InPost scales AI-driven operations to meet next-day delivery goals with Dynatrace

InPost is expanding rapidly across Europe, having more than doubled its network of parcel lockers over the past two years to 61,000, and now processing around 1.4 billion parcels annually. This expansion has increased both the scale and complexity of its operations, further driven by acquisitions and the shift to a cloud-first operating model. To support the next phase of its growth, InPost is working to replicate the high-performance service levels it has in its home market of Poland to the rest of the countries it serves, with ambitions of achieving 98% of parcels being delivered the next day across Europe.

To achieve this goal, InPost is accelerating its use of AI across the business, including logistics forecasting, workforce planning, AI-powered shopping agents, and AI-assisted software development. As these AI capabilities have matured, so has the challenge of monitoring them effectively. With large volumes of data, multiple models, and diverse use cases, the business needed clear insight into how AI systems were performing, how users were interacting with them, and where inefficiencies could arise. Ensuring reliable, observable AI-driven services is now essential to maintain performance, improving operational efficiency, and delivering consistent customer experiences at scale.

Dynatrace was already delivering value across InPost’s cloud-native environment, which is underpinned by a partnership with Google Cloud. Given this success, the company decided to extend its use of Dynatrace to include AI observability, to unlock insights into its expanding network of agents. Grail provides a scalable data lakehouse, ingesting the large volumes of telemetry data being generated by InPost’s AI workloads, which is analyzed by Dynatrace Intelligence and operationalized. With Dynatrace MCP, teams can use natural language to query the data stored in Grail quickly and intuitively. For example, they can ask whether an AI application is performing as expected, analyze user interactions with AI models, or identify abnormal behavior such as unusual token consumption.

“As InPost expands across Europe, our IT teams’ use of Google Cloud and Dynatrace enables us to apply AI at scale without compromising on service quality. The partnership gives us the robust infrastructure and visibility needed to understand model performance issues and drive operational efficiency, so we can meet our next-day delivery goals,” said Mateusz Piasta, Lead Site Reliability Engineer at InPost. “Dynatrace’s support for OpenTelemetry and seamless integration across multiple AI models and vendors means our teams can continue using the tools they are familiar with, making onboarding easier and reducing friction for developers.”

• Greater developer productivity: InPost is using AI to support software development and improve the efficiency of internal teams. Dynatrace provides visibility into internal AI usage, helping teams understand how these tools are being adopted and where they are creating value. This allows InPost to monitor performance, identify opportunities to improve developer productivity and ensure AI-enabled workflows support faster, more reliable service delivery.
• Efficient AIOps: Dynatrace enables teams to investigate and resolve issues faster by combining AI-driven root cause analysis with natural language querying through an MCP. Developers can explore observability data, assess service performance, and identify anomalies without manual queries, improving productivity, and ensuring more reliable services across the platform. In one example, an incident affecting Kubernetes nodes was identified and resolved within just seven minutes, before any end users were impacted.
• Optimized shopping experiences: InPost’s AI shopping agent enables faster browsing and one-click checkout, to encourage better customer experiences and higher checkout rates. However, the agent increases demand on back-end systems and token usage. Dynatrace provides visibility into user interactions, helping teams identify abnormal behavior such as excessive token consumption, so they can refine prompts and guardrails to ensure efficient, cost-effective performance of the shopping agent.
• Smarter forecasting: AI-driven forecasting models help InPost predict parcel volumes and workforce requirements, supporting decisions on courier capacity and staffing levels. Dynatrace ensures these systems operate reliably by providing visibility into their performance and dependencies, helping InPost to maintain efficient operations as it continues to work towards its next-day delivery targets.

“Personally, Dynatrace is a big part of why I’m here and why my team is thriving,” continued Piasta. “The Dynatrace platform has become central to how we work – operating at this scale without it would be difficult. Observability is mission critical in an AI first world and Dynatrace gives us the visibility to reduce the blast radius of issues in our agentic systems early. With fewer teams tied up in incident response, we can focus more of our time on improving services and supporting the next phase of our growth.”