propose the next processes to automate. We respond to a critical technical incident within six business hours, a standard issue within one business day, and estimate a change request within two business days. Ongoing technical support starts from RUB 15,000 as a one-time payment; the scope, communication channel, and release schedule are recorded in the selected SLA.

Question 9

What guarantees does Zolt provide?

Accepted Answer

After project acceptance, a 45-calendar-day warranty applies. During this period, we fix reproducible errors free of charge when they may cause the solution to diverge from the agreed specification or fail an approved checklist: for example, by not creating a lead in the CRM, handing a conversation to a manager incorrectly, or producing an error in configured logic. The warranty covers the code, scenarios, and integrations we develop within the agreed access scope. New features, business-logic changes, additional channels, expanding the knowledge base with new materials, and incidents in CRM, Telegram, WhatsApp, AI providers, or other external services are scoped as separate work. Before a migration, we agree on a backup of available data, a test launch, and a rollback plan. The contract records the scope of work, agreed scenarios, acceptance process, warranty period, communication channels, and support terms. During support, we respond to a report of a critical technical failure within six business hours, share the status, and follow the agreed SLA. Issues on the side of CRM systems, messengers, AI providers, and other external services are outside the warranty, but we help diagnose them and prepare a support request for the provider.

Question 10

Can you improve an AI agent after launch?

Accepted Answer

Yes. We can add channels, roles, reports, integrations, scenarios, and knowledge sources, or change the model and infrastructure as traffic grows. We review logs and feedback first, then build an improvement backlog and release updates in small versions. Each version receives its own estimate and passes testing before publication.

Question 11

Can you migrate an existing bot or automation to Zolt?

Accepted Answer

Yes. We begin by reviewing the current platform, scenarios, integrations, code, and available exports. The audit tells us whether to move the solution directly or rebuild components that the previous platform cannot export. The migration plan covers a test environment, validation of critical scenarios, a parallel launch, and a rollback path. This sequence reduces downtime and lets us compare the new version with the active system.

Question 12

Will our scenarios, knowledge base, and data survive the migration?

Accepted Answer

We can usually preserve conversation logic, documents, tags, user records, and history when the source system provides an export or API and your organisation has the right to transfer the information. Before migration, we validate formats, remove duplicates, map fields, and agree on retention rules. If the platform blocks part of the data, we list the transferable layers and potential gaps before work begins.

FREQUENTLY ASKED QUESTIONS

READY TO STOP
LOSING LEADS?

FREQUENTLY ASKED QUESTIONS

01How much does an AI agent for business cost?

02How long does AI-agent implementation take?

03What does your team need from us to start?

04Can you integrate an AI agent with a CRM, website, Telegram, and WhatsApp Business API?

05How does the AI agent handle personal data?

06Can the AI agent make mistakes, and how do you reduce the risk?

07When does the AI agent hand a conversation to a manager?

08What happens after launch?

09What guarantees does Zolt provide?

10Can you improve an AI agent after launch?

11Can you migrate an existing bot or automation to Zolt?

12Will our scenarios, knowledge base, and data survive the migration?

READY TO STOPREADY TO STOPLOSING LEADS?LOSING LEADS?

READY TO STOP
LOSING LEADS?