Operations | Monitoring | ITSM | DevOps | Cloud

Many revenue-critical interactions, such as ecommerce checkouts and customer portals, run on Shopify and Salesforce. But engineering teams have less control over the frontend runtime on these platforms, and this lack of control can make user monitoring difficult to implement and maintain. These monitoring limitations can leave gaps in visibility across important parts of the user journey.

TypeSafe AI released Jev in September 2026 to do one thing: make decisions. Give it a state (a string or a JSON object) plus a set of typed questions, and it returns typed answers with probabilities. It never explains itself, and that constraint is the whole idea. Evaluation pipelines have spent the last two years asking text generators for yes/no verdicts, wrapping the reply in a JSON schema, and paying generation prices for what amounts to a single bit.

Datadog Real User Monitoring (RUM) SDK settings live in your application code, so changing how the SDK collects RUM data has traditionally required shipping a new application version. These configuration changes can include adjusting sampling rates, enabling Session Replay, or changing which events the SDK collects. For mobile teams, this means that updates often sit in app store review for days or weeks before users start adopting the new version. Full user adoption can take weeks or months longer.

Logs are easiest to investigate when the values that matter are already captured as attributes. When those values are buried in a log message, even a straightforward question such as filtering on a status code, graphing the duration of a request, or following a unique transaction across a set of logs requires writing complex regular expressions or Grok patterns.

Observability investigations rarely follow a straight line. A latency question might cause an AI agent to start with a metric, pivot into traces, compare a deployment window, and finish by reducing thousands of logs to a few patterns. Each individual query is easy, but propagating context throughout an entire investigation can be tricky and expensive.

Collecting high-quality user feedback on agents, like from thumbs-up or thumbs-down buttons, is an important part of agent development. User feedback is needed for everything from basic gut checks on whether your agents are behaving well to planning and creating robust eval sets. It’s a critical part of Datadog’s Agent Observability, which provides explicit end-user feedback features for collecting and analyzing it.

A funnel can tell you that 40% of users dropped off between checkout and payment. What it can’t tell you is what those users did instead, such as return to an earlier form field, leave the flow for a support page, encounter an error, or take another route entirely. Because actions and views between funnel steps don’t affect the conversion calculation, two very different experiences can produce the same funnel result.

When an issue in production triggers an alert, the people responding to it are often working in Slack while the evidence they need is elsewhere. Responders need to move between conversations, telemetry data, source code, and incident tooling as they form hypotheses, coordinate actions, and keep stakeholders informed. That context switching can slow down a time-sensitive investigation and make updates harder to follow.

A platform engineering team can design robust Golden Paths for agent use yet still be unprepared for what happens after adoption. An agent may authenticate properly, call the correct tools, adhere to approval gates, and complete tasks without incident, but new operational risks arise once multiple teams begin running agents continuously and in parallel. We’ve encountered these risks firsthand at Datadog.

Distributed AI training workloads impose complex scheduling requirements that Kubernetes’s built-in scheduler can’t meet. Kubernetes schedules pods individually and independently, but distributed training introduces two requirements that break this model: Pods must land on hardware with the right inter-GPU bandwidth, and all pods must be scheduled simultaneously. If either requirement goes unmet, training stalls or runs far below the hardware’s potential.