Anthropic Launches Claude Enterprise Cost Tools — But the ROI Challenge Persists

Last week, when Anthropic disclosed its financial results for the second quarter, what drew the most attention was the 900% revenue retention number: spending from the same customers in Q3 2025 grew nine times compared to Q3 2026. It's growth that no CFO planned for — and that many companies still can't justify to their boards.

Yesterday (October 28, 2026), the company took the next logical step: it made generally available a suite of financial reporting and control tools for Claude Enterprise. Three features — smart reports, in-app surveys, and pooled group budgets — promise to give IT administrators granular visibility into how teams are using Claude, what it costs, and whether it's worth it.

What the tools actually do

Smart reports were released in beta on September 10 and are now available to all Enterprise customers. The workflow is straightforward: an administrator selects a team, a period (up to 28 days), and the Claude applications (chat, Claude Code, Claude Cowork). The system then analyzes samples of team transcripts, categorizes tasks, calculates costs at list price, and flags where sessions encountered friction — connectors that were never set up, outputs that didn't meet the request, rework loops.

Anthropic's stated intent is clear: help companies that are already spending millions on AI understand where that money is going. The smart reports are explicitly not designed — and the company emphasizes this repeatedly — for evaluating individual performance or making employment decisions. They are adoption and investment planning tools.

In-app surveys let administrators send questionnaires directly to Claude users, capturing feedback about how the tool is enabling new work. Responses appear in real time and can be exported to spreadsheets.

Pooled group budgets solve a common practical problem: a highly active user hits their individual monthly limit mid-month while other team members leave budget unspent. With a group-level shared budget, spending flows where the work happens — and administrators can set limits based on the previous month's actual spend.

The math behind the problem

To contextualize the urgency of these tools, the numbers are compelling. According to a Jellyfish report covering 276,000 customers, spending on Claude Code and other Anthropic-based tools grew 10x since the start of 2026, now representing about 2% of total engineering cost. "Even if you say 2% is still small, that 10x growth is something most CFOs didn't plan for," Nicholas Arcolano, chief scientist at Jellyfish, told TechCrunch.

The corporate picture is even bleaker. A KPMG survey of 2,131 senior leaders across 20 countries found that 89% of executives say their companies benefit from AI — but only 12% consistently compare those benefits to cost. For companies already generating some value, that number rises to 48%. In other words: most companies don't know how much they're spending or what they're buying.

Federal Reserve research also identifies this gap: perceived productivity gains frequently outrun measured improvements. BCG found that nearly half of surveyed companies were generating value from AI, even with spending reaching 3.3% of revenue.

The IPO context

The timing of these tools is no accident. Anthropic filed its S1 for its IPO on October 26, and the numbers presented to the SEC showed operating profit in the second quarter. But to convince investors it can sustain a billion-dollar valuation as a public company, Anthropic needs to demonstrate its customers are engaged, not just subscribed — and that they're seeing returns on investment.

The company also has defensive motivation. According to Ramp's AI spend index, Anthropic's share of total token spending rose from 51.3% on September 25 to 60.5% on October 23. Competitors like OpenAI with Astra and Google with Gemini aren't standing still — but so far, switching platforms for Claude remains low. "It took a long time for us to see Codex really pop in our data at all, and it came up with Astra," said Nikolas Albarran, AI product lead at Jellyfish, noting that Anthropic adoption worked well and companies standardized workflows around it.

The question that remains

Something subtle is happening here. As tools like smart reports become standard across generative AI platforms, they create a new category of business metric: the ratio of token cost to delivered value. This fundamentally changes how companies evaluate AI — from an engineering experiment to an infrastructure investment with measurable ROI.

But measuring spend and activity is not the same as proving verified productivity or profit impact. While Anthropic's tools capture transcripts and estimated costs, the central question about how AI transforms (or doesn't) actual productivity remains without a definitive answer — and without a universal metric that all companies agree to use.

As global AI spending is projected to reach $2.7 trillion in 2026, according to Gartner, the need for transparent reporting extends far beyond Anthropic. It's a trend that will shape how companies evaluate technology for years — and yesterday's launch is only the beginning of a much more complex conversation still to come.

Sources: TechCrunch, Cryptopolitan, ZDNET, Unite.AI

✓ Independent sources cross-checked and verified before publishing