Apollo.io, an AI-powered sales solution for mid-market and growing businesses, set out to improve productivity through internal AI agents. In 2025, the CEO launched a company-wide mandate focused on measurable productivity gains. Apollo’s Head of IT, Ramiro Meyer, joined the core team leading the initiative. “The initial goal was to reduce the time people were spending on their work, and that quickly shifted to productivity. How much more productive could we be with AI?”

To identify opportunities, IT partnered with Business Operations to interview every department, surfacing more than 80 potential agent use cases.

The team began in IT, where they had strong operational context, fast feedback loops, and clear signals for success or failure. To scale safely, Apollo needed an AI orchestration platform that could standardize agent behavior, escalation, and actions across systems while enforcing clear guardrails.

During evaluation, other vendors offered prebuilt agents with limited flexibility. “The glitch there was what happens if I need to move outside the boundaries of the agents that you’ve built? The answer was always the same: ‘Oh, we can build it for you.’ That’s when we realized that wouldn’t work for us. With Tray, we can build it ourselves,” says Meyer.

Apollo selected Tray Merlin Agent Builder because it allowed them to design agents around their existing processes rather than adapt to rigid tools. Tray’s vast library of prebuilt integrations lets agents work across Apollo’s full tech stack, with the option to extend coverage using custom connectors when needed. This gave the IT team control to deploy agents across IT, sales, fraud, and support without relying on vendor services.

The first deployment was an IT Service Management (ITSM) agent in Slack, designed to resolve employee IT issues and deflect tickets. A key design principle was confidence: The agent triages every request and only responds when it can fully or partially resolve the issue. When it can’t, it escalates cleanly to a human.

Tasks such as reviewing Okta or Kandji accounts and provisioning licenses are handled automatically, while requests requiring judgment are converted into tickets. Employees can rate responses directly in Slack, and logs are reviewed to identify gaps and refine prompts.

“The agent only responds when it knows it can fully or partially solve the problem. If it can’t, it escalates. That’s why people trust it,” says Meyer.

With the IT agent establishing trust and proving the model, Apollo began extending the same interaction patterns and guardrails into other parts of the business, including agents such as:

• A virtual sales coach that integrates with Gong to review and evaluate every sales call, replacing manual call reviews.
• A customer insights partner that summarizes the last six months of a customer’s support history, giving support reps faster, more context-rich answers to recurring questions and ongoing issues.

“Tray Merlin Agent Builder gave us something the others couldn’t. It’s not just answering questions. It’s taking action, it’s unlocking accounts, it’s provisioning software. It feels less like a chatbot and more like another member of the team, in fact, the IT agent closed out September 2025 as the number one in ticket resolutions,” says Meyer.

Within months, Apollo’s IT agents were delivering measurable outcomes while maintaining IT’s near-perfect CSAT score of 4.95 out of 5, and proving that speed did not come at the expense of quality. Their CEO’s ambitious plan is already resulting in tangible wins from their first wave of agents, including significant improvements in responsiveness, deflection, and time savings:

• 40% ticket deflection in the ITSM agent’s first two weeks, double the typical first-month benchmark of ~10%.
• 1-second first-response SLA, including a weekend incident where a Director’s device was unlocked in three seconds without on-call staff.
• Faster, context-rich support responses powered by the customer insights agent, with strong early feedback from support teams.

“The moment that really stuck with me was when one of our Directors of Product got locked out of his laptop on a Saturday night. The agent had him back online in seconds. That’s when everyone on the team realized this isn’t just a pilot. It’s real value,” says Meyer.

Looking forward, Apollo sees agents as central to how work gets done across the company. “We’ve only scratched the surface,” says Meyer. “We have over 80 use cases lined up. The more we build, the more people come to us with ideas. Agents aren’t just supporting the business anymore. They’re becoming part of how the business runs.”

Dive deeper into the Apollo story and hear first-hand how they deflect 40% of IT tickets with Tray AI agents.

Watch the webinar.