Nebuly Series A Raises $21M to Measure AI Value
Nebuly raised $21 million to measure the business value of enterprise AI, where more activity does not automatically mean better results. Announced November 12, 2026, the Nebuly Series A funds software that connects AI use with estimates of time saved and business impact. The central question is how closely those estimates match work completed, capacity recovered, and financial outcomes.
That puts enterprise AI owners under a different kind of pressure. Deploying an assistant and reporting adoption can establish that employees use it. Defending its budget requires evidence about what changed afterward. Microsoft already offers estimated impact metrics through its Copilot Dashboard, giving Nebuly an established platform vendor to compete alongside.
Nebuly’s opportunity sits between those usage reports and the decisions finance teams need to make. Its approach brings assumptions into view, allowing customers to examine how activity becomes an estimated return. The distinction matters because an estimate can inform a decision without proving that AI caused a productivity improvement.
What the Nebuly Series A funds
Planven led the financing, with participation from Exor Ventures, Tenacity Capital, TGVG Capital, and Fastweb+Vodafone through Founders Factory. Returning investors and angels also participated. CDP Venture Capital supported the round through its Artificial Intelligence Fund and Digital Transition Fund.
Nebuly says the capital will support product development and expansion of its sales and marketing organization across the United States and Europe. Its Series A announcement identifies Oura, Iveco Group, Fastweb+Vodafone, CNH, and Zeiss as customers. Those names establish company-reported adoption, rather than independently verified customer savings.
The financing backs a business measurement platform, rather than another assistant competing to perform employees’ tasks.
That positioning changes the sales argument. An assistant vendor needs to persuade a buyer that its product can do useful work. A measurement vendor needs to persuade the same buyer that it can judge that work credibly. Customers must trust both the information collected and the interpretation applied to it.
The distinction also shapes what investors are backing. Nebuly does not need every enterprise to standardize on the same AI provider. Its proposed role becomes relevant when organizations need a shared view across different tools. That creates a commercial opening wherever separate deployments produce reports that leadership struggles to compare.
Tamara Djurickovic’s reporting describes a platform spanning employee tools and customer-facing assistants. For employees, the focus includes efficiency and additional capacity. For customer interactions, it includes revenue influence and potential retention or sales opportunities. These are related business questions, but they require different evidence.
The enterprise AI funding therefore carries a narrower meaning than a verdict on AI productivity. It gives Nebuly resources to expand its measurement business. It does not establish that enterprise AI consistently delivers positive returns, or that one measurement approach can settle every disputed investment.
For buyers, that leaves a practical distinction between commercial traction and measurement accuracy. A customer can find a platform useful for organizing decisions before every estimate has undergone independent validation. The funding story becomes more informative when those claims remain separate.
The announcement also contains supportive statements from a customer and the lead investor. Their support explains why the product has attracted commercial interest. It should be read with their relationships in mind, especially when judging claims about the platform’s ability to quantify business value.
AI owners face a budget question beyond adoption
The immediate pressure falls on teams that must explain why an AI deployment deserves continued investment.
Usage is an understandable starting point. An unused assistant cannot help employees, and adoption data can reveal a rollout that needs attention. But a busy assistant can still produce work that requires extensive correction. Activity alone leaves that difference unresolved.
Consider an illustrative internal research workflow. An employee asks an assistant to find a policy, summarize its implications, and prepare a response. A conversation log can show that the exchange happened. It cannot automatically establish that the policy was current, the summary was accurate, or the response was usable.
The business outcome depends on what happens after the exchange. The employee might accept the answer, check its sources, rewrite it, or start again. Each path has a different effect on time and quality. A measurement system needs some way to distinguish them before assigning value.
That creates pressure for AI program owners because adoption and outcomes can move in different directions. Training might reduce unnecessary prompts while improving completed work. Better source material might shorten conversations. An assistant that needs fewer interactions can become more useful even as its activity count declines.
Finance teams face the same ambiguity from another direction. A report describing recovered employee capacity does not automatically explain whether spending fell. It also does not show whether the organization used that capacity productively. Those are subsequent operating decisions, with evidence requirements of their own.
This distinction is particularly relevant to hiring discussions. A task-level estimate does not establish that an entire role can disappear. Jobs combine activities, responsibilities, coordination, and judgment. Removing time from one activity changes only part of that larger system.
The same issue affects knowledge work that depends on finding reliable information. A searchable knowledge base can support the workflow being evaluated. Its business contribution still depends on whether people find the right material and complete useful work with it.
Revenue introduces another layer. A customer might speak with an assistant before buying, but that sequence alone does not establish additional sales. The customer might already have intended to purchase. The interaction can be relevant without being the decisive cause.
Nebuly’s financing matters because it targets this gap between evidence of use and evidence suitable for budget decisions. The buyer’s need is specific: identify which deployments merit expansion, which need improvement, and which no longer justify their demands on the organization.
A common reporting system can help make those choices more consistent. Its value increases when it preserves distinctions between observed events, estimated benefits, and realized outcomes. Combining them into a single impressive total would make the budget discussion easier to present but harder to assess.
How Nebuly works: estimating value from completed tasks
Nebuly’s published method starts with user conversations and the tasks represented within them. It says only interactions meeting the customer’s success criteria contribute to its calculation. A proprietary model estimates the manual work involved, then applies an hourly cost assumption to express value.
Its AI ROI methodology describes adjustable assumptions and reports covering completed tasks, estimated hours saved, and value by activity. Here, ROI means return on investment, although a time-based value estimate remains distinct from a complete financial return calculation.
The useful feature is the ability to inspect the assumptions connecting a completed task to an estimated benefit.
A conversation classified as successful is a starting point for that estimate. The classification itself needs interpretation. An employee keeping a draft can signal usefulness, but acceptance does not necessarily establish factual accuracy. Similarly, the absence of another question does not always prove that a customer’s problem was resolved.
The manual-work baseline carries equal weight. This baseline estimates what the work would have required without AI. If it assumes an unnecessarily slow process, the estimated benefit grows. If it assumes an experienced employee using existing shortcuts, the same interaction can produce a smaller result.
Adjustability makes those disagreements visible. A finance team can challenge the assumed duration of a task rather than reject an unexplained total. A department can distinguish activities that require different levels of effort. That provides a more concrete discussion than a broad claim that everyone feels faster.
However, changing an assumption is different from validating it. Agreement between stakeholders can make reporting consistent without making its baseline accurate. The strongest interpretation depends on comparisons with actual work, including the time spent preparing requests, checking outputs, and correcting mistakes.
That is the central tension within Nebuly AI ROI reporting. A platform can observe interactions at scale while still estimating the alternative process. Directly observed activity and modeled time savings belong to different evidence categories. Keeping both visible makes the resulting number easier to evaluate.
The distinction becomes sharper for newly possible work. An employee might use AI to undertake an analysis that previously remained undone. That can create value, but describing it as time saved implies a manual task would otherwise have occurred. Additional capability deserves its own explanation.
Task boundaries also matter. A research request can involve several conversations, while one conversation can contain multiple requests. Treating each exchange as an independent completed task risks misrepresenting the workflow. Buyers need to understand how any measurement product handles that boundary.
A similar issue arises when several assistants contribute to one output. A search assistant might locate information, another tool might draft text, and an employee might assemble the result. Adding separate estimates together requires care because some claimed benefits can overlap.
The funding therefore supports a product whose usefulness depends on details beneath the headline dashboard. Classification, baselines, exclusions, and overlap determine how readers should interpret the total. Explaining how Nebuly works is ultimately about identifying which parts of the result were observed and which were inferred.
This is also where the platform can support operational decisions beyond a single return figure. A breakdown can direct attention to a workflow that performs poorly or needs better source material. That diagnostic use can remain valuable even when the financial estimate carries substantial uncertainty.
Microsoft already offers estimated AI impact
Nebuly enters a market where platform providers already report more than basic adoption. Microsoft’s Copilot impact metrics include assisted hours and assisted value. Its documentation describes these as estimates derived from usage, research-based assumptions, and an hourly rate.
Microsoft also explicitly describes its calculations as broad approximations. It warns that organizational factors can influence differences in workplace metrics. That qualification matters because native dashboards already acknowledge part of the measurement problem Nebuly addresses.
Nebuly’s proposed advantage is a shared view across providers, while Microsoft supplies an established view within its own environment.
According to Nebuly, its platform connects to custom tools and assistants based on Claude, ChatGPT, and Microsoft Copilot. The distinction is coverage. An organization using several systems has a reason to seek comparisons that do not stop at the boundary of one provider’s products.
That does not make an outside platform automatically more accurate. Independence from an assistant vendor can reduce one commercial conflict, but the measurement company still sells its own interpretation. Customers need to evaluate its method rather than equate a separate supplier with independent scientific validation.
Native reporting has practical advantages as well. A platform provider understands its own product events and can connect them with established administration tools. Buyers may already have access to those reports. An additional vendor must justify the work required to integrate another system and interpret another set of metrics.
Cross-provider reporting introduces its own challenge: similar labels can hide different underlying events. A completed action in one product might represent something unlike a completed action elsewhere. Meaningful comparison requires a consistent understanding of tasks, outcomes, and the human effort surrounding them.
A shared monetary unit does not solve that problem by itself. Converting different estimates into currency can make them appear comparable before their assumptions align. The same caution applies to rankings between departments whose tasks, experience levels, and quality requirements differ.
For example, an illustrative legal review workflow and an internal scheduling workflow have different consequences when an answer is wrong. Ranking them only by estimated time recovered leaves out those consequences. A useful comparison needs to preserve context rather than flatten every activity into the same score.
This is where Nebuly AI ROI competes with the buyer’s existing reporting habits as much as with individual products. The question is whether a shared measurement approach improves decisions enough to warrant another platform. A dashboard’s breadth matters only if its categories remain understandable.
The competitive pressure runs in both directions. Native platforms can expand their reporting, reducing the need for separate analytics in simpler environments. Independent vendors can emphasize mixed deployments, consistent assumptions, and workflows crossing product boundaries. Buyers should judge those routes against their actual reporting needs.
For enterprise teams, the choice is therefore more specific than selecting the dashboard with the largest reported benefit. The relevant comparison concerns coverage, transparency, integration effort, and the decisions each report supports. Those factors determine whether another measurement layer adds useful information.
Estimated savings still need an outcome check
A credible estimate needs evidence about completed work, quality, and the alternative process used for comparison.
Independent productivity research illustrates why this remains difficult. In its February 24, 2026 productivity research update, METR described problems interpreting a follow-up developer experiment. Participants’ reluctance to work without AI affected recruitment and task selection.
The researchers also identified difficulties measuring time when developers used agents concurrently. They believed developers were likely benefiting more than in their earlier study, but described the newer evidence as weak. This was research about developer productivity, rather than an evaluation of Nebuly.
Its relevance is methodological. The work people choose, the participants willing to be studied, and the treatment of overlapping activity can change the result. Even a careful comparison becomes difficult when AI changes the workflow being compared.
That complicates simple time accounting. An agent can run while an employee performs unrelated work, making elapsed duration different from human effort. Conversely, a quick generation can trigger extensive review later. A measurement focused on the interaction itself needs context about the surrounding process.
Quality can also move separately from speed. An employee might produce a more complete document in the same time, gaining an outcome that a time-saved metric misses. Another might finish sooner while leaving more errors for colleagues. Both situations challenge a single measure of productivity.
Customer-facing revenue claims require comparable discipline. Revenue influence describes an association between an interaction and a commercial outcome. Incremental revenue means sales that would not otherwise have occurred. Establishing the second requires a stronger comparison than showing that the assistant appeared somewhere in the customer journey.
Potential churn and upselling signals also remain signals until subsequent events support them. A conversation can identify dissatisfaction or buying interest without establishing that an intervention changed the outcome. Their operational usefulness should be assessed separately from claims about realized financial benefits.
Deployment choices answer another set of questions. Nebuly says its models can run on customer infrastructure or in a private cloud without sending enterprise data to external model APIs. That offers a stated deployment option, rather than proof that every implementation meets every privacy or compliance requirement.
The analysis still depends on what information enters the system and who can access it. Conversation data can contain sensitive business or employee material. Buyers need to understand collection scope, retention, permissions, and how results are presented within their organization.
These questions also affect employee trust. A tool introduced to assess workflows can be interpreted as a way to rank individuals. Comparisons become especially difficult when employees handle different tasks or levels of complexity. An aggregate figure should not acquire more precision merely because it appears beside a person’s name.
The reviewed announcement does not provide an independent validation establishing the accuracy of Nebuly’s time estimates across all named customers. Nor does investor support supply that validation. Those limits leave room for useful reporting while preventing a funding announcement from becoming proof of measured customer outcomes.
The Nebuly Series A supports further development within these constraints. Better measurement can make uncertainty easier to inspect and decisions easier to explain. Its credibility depends on preserving that uncertainty when the available evidence cannot justify a definitive savings claim.
What to watch after the funding
The next useful signals concern validation, reporting detail, and competition, rather than another headline total.
The first signal is a customer account connecting estimated benefits with independently checked work. A persuasive example would describe the task, the comparison process, quality requirements, and human effort surrounding AI use. It would distinguish expected capacity from an observed operational change.
Such evidence would strengthen the case for using the platform in consequential budget decisions. A customer endorsement without those details would provide a narrower signal of satisfaction. It would not resolve whether the reported value accurately represents the work or financial outcome.
The second signal is greater product detail about uncertainty and incomplete coverage. Buyers need to understand unsuccessful attempts, missing interactions, overlapping tasks, and changed assumptions. Reporting that makes these visible would help explain why a figure moved between review periods.
That distinction matters when an apparent improvement comes from revised settings rather than better performance. A change history could let a reviewer separate methodological updates from changes in the underlying work. Clear treatment of those differences would strengthen confidence in comparisons over time.
The third signal is how native platforms develop their own measurement tools. Expanded reporting could make separate analytics less necessary for organizations concentrated on one provider. Broader mixed deployments would strengthen the argument for a shared approach across tools.
Neither outcome automatically determines Nebuly’s prospects. The more informative question is whether buyers use its reporting to make better-supported decisions than their existing systems allow. Evidence of that contribution would explain the product’s role more clearly than a claim to measure every benefit of AI.
These are observation points for the coming months, rather than scheduled announcements. The verified event is the financing and Nebuly’s stated expansion plan. Future product capabilities and customer results should be evaluated when supporting evidence becomes available.
For developers and knowledge workers, the reporting choices will shape how organizations interpret their use of AI. A system that recognizes completed work and quality can support a more useful discussion than one rewarding activity. Poorly framed metrics can encourage behavior that improves the dashboard while leaving the work unchanged.
For enterprise buyers, the Nebuly Series A is a reason to examine how AI value gets reported. Before expanding a deployment, ask which outcomes were observed, which benefits were estimated, and which assumptions changed the result. Keep recovered capacity separate from spending reductions, and revenue influence separate from additional sales. Then compare the report with the workflow employees and customers actually experienced. The strongest investment case will explain both the useful work and the uncertainty around its value. When the next AI budget reaches review, can its owner connect the headline number to evidence that another team can inspect?



