Agentic AI is redefining what it means to support software in production. Traditional customer success models were designed for products that were “set-and-forget”, but agents don’t work that way. They are iterative by nature, embedded in real workflows, and expected to improve continuously. Driving outcomes requires proactive ownership to expand the workflows an agent can handle and improve how reliably it performs.

At the same time, AI has changed who can contribute to building the product. Success teams are no longer limited to engineering escalations and feature requests; we can directly design, test, and deploy improvements ourselves. That ownership comes with higher expectations: deeper technical fluency in agent architecture, faster iteration cycles, and the ability to translate customer needs into tactical upgrades without requiring an engineering sprint for every change.

This shift is why we’ve renamed our Success team at Decagon to Agent Success.

Building the agent and building the customer relationship

At Decagon, we think of the Agent Success Manager (ASM) role as having two core pillars: building the agent and building the customer relationship.

Building the agent requires a deep, end-to-end understanding of how the system works in production. ASMs develop mastery of the underlying infrastructure so we can guide customers on how to build, optimize, and scale agents effectively. When performance dips or something breaks, we treat these moments as opportunities to improve reliability and expand the agent’s capabilities over time. This work depends on continuous iteration informed by real usage patterns, customer feedback, and evolving business needs.

That technical depth creates space for meaningful problem-solving and creativity. In my second month, I identified a high-volume workflow that was consistently escalated to a customer’s live support team, dragging down NPS. Partnering closely with the customer, I scoped the problem, designed a solution, and implemented a targeted integration that allowed the agent to resolve these requests autonomously. The impact was immediate: near-complete deflection of these inquiries, a significant reduction in support burden, and an increase of CSAT from 2.0 to 4.2.

The second pillar of the role is what turns agent building into durable business impact: long-term customer partnership. The most effective agents aren’t one-size-fits-all, but deeply tailored to each customer’s workflows, data, and operational reality. A core part of the ASM role is helping customers build internal capability around agentic AI by enabling them to customize and improve the agent themselves. That partnership is collaborative and strategic, built on trust across the organization.

A cross-functional hub

Agent Success doesn’t work in isolation. One of the defining traits of an AI-native company is that Success becomes a true cross-functional hub. ASMs work closely with Product, Engineering, and Sales to shape the roadmap, accelerate adoption, and unlock new value.

Because the Agent Success team has the clearest visibility into the day-to-day needs of our customers, we’re uniquely positioned to shape product direction with real-world impact. We see how customers actually use the platform, where agents struggle in production, and which gaps matter most. As some of the heaviest users of the product, ASMs can calibrate customer feedback against hands-on experience, separating one-off requests from recurring patterns and identifying the highest-leverage improvements.

Beyond shaping what gets built, ASMs ensure customers fully realize the value of those improvements. They enable customers on new features, coordinate rollouts, and drive adoption within each customer’s workflows. With features shipping daily, this guidance is essential. Doing it well requires deep feature understanding and the ability to translate new capabilities into clear, practical value, showing customers how each update directly supports their goals and operations.

ASMs are equally critical partners to Engineering. Many customer requests can be resolved directly within the platform, but it’s important to distinguish when self-serve support is sufficient and when deeper engineering involvement is required. When engineering is necessary, ASMs own scoping, triage, and prioritization, ensuring requests are well-defined, contextualized, and clearly tied to customer impact.

Finally, the proximity to customers also makes Agent Success a core driver of our go-to-market motion. ASMs often have the strongest pulse on expansion opportunities because of their deep understanding of customer goals, success criteria, and where additional automation can drive ROI. That same insight underpins renewals and long-term retention. In our experience at Decagon, retention is driven by how consistently the agent delivers outcomes, and the ASM’s job is to make that value explicit.

Defining the Agent Success team

Agent Success was created to be the long-term home for our customers and their agents. With consistent ownership, we can focus on what matters most in an agentic world: continuous optimization, expanding the agent’s scope over time, and enabling customers to become increasingly self-sufficient on the platform.

This evolution is still in its early stages. Decagon hired me as the first ASM in September, and we’re already a team of six and counting. We’re small enough to stay nimble, but large enough to start defining what “Success” should look like for AI agents at scale. For me personally, it’s been a unique opportunity to build from scratch: taking processes from 0 to 1, designing new systems, and shaping the foundation of a team that will grow alongside the category. Watching this function take form in real time has been incredibly rewarding.

If you’re ready to help us define success in the world of agentic AI, come join us in San Francisco or New York!