Anthropic has disclosed that Claude now resolves 95% of internal analytics queries across the company, according to a report from InfoQ. The figure is striking for a company that builds and trains frontier AI models, but it fits a broader pattern of Anthropic deploying its own tools aggressively in-house before pushing them to external customers.
The disclosure is one of several recent signals that Anthropic is leaning harder on Claude for operational work, not just research. Earlier this year the company reported that Claude writes the majority of its own production code, a trend that appears to be extending into business intelligence and data analysis functions.
What the 95% Figure Actually Means
Analytics queries inside a technology company can range from simple dashboards to complex, multi-step data investigations. Handling 95% of those through an AI system suggests Claude is doing more than surface-level lookups. It implies the model is interpreting business context, writing and running queries against live data sources, and returning results that analysts consider accurate enough to act on.
Key Facts
- Claude now handles 95% of Anthropic's internal analytics queries, per an InfoQ report.
- The figure covers day-to-day data requests inside the company, not just automated pipelines.
- Anthropic has previously reported Claude writes 80% of its production code, suggesting wide internal adoption.
- The trend reflects a broader move toward AI-assisted knowledge work at AI-native companies.
- No external audit of the 95% claim has been published; the number comes from Anthropic internally.
It is worth noting that the 95% figure is self-reported. Anthropic has not published a methodology or third-party audit behind that number. That does not make it implausible, but readers should treat it as an internal benchmark rather than a formally verified metric. Companies have strong incentives to highlight high adoption numbers for their own products.
The pattern of using Claude internally at scale appears to be a deliberate strategy, giving Anthropic direct feedback loops on where the model succeeds and where it struggles before those gaps surface in customer deployments.InfoQ analysis of Anthropic internal usage data
A Company Using Its Own Product at Scale
This is not an isolated data point. In a separate disclosure covered here, Anthropic noted Claude now writes 80% of its production code, a figure that raised eyebrows when it was first reported. Together, these numbers sketch a picture of a company where AI assistance has become the default mode for technical and analytical work, not an optional add-on.
There is a practical logic to this approach. When Anthropic's own engineers and analysts use Claude daily for real work, the company gets ground-truth data on failure modes, latency issues, and edge cases that synthetic benchmarks miss. That feedback presumably feeds back into model training and product decisions. It also makes for compelling marketing, though the line between genuine internal use and demonstration-ready statistics can be thin.
The analytics use case is particularly interesting because it requires the model to interact with structured data, understand domain-specific terminology, and produce outputs precise enough to inform business decisions. Getting that right at 95% coverage implies either a very capable model, a well-engineered tooling layer around it, or both. Anthropic has confirmed AI is now building AI systems internally, so it is plausible that the analytics infrastructure itself was partly designed with Claude's assistance.
Implications for Enterprise Customers
For companies evaluating Claude for their own analytics workflows, the internal adoption numbers offer a reference point, though context matters. Anthropic's data environment, team size, and query types may differ significantly from a typical enterprise deployment. Still, a company willing to stake its own operational decisions on Claude's outputs is making a meaningful implicit claim about reliability.
The broader trend of AI-native companies eating their own cooking is accelerating. As more organizations report high internal adoption rates, the conversation will shift from whether AI can handle knowledge work to how companies structure teams, workflows, and accountability around AI-generated outputs. Anthropic's 95% figure is a data point in that larger story, one worth watching as independent verification and longer-term performance data eventually emerge.
