March 2026 Summaries

An analysis of 63 recent company earnings calls revealed that while many companies, such as HubSpot, Google, and Spotify, report high percentages of AI-generated code, none have tied these figures to concrete engineering outcomes like deployment frequency or incident resolution times. Despite widespread AI adoption in coding, with reports of increased individual productivity, organizations have not seen a corresponding increase in overall delivery velocity due to bottlenecks in code reviewing and integration processes. Companies like Pfizer and Amazon were able to connect AI usage to significant outcomes by having clear metrics for comparison, but for engineering, isolating AI's impact remains challenging due to the variability in project scopes and complexities. The report suggests that to truly demonstrate the value of AI in engineering, companies need to establish baseline metrics and clearly connect AI adoption to specific engineering improvements.

Defining a "service" within a software catalog is crucial for clarity and efficiency, particularly as organizations grow and adopt agentic workflows. A service is considered a deployable unit, owned by a single team, with a stable interface that can be independently managed without requiring coordination with other teams. This definition helps eliminate ambiguity that can lead to operational friction, such as unclear incident management and inefficient debugging. Key properties of a service include its owning team, lifecycle stage, criticality tier, deployment target, and dependencies, which are essential for understanding its role and impact within the broader system. With clearly defined services, companies can improve incident response times and maintain more effective engineering practices, as AI agents rely on precise data to function optimally. Establishing a company-wide agreement on what constitutes a service enhances collaboration, reduces downtime, and ensures that both humans and AI agents can accurately navigate the software ecosystem.

Checkmarx, a software and cybersecurity company, faced challenges with inefficient developer environments which led to high costs and operational bottlenecks. To address these issues, they implemented an internal developer portal using Port, which provided developers with autonomy and streamlined the creation and management of development environments. This approach enabled developers to set up environments quickly with self-service actions, ensuring security and cost-effectiveness, while significantly reducing infrastructure costs by standardizing processes. The portal's automation features allowed for better management of resources and reduced the reliance on DevOps teams, leading to a more consistent and productive workflow. Additionally, the move towards a GitOps model and plans for a comprehensive software catalog and scorecards reflect Checkmarx's commitment to continuous improvement and maintaining high standards. The high adoption rate of the portal underscores its value in enhancing productivity and efficiency within the organization.

During a recent incident at Port, three separate teams were alerted to the same issue, resulting in redundant efforts and delayed resolution. The incident involved a customer generating 1.7 million automation runs in 90 minutes, causing Kafka offset lag and triggering multiple PagerDuty alerts. The teams worked in isolation, unaware of each other's activities, and it took 77 minutes to identify the common root cause. In response, Port is developing an autonomous incident resolution system starting with a triage agent that utilizes a Context Lake to gather comprehensive service and deployment data, enabling swift and informed responses. This agent aims to streamline incident management by correlating alerts, suggesting fixes with human approval, and executing solutions with built-in safeguards. The initiative seeks to enhance efficiency and documentation by automatically compiling post-mortems from existing data, thus moving towards a future of autonomous incident management with controlled human oversight.

A recent incident at Port involved three teams investigating the same issue independently, leading to inefficient use of time and resources. The incident was triggered by a customer creating 1.7 million automation runs in 90 minutes, causing Kafka offset lag and multiple PagerDuty alerts. Each team attempted to resolve the issue separately through service restarts, but it was only after 43 minutes that one team connected the alerts as related, leading to the realization that the root cause was shared. To prevent such incidents in the future, Port has developed a triage agent equipped with a Context Lake and a triage skill file to handle incident resolution more effectively. The Context Lake centralizes necessary information about services, deployments, and dependencies, while the triage skill file guides the agent's actions. This agent aims to assess incidents, suggest remediation, and execute fixes with human approval, ensuring faster and more coordinated responses. Port's approach emphasizes autonomous incident resolution with controlled oversight, aiming to efficiently handle incidents while maintaining a full audit trail and quantifying the agent's impact on incident management.

Port's "Support AI Hub," built on the Agentic Engineering Platform (AEP), transforms support ticket data into actionable insights across organizations, enhancing collaboration between teams such as Research & Development and Customer Success. By utilizing Port's Context Lake, the hub's specialized AI agents streamline processes like ticket summarization, escalation, and proactive health checks, while also offering trend analysis and usage metrics to improve efficiency and decision-making. These agents facilitate faster ticket resolution, effective communication, and data-driven problem-solving, enabling a comprehensive understanding of customer experiences and fostering informed decision-making across the organization. As a result, the hub not only increases operational efficiency but also empowers organizations to unlock valuable insights from everyday support interactions.

Engineering intelligence tools provide valuable insights into team performance by identifying bottlenecks and inefficiencies, such as prolonged code review cycles and time spent on manual tasks. However, these tools often fall short in enabling actionable outcomes, as they focus on measurement rather than providing mechanisms to implement solutions. Port addresses this gap by offering a platform that integrates both measurement and action, allowing teams to efficiently address issues without the need for multiple tools or extensive ticket systems. By connecting insights to actions, such as rolling back a failed deployment or triggering a fix directly from the platform, Port transforms static dashboards into dynamic solutions. This approach not only optimizes workflow but also prepares teams for an agentic future where AI can actively resolve problems and enforce standards, ensuring engineering intelligence translates into tangible improvements in productivity and team satisfaction.