Dash0 Raises $110M Series B at $1B Valuation

Last updated: July 5, 2026

About Log AI

Dash0 uses AI to identify patterns in your log messages and extract type-safe, named attributes for use in queries, filters, grouping, and triage.

Even today, the majority of logs are typically not structured very well, or not structured at all. They are written to files as plain text, and any system that needs to work with them has to parse that text — typically through a painful web of regular expressions that nobody wants to maintain.

This creates two key problems:

  • Missing severity. Many log records don't include a severity field (e.g., error, warning, info). The severity information is buried somewhere in the log body, making it invisible to any tool that relies on structured fields. Without it, you can't filter for errors, you can't alert on them, and you don't even know they exist.
  • No structured attributes. Important data — IDs, names, status codes, parameters — is embedded in free-text log messages. You can't filter by it, you can't group by it, and you can't run any meaningful analysis over it.

Log AI in Dash0 solves both of these problems automatically, without requiring you to define a single regular expression.

What Log AI Does

Log AI eliminates the need to maintain regular expressions for parsing unstructured logs. It automatically infers severity when it's missing, mines meaningful patterns from your log data, and extracts named attributes from dynamic log segments — all without any configuration. If you've ever had to maintain regex-based log parsing, Log AI handles it for you.

Note
  • There is nothing to configure. Log AI is always running. As soon as you send logs to Dash0, Log AI begins processing them.
  • Log AI requires a warm-up period to build up enough data for accurate pattern mining. If you send a single log record and immediately check for patterns, they won't be there yet. Pattern mining needs a representative sample of log data to work with. This can take a few hours depending on log volume.
  • Severity inference also takes some time to become fully active, and is not 100% guaranteed for every record — but it covers the vast majority of cases.

Log AI has two core capabilities that work together: severity inference and log pattern mining.

Severity Inference

When Dash0 receives a log record that has no known severity, Log AI analyzes the content of the log body and infers what the severity should be.

For example, a log line that clearly describes an error condition will be classified as an error — even if the original log record didn't include a severity field.

Log Sidebar showing severity field with AI inference indicator noting the value was determined by Log AI

Without Log AI, that record would show up as "unknown" severity, and you'd never find it when filtering for errors.

Log Pattern Mining

Log AI analyzes your incoming logs and mines patterns from them. But it goes beyond simple pattern detection — it also extracts the dynamic segments within each pattern and assigns them sensible, meaningful names.

For example, given log messages like these:

Log messages showing Product Found entries with varying product IDs and names

Log AI will:

  1. Identify the log pattern: Product Found - ID: *, Name: *
  2. Extract the dynamic segments and name them meaningfully, such as product_id, product_name, rather than using generic labels, such as parameter_1 or dynamic_segment_1.

You can see this in the Log Explorer's Patterns tab, where dash0.log.pattern is used:

Log Explorer Patterns tab showing dash0.log.pattern attribute with extracted dynamic segments like product_id and product_name

Once these dynamic parts have names, they become real attributes you can filter, group, and run pattern analysis on.

Tip

Log patterns are scoped to specific workloads, which could be a Kubernetes Deployment, a Vercel Project, or a Service. For example, a pattern mined from an ad service is distinct from a pattern mined from a product catalog service — they each have their own set of patterns. Dash0 automatically identifies and groups workloads so that patterns remain relevant and accurate within their context.

Further Reading