Anthropic has introduced new analytics tools that let companies assess exactly what they are getting for the money they spend on AI. The reports show what employees are working on using the company's tools, what results they achieve, how much it costs, and where problems arise. The new offering is intended to help businesses measure the effectiveness of AI adoption and control rapidly growing spending on it.

For Anthropic, this is also a way to strengthen its position among enterprise customers. Financial data the company shared with investors this week points to the success of its coding tool, Claude Code: it helped the company achieve an operating profit in the second quarter, while net revenue retention from existing customers reached 900%. This means that in the third quarter of 2026, the same customers generated nine times as much revenue for the company as a year earlier, TechCrunch reports.

What companies will get from the new reports

Anthropic's new suite of tools uses internal analytics and employee surveys. Managers will be able to review their team's work over a selected period, evaluate the results, and compare them with the cost of using AI. The reports are also intended to identify weaknesses in workflows and suggest ways to improve efficiency.

By default, Anthropic will not use this data to train its models.

The company is also introducing shared budgets for groups of employees. These will allow teams to jointly manage spending on tokens—the units used by language models to process text and other data, whose volume largely determines the cost of using AI.

In this way, Anthropic is trying to address one of the main challenges of enterprise AI adoption: companies are using AI tools more actively, but cannot always prove that the benefits justify the costs.

AI spending is growing faster than companies can assess it

A new report from Jellyfish, a developer of tools for software development teams, illustrates the scale of the problem. The study is based on data about the use of agentic coding tools by 276,000 users.

According to Jellyfish, Claude Code and other tools powered by Anthropic's models remain the most widely used among the solutions studied. They account for nearly a third of all proposed code changes submitted for review.

Meanwhile, spending on such tools has increased tenfold since the beginning of 2026. They now account for around 2% of total software development costs.

“Even if 2% seems like a small share, a tenfold increase is not something most CFOs had factored into their plans,” Jellyfish chief scientist Nicholas Arcolano told TechCrunch. He said companies are gradually realizing that simply spending money on AI is not enough: they need to optimize costs and achieve measurable results.

Anthropic's new reports are designed to address precisely this task. They are intended to help businesses understand which AI use cases genuinely work and which need to be reconsidered.

Why enterprise customers are in no hurry to switch tools

Anthropic has a strong position in the enterprise market, and its advantage is not just down to the quality of its models. Established workflows, the integration of tools into company infrastructure, and employee habits also play an important role.

The launch of OpenAI's new Astra model over the summer appeared to pose a potential threat to Anthropic, but so far it has not led to a noticeable slowdown in the adoption of its products among enterprise customers.

Nicholas Albarran, head of AI products at Jellyfish, told TechCrunch that OpenAI's Codex became more prominent in the company's data after Astra's release. However, he said social media discussions create the impression of fiercer competition than actual enterprise deployments suggest.

Anthropic has managed to establish a foothold in companies because its tools worked well and users had time to build standardized processes around them. As a result, customers often lack sufficiently compelling reasons to switch to an alternative.

The more deeply an AI tool is embedded in an organization's operations, the more expensive replacing it can be. Even when competing products are broadly similar, they still differ in behavior and settings, including how they handle files such as agent.md, which are used to provide instructions to coding agents.

From analytics to customer support

Technology is not the only factor helping retain enterprise customers. Major AI developers are investing heavily in engineering teams that work directly within client organizations. These specialists, known as forward-deployed engineers (FDEs), help deploy tools, adapt them to existing processes, and achieve initial results.

Anthropic's new reports can be seen as a lighter version of this kind of support: instead of ongoing assistance from specialists, the company provides customers with data that allows them to identify problems and improve workflows on their own.

This is becoming particularly relevant for businesses as AI spending rises rapidly. While companies experiment with agentic tools, token consumption can grow faster than budgets and methods for evaluating effectiveness can be put in place.

Anthropic's task is twofold: to demonstrate that its products deliver measurable value and to make switching to competitors less attractive. If the new tools help customers better control spending and achieve more predictable results, they could strengthen enterprise customers' ties to the company's ecosystem.