Arkadia Events case study: 200 leads from 677 website and Instagram conversations

Arkadia Events focuses primarily on planning and decorating weddings and celebrations. On its website and Instagram it has a pre-sales AI agent called Mila, which handled 677 conversations in six months and created 200 leads. The conversation-to-lead conversion rate is 30%: 33% on Instagram and 22% on the website.

The data comes from the Clover4 client portal: as of October 2026 for Instagram and November 2026 for the website.

Why did Arkadia introduce an AI agent?

Arkadia wanted to improve its sales process. It had three goals:

• The team's time: to free up the time spent on inquiries from unqualified leads.
• Availability: to have 24-hour coverage and a fast reaction to every inquiry.
• Triage: to make lead qualification automated and reliable.

Before the agent, the in-house team replied to website and Instagram messages whenever someone had time. All questions went through the team, from basic information to inquiries that led nowhere.

What role does the AI agent have?

Mila runs the complete pre-sales process on the website and Instagram. It answers questions about services, qualifies the buyer and asks for contact details or books a meeting. The rules and the flow of the conversation are clearly defined in a behavior map, a document that sets out the agent's job and how it runs a conversation. How such a map is built is described in How to build an AI sales agent.

The agent does the job of a setter: the Arkadia team gets a qualified buyer, while the meeting and the agreement stay with people.

How did we approach building the AI agent for Arkadia?

For Mila to deliver results like these, we had to understand each of Arkadia's services well and build a behavior map for each one. It defines what the agent needs to know about that service, which questions it asks, how it qualifies the buyer and when the team steps in.

Arkadia has a set of services, so one agent with one set of rules was not enough. That is why we built a complex multi-agent system: each service has its own agent with its own rules, and orchestration routes the conversation to the right one. The buyer talks to Mila throughout.

How were the results measured?

• Conversation: the buyer sent at least one message.
• Created lead: a qualified buyer who left their details or booked a meeting.
• Conversion rate: the number of created leads divided by the number of conversations.

This rate is different from website conversion, which is calculated across all visitors.

How many conversations became leads?

• Instagram: 454 conversations, 152 created leads, 33% conversion.
• Website: 223 conversations, 48 created leads, 22% conversion.
• Total: 677 conversations, 200 created leads, 30% conversion.

How do the website and Instagram differ?

The same client and the same agent give a different picture on two channels. Instagram has the higher conversion rate, 33% against 22%. The website has a larger share of buyers who send three or more messages and show interest: 35% against 10%. Conversations that stop after one or two questions make up 42% on the website and 50% on Instagram.

This data does not show why. It shows that an agent's result should be judged separately for each channel.

What happens to conversations that did not become leads?

477 conversations did not become leads. The portal sorts them into three groups:

• 34 conversations: the buyer said they would get in touch (31 on Instagram, 3 on the website).
• 123 conversations: three or more messages and shown interest (45 on Instagram, 78 on the website). These are informed potential buyers who may later become leads through another channel. Such leads are not included in the conversion rates.
• 320 conversations: one or two questions, with no follow-up (226 on Instagram, 94 on the website).

How many conversations come in outside working hours?

Of 454 Instagram conversations, 62% came in outside working hours, counting working hours as Monday to Friday, 9 am to 5 pm. 24-hour coverage was one of the goals of introducing the agent.

What changed in Arkadia's sales?

The biggest change is a shorter path from the first inquiry to a meeting, and that meeting is with a qualified potential client. The team no longer spends time on basic qualification and on conversations with leads who are not the right fit. It goes into the meeting with clear context: who the client is, what interests them and what they need. A large part of the communication that used to happen during the meeting is now automated.

The Arkadia team now relies on Mila and counts on it to answer reliably and generate leads. Gabrijela Brujić, General Manager of Arkadia, describes it this way:

"Working with Clover4 is exceptional. The Clover4 team made a thorough effort to understand our business and create an AI agent that now runs the complete pre-sales process on our website and Instagram. The result: significant time saved for the entire sales team and a constant flow of qualified inquiries."

Gabrijela Brujić
General Manager, Arkadia Events

Time savings can only be estimated. The estimate rests on two assumptions: that without the agent, half of the 477 buyers without a lead would have needed a reply from the team, and that handling one such inquiry takes 20 minutes. On those assumptions, the saving is about 80 hours, or about 10 working days of one salesperson over six months.

What does this mean for other companies?

Companies face a similar challenge when inquiries arrive on several channels and the team replies to each whenever it has time. This also applies when a company has several services, since each can have its own rules within the same agent. Costs and choosing a solution are covered in AI agent for your website.

To estimate the potential in your company, four figures are needed: the current conversion rate, the share of inquiries outside working hours, the time to first response, and the time the team spends on buyers who are not ready. If you would like to see how this could work for you, book a free demo.