How Automox leverages AI customer support to streamline their complex IT automation platform

Jonathan White is a Staff Support Engineer at Automox and the internal point person for the company's Duckie deployment. His team handles a wide range of tickets: questions about platform functionality, firewall and antivirus interactions, single sign-on configuration, patch completion, policy best practices, and — most demanding of all — custom worklets, the PowerShell and bash scripts customers write to extend the platform.

Automox first deployed Duckie around summer 2025 on the V1 platform, starting internally and inside Zendesk tickets, then expanding to the help portal widget, and eventually into the Automox console itself. That's where adoption took off. Today, Duckie is achieving a 90–92% ticket deflection rate.

Before Duckie, Automox had spent about a year and a half on an incumbent AI support tool. The contrast stayed with Jonathan.

"It was very stale and stagnant. They weren't really adding any new features or implementing anything that we were asking for. It was a pretty night and day difference in our experience between the Duckie and our previous vendor."

Before moving on, Automox evaluated a number of other tools and chose not to move forward with any of them.

"The other tools were very lackluster. They weren't worth the cost, and were extremely limited in their functionality."

The numbers told the story. The other AI tools struggled to hit 30% resolution while Duckie was able to reach 90% ticket deflection rates on technical tickets. Other incumbent tools lacked tools integration, configurability, and sounded robotic when interacting with customers. They didn't move the needle for Automox's needs: worklet scripting, internal ops in Slack generating delta summaries for complex tickets, and much more.

Duckie excelled at handling tickets and also became an integral part of how the entire company operated.

When Automox first saw Duckie demoed, the difference from the previous tool was immediately obvious.

What particularly stood out was Duckie's ability to go beyond static knowledge and handle complex, multi-layered technical issues.

"Yeah, I would say that number one Duckie is able to handle more complex questions… But it also offers the custom tools feature where you essentially create tools where it connects to API endpoints and they can be external to other platforms that you use as a company and it can pull information from those API calls to help answer questions that can be environmental related or specific platform related. And that adds another layer of the complexity that it can answer... where the tools we were using before were not able to."

Unlike basic chatbots, Duckie operates as a true AI support agent — understanding context, making complex tool calls across multiple systems, and actually solving the intricate problems Automox customers face for the customer.

This wasn't just an incremental improvement — it fundamentally changed what AI customer support could do for Automox. For the first time, they had a tool that could actually understand their product's complexity and answer the nuanced questions their customers were asking. It proved that an AI support agent could actually tackle the complex, technical problems that matter, the kind that typically require escalation to senior engineers or support reps.

Automox embeds Duckie throughout the entire customer journey, leveraging it both externally and internally. For new users, Duckie powers product onboarding and drives adoption through an intuitive widget that guides customers through setup and best practices. When customers need help, Duckie surfaces relevant Worklets and integrates directly with Zendesk to resolve support tickets faster. Behind the scenes, the internal team uses Duckie via Slack to quickly access documentation, troubleshoot issues, and share knowledge — creating a unified AI customer support experience that benefits everyone.

What surprised Jonathan most was Duckie's ability to help customers with complex technical configurations — particularly writing worklets, the custom scripts that drive advanced Automox usage.

A worklet has two pieces: evaluation code that runs every device scan as a criteria check, and remediation code that runs if the criteria isn't met. Duckie understands that structure and can not only troubleshoot implementation issues but help write scripts when customers get stuck.

"You can ask it to give you a worklet to perform a task, like create a registry key if one isn't there, or flag a device if some criteria isn't met, and it will give you both the evaluation code and the remediation code for them to copy and paste into a new policy, into a fresh worklet."

This kind of deep technical troubleshooting and structured output — actually helping customers implement complex configurations — is exactly what the previous AI tool couldn't do.

Duckie isn't only customer-facing. The team uses it inside Zendesk as an AI support agent to search knowledge, draft response frameworks, and — critically — tackle one of support's most time-consuming challenges: long-running tickets.

Jonathan built a runbook that generates a delta summary every three responses. When tickets spiral into 20 to 30 back-and-forth replies over weeks, new engineers no longer waste time parsing the entire thread and instead get instant context.

"If another TSC comes in or if they need to hand that ticket off, it's a lot easier for them to get caught up on where the ticket is. And if the delta summary isn't enough, you can always just ask Duckie to summarize the whole ticket for you."

This is where Duckie's real power shows — it's not just driving ticket deflection from customers. It's transforming how support teams actually operate internally, eliminating the friction that makes complex tickets drag on forever.

Duckie also lives in Automox's internal Slack, where employees who aren't deep in the product — sales, CSMs, SEs — can ask product questions without pulling support into the loop.

This setup transforms how internal knowledge flows across Automox. Previously, support fielded constant pings from sales, CSMs, and SEs needing quick answers. Those questions ate significant support bandwidth.

Now those teams get instant, accurate answers directly from Duckie. A sales rep can verify patch scheduling capabilities mid-demo. A CSM can confirm policy configuration options while on a customer call. An SE can check API endpoint behaviors without waiting for support to dig through documentation.

"It alleviates our cycles where we used to have to manage or answer those questions ourselves in our support channel. If there's something Duckie can't answer, that's the only time we have to intervene."

The impact compounds — when internal teams get faster answers, they close deals faster, resolve customer issues faster, and the entire organization moves with less friction. Support only steps in for the truly novel questions that push beyond Duckie's knowledge, which means those interactions are higher value for everyone involved.

The headline number is ticket deflection, but Jonathan keeps returning to a different shift: what the support team is able to do now that the AI support agent handles the queue.

"We are able to take that time and invest it into these other things, like tooling and improvements that we weren't able to do before, because we were so weighed down by the turn-and-burn life of tickets."

Recently, Jonathan's been building a dashboard that unifies data sources for the CX team, apps for incident tracking, and other tools across the company — work that wasn't possible when the queue was the only priority.

Part of his ongoing role also involves tracking what Duckie can't answer, finding and filling gaps in the knowledge base, so Duckie handles those questions next time.

Beyond AI customer support, Jonathan is most excited about integrating Automox's own native MCP directly into Duckie.

"If Duckie can integrate with our new MCP server, that could be very powerful for the end user — because then it can start doing things like environment configurations for them. I'd really like to explore that and see just how far that can go."

Through this type of deep integration, Duckie would become an actual element of the Automox product offering, not just a standalone chatbot.

Jonathan talks regularly with Duckie's founding team. Feature requests get implemented. The pace of the product matches the pace of his asks.

"Every step along the way I think we've felt very greatly supported. Through our feedback, we've helped shape Duckie also. We've got a lot of the feature requests we've asked for put in. It's been really cool to see that evolution within the product itself and the pace that it's happened at."

We're incredibly proud to be partnering with Automox. They're tackling one of IT's most critical challenges — keeping thousands of distributed endpoints patched and secure across Windows, Mac, and Linux — all from a single cloud-native console. Their worklet system gives IT teams programmable control over their entire infrastructure, turning what used to be manual, error-prone processes into automated policies that just work. As they continue scaling their platform and pushing the boundaries of what endpoint management can do, we're excited to grow alongside them, ensuring their support experience matches the innovation of their product. Companies like Automox, who deeply understand their customers' technical needs and aren't afraid to build sophisticated solutions, are exactly why we built Duckie.