How Zenjob went live with telli within a month and scaled its candidate screening

About Zenjob

Zenjob is a digital staffing platform based in Berlin, founded in 2015. It places students with companies that need extra staff, short-term or long-term. For short-term needs, students, known as Zenjobbers, take on individual shifts, for example in warehouses, retail or hospitality. They apply in the app, and selection is automatic. For long-term needs, Zenjob places working students through Zenjob Pro, who work at the same company for 3 to 18 months, for example in back office or customer service.

‍

Challenge

Every long-term placement starts with a phone call

Most Zenjob jobs are filled without anyone stepping in. Zenjob Pro is different: before working students are proposed to a company, Zenjob calls them to check that they really fit the role.

Those screening calls sat with a small team. Martin Fricke, Product Manager for Zenjob Pro, says a team can only conduct so many interviews a day, so Zenjob could not scale these placements internally.

The bar for an AI agent was high. Zenjob wanted calls that felt like an easy, natural conversation, without a robotic voice. What convinced the team about telli from the start was how authentic the voice sounded on the phone.

Solution

AI agents built into Zenjob's own systems

Zenjob built its setup with telli in three to four weeks and went live in summer 2025, starting with screening calls for Zenjob Pro. Since then, Zenjob's Pro engineering team, including AI & Automation Engineer Sergen Becerik, has built telli into Zenjob's backend: when someone applies, Zenjob's system filters the application, adds the job-specific questions and hands the contact to telli.

From there, the agents take over the calls:

• Fast first call: telli calls applicants as soon as Zenjob hands them over.
• Questions answered: the agent answers questions about Zenjob or the contract during the call.
• Calls that fit student life: the agent also calls after 6 p.m. and on weekends, when students are not in class.
• Callbacks: most agents also take callbacks, so people who miss a call can usually just call back.

A person at Zenjob then reads a summary of every answer and decides who moves on.

New agents are quick to set up. For a job with very specific requirements, a new agent was ready within an hour and made almost 200 calls in the hour after that. Zenjob needed 20 suitable applicants, meaning calls with a positive outcome. By the end of that second hour, it already had about 40.

“No account manager here could get through that by phone. It's not possible.”

Results

From Zenjob Pro to other parts of the business

After Zenjob showed telli internally, other teams wanted it too. In September 2026, 40 agents worked in six areas: Zenjob Pro, short-term jobs, project recruiting, shift filling, the Zenjobber hotline and calls from companies.

Around April 2026, a large client asked for about 500 students for the following week. Commercial Enablement Manager Rafael Coelho pulled colleagues together, and a group of nine or ten spent a night calling applicants by hand. It was still not enough to brief everyone on where to go, what to bring and what the client required. The next day, Rafael sat down with the Zenjob Pro team, who were already using telli. telli then called all applicants at the same time. Since then, Rafael says, Zenjob can take on orders it could not have handled before.

The results:

• 4,000+ conversations in September 2026, up from about 150 in January
• 88% of the students called are reached, more than one in five of them because they called back
• Under a minute (median) until telli calls a pre-filtered applicant
• Almost 1,800 hours of conversation by October 2026

Zenjob builds much of its technology itself and evaluated building voice AI too.

“Of course we could have developed voice AI ourselves. In today's world there are incredible possibilities, and that's very tempting. But we made a conscious decision to focus on our core competencies and leave voice AI to a partner.”

Zenjob's teams now have more time for Zenjobbers and companies. Even colleagues without an engineering background set up new agents from a job description. Martin uses Charlie, the AI assistant in telli, for this: he types what he wants to change about an agent, and Charlie rebuilds it.

Zenjob also sees great potential in working with telli in the future. Further use cases are already being considered, such as automated feedback calls during long-term placements. This lets both companies keep expanding their collaboration and together open up new ways to provide more efficient and more personal support.