Nebuly raises $21M to measure AI returns across enterprise agents

Nebuly says its platform brings employee-facing tools and customer-facing assistants into a single view. For employees, it measures hours saved, capacity freed and tasks made possible by AI. It also identifies higher-value uses and teams that could benefit from further training. For customer assistants, it tracks revenue influenced by AI and flags potential churn and upselling opportunities.

Leadership and finance teams can inspect and change the assumptions behind the figures, the company says. That makes the assumptions open to challenge, but also means the output depends on judgments about the work AI replaces and the value assigned to it. Time saved need not reduce spending, and revenue influenced by an assistant does not by itself establish that the assistant caused the sale.

Nebuly says it connects to custom-built tools and assistants based on Claude, ChatGPT and Microsoft Copilot. Its purpose-built analysis models can run on premises or in a customer's private cloud without sending enterprise data to external model APIs, according to the company's description outlined in coverage of the platform.

Private deployment offers a way to process data within a customer's infrastructure. The announcement does not specify pricing or the infrastructure and integration costs of that option. Buyers therefore still need to account for the cost of measurement itself when evaluating the benefits it identifies.

Nebuly’s diagram groups integrations into employee-facing and customer-facing agents. Diagram: Nebuly. Source: Nebuly.

A market with returns—and competing ways to count them

A survey released September 1 by Gartner, the business and technology research company, provides evidence of the visibility problem beyond Nebuly's sales pitch. It found that roughly 11% of organisations were entirely unaware of what their function spent on AI in 2025, while 85% of functional leaders planned to increase spending in 2026.

The January–April 2026 survey covered 1,303 respondents from organisations with at least $50 million in enterprise-wide annual revenue in fiscal 2025. These are findings about larger organisations, including spending intentions, rather than a measure of realised returns across all businesses.

The survey also offers a counterweight to a blanket account of unproven AI value. Companies Gartner classified as high performers—those that constantly track AI returns, manage initiatives as a portfolio and regularly assess projects to reallocate resources or discontinue weak ones—reported positive returns in 81% of their AI initiatives. That is a reported result for a selected group, not proof that measurement software causes better returns or that Nebuly delivers them.

Nor is cross-enterprise measurement an uncontested proposition. Ascerta, an AI management company formerly called Pay-i, said in its September 30 funding announcement that it was expanding from AI cost management to a single view of cost, adoption and business value. It announced an $18 million Series A led by Dell Technologies Capital. The overlapping scope limits any claim that Nebuly alone can connect spending with outcomes.

The financing report also identifies customer-service software provider Zendesk's move towards pay-per-result pricing. Charging for results and independently measuring returns are different approaches, but both give enterprise buyers alternatives to judging AI by activity counts alone.

Ascerta’s illustrative executive scorecard combines AI spending, adoption and value metrics. Displayed figures are company-reported examples, not verified customer returns. Image: Ascerta. Source: Ascerta.

What buyers still need to establish

Nebuly plans to use the capital to develop its platform and expand its go-to-market team across the US and Europe. The next consequential evidence would be customer results that expose the baseline, assumptions and costs behind a claimed return.

For buyers, the decision is whether those calculations hold up well enough to guide spending: which hours saved become usable capacity or lower costs, which revenue gains can be attributed to AI, and what remains after deployment and operating expenses. Adjustable assumptions let finance teams examine those questions. Published validation would help show whether the answers deserve confidence.