Home  >  Companies  >  Nebuly
Nebuly
An enterprise platform that measures AI agents’ adoption, effectiveness, and business return
View PDF
Details
Headquarters
San Francisco, United States
CEO
Francesco Signorato
Website
Milestones
FOUNDING YEAR
2022

Revenue

Nebuly does not disclose revenue or annual recurring revenue (ARR). In November 2026, it reported multiple six-figure contracts and named Oura, Iveco Group, Fastweb+Vodafone, CNH and Zeiss as customers. Most of its customers are large European enterprises.

Several deployments have expanded beyond a single assistant. Iveco uses Nebuly across dozens of AI copilots, and a bank expanded from an HR assistant into internet banking and other applications.

Valuation & Funding

In November 2026, Nebuly raised a $21M Series A led by Planven. Exor Ventures, Tenacity Capital, TGVG Capital, Fastweb+Vodafone (through Founders Factory) and returning investors also participated. CDP Venture Capital invested through its Digital Transition Fund, which is backed by the EU's NextGenerationEU program.

Nebuly plans to use the capital to build sales teams in the US and Europe and to continue developing its platform.

Francesco Signorato (CEO), Diego Fiori (CTO) and Julien Roux (CRO) founded Nebuly in 2022. Signorato researched AI at MIT, Fiori was an early employee at Giotto.ai, and Roux previously invested at Exor. Nebuly began as an open-source tool for improving AI model performance before shifting to enterprise AI analytics, and it now operates from Turin and New York.

Product

Nebuly reads the conversations that employees and customers have with a company's AI assistants and estimates whether those assistants are getting useful work done. AI teams, support leaders and finance staff use the results to decide which assistants to fix, expand or stop paying for.

A company first connects conversation logs from Claude, ChatGPT, Microsoft 365 Copilot, Gemini Enterprise or a custom-built assistant. It can do this through an API, scheduled sync scripts or existing Langfuse traces. Nebuly's own models remove personal data, group messages into tasks and flag failures using signals such as repeated rephrasing, corrections and abandoned chats.

An operator then opens the Failure Intelligence report. They might see that an IT help-desk bot keeps failing on one application, read the underlying chats and adjust the categories to fit the business. The team fixes the assistant outside Nebuly, then tracks whether that failure rate falls.

AI ROI assigns each successful employee task an estimated manual duration and hourly cost, both of which finance can override. AI Proficiency scores employee prompts from 0 to 100. For customer-facing assistants, Nebuly flags conversations that signal churn risk or upsell interest.

Business Model

Enterprises pay Nebuly a negotiated license covering agreed assistants and volumes. Customers buy after a demo rather than from a public price list. When a contract commits to certain quantities, the customer owes those fees even if it uses less.

Nebuly does not take a share of the savings it identifies. The AI spending shown in its dashboards is money customers already pay to model and assistant vendors, not Nebuly revenue. Nebuly can even recommend that a customer cut spending, which supports its claim to be a neutral evaluator.

The main cost is turning messy conversations into structured judgments. Each additional category set raises recurring processing costs, and reclassifying historical data consumes extra compute. Nebuly uses purpose-built small models instead of sending every conversation to a frontier model. If those models are accurate enough, Nebuly could analyze more conversations per dollar of fees, though no public benchmark shows that advantage.

Deployment choice shifts costs. In hosted deployments, Nebuly pays for infrastructure. In self-hosted deployments, customers run the infrastructure, but Nebuly takes on installation and upgrade work.

Accounts grow by adding assistants, business units and conversation volume rather than seats. Nebuly also has a reseller channel: an Italian telecom operator that used the product internally now resells it to its own customers.

Competition

Nebuly's closest independent rivals are Workhelix, Larridin, Ascerta, Worklytics and Olakai, and IBM Apptio is moving in from technology finance. The central contest is over whose evidence finance leaders trust when deciding which AI tools to renew or cut. Worklytics avoids reading conversation content entirely, and Olakai combines ROI measurement with governance and on-premises deployment, so neither neutrality nor private hosting is unique to Nebuly.

Workhelix

Workhelix pairs a taxonomy of more than 250,000 tasks with control groups and staged rollouts to estimate whether AI caused an improvement. That gives finance buyers a stronger answer than modeled time savings, and Accenture's investment puts Workhelix inside its workforce training services.

Larridin and Ascerta

Both rivals combine measurement with cost control. Larridin observes workflows through browser and desktop agents and routes coding requests to cheaper models. Ascerta, formerly the AI cost-management company Pay-i, rebranded in September 2026 around ROI. Each one can lower customer spending directly, while Nebuly mostly makes recommendations.

IBM Apptio

Apptio launched AI Value & ROI in public preview in August 2026, inside the software finance teams already use for technology budgeting. The risk is that Nebuly ends up as a data feed into Apptio rather than the tool executives use to allocate budgets.

Native platforms and observability tools

Microsoft, OpenAI and Salesforce now ship their own adoption and impact analytics, though each covers only its own agents. Langfuse, now owned by ClickHouse, launched an AI gateway, and LangSmith is expanding toward enterprise platform buyers. Both observability tools supply data Nebuly ingests while also edging toward its buyer.

TAM Expansion

Nebuly's measurement products already cover AI adoption, value and proficiency, so selling more of them is account growth rather than new market. The incremental opportunities come from turning existing measurement capabilities into actions with separate budgets: controlling AI spending, delivering training and selling privacy tooling. None of these is an announced plan.

Active AI spending controls

Finance and AI infrastructure teams buy tools that enforce spending policies. Nebuly's Routing Intelligence work, announced in August 2026 and still under development, matches recurring tasks with the cheapest model that meets a required quality bar. Nebuly currently says it does not set or enforce budgets.

A control product, possibly delivered through AI gateway partners, could tie model choice to whether tasks actually succeed. Building one would require gateway integrations, permissions, rollback mechanisms and proof that cheaper models preserve quality.

Workforce training delivery

HR and learning budgets pay for training, not dashboards. Nebuly already identifies weak prompting habits by team, and its September 2026 roadmap extends measurement to delegation, judgment and agent orchestration.

Standalone privacy software

Nebuly's locally deployed model for removing personal data already runs for enterprise customers and accepts custom categories, such as portfolio company names. Packaging it separately could reach security and data teams that do not need ROI analytics. Nebuly's own benchmark missed some sensitive data in about 5% of documents, however, so the product would need independent validation and audit features first.

Risks

Modeled value: AI ROI starts from adjustable defaults of 15 minutes per action and $75 per hour rather than observed cash savings. If finance teams cannot reconcile these estimates with actual headcount or cost changes, Nebuly looks like an advocate for AI spending and becomes vulnerable at renewal.

Platform substitution: Microsoft, Salesforce and IBM Apptio are bundling AI value reporting into contracts customers already hold. Companies whose AI use is concentrated in one ecosystem have less reason to buy a separate cross-vendor layer.

Privacy bottleneck: Nebuly says data redaction once stretched deployments to months, and its model still missed 80 of 3,027 sensitive values in testing. Customer-specific validation work could slow rollouts and weigh on margins even when customers see strong ROI.

DISCLAIMERS

This report is for information purposes only and is not to be used or considered as an offer or the solicitation of an offer to sell or to buy or subscribe for securities or other financial instruments. Nothing in this report constitutes investment, legal, accounting or tax advice or a representation that any investment or strategy is suitable or appropriate to your individual circumstances or otherwise constitutes a personal trade recommendation to you.

This research report has been prepared solely by Sacra and should not be considered a product of any person or entity that makes such report available, if any.

Information and opinions presented in the sections of the report were obtained or derived from sources Sacra believes are reliable, but Sacra makes no representation as to their accuracy or completeness. Past performance should not be taken as an indication or guarantee of future performance, and no representation or warranty, express or implied, is made regarding future performance. Information, opinions and estimates contained in this report reflect a determination at its original date of publication by Sacra and are subject to change without notice.

Sacra accepts no liability for loss arising from the use of the material presented in this report, except that this exclusion of liability does not apply to the extent that liability arises under specific statutes or regulations applicable to Sacra. Sacra may have issued, and may in the future issue, other reports that are inconsistent with, and reach different conclusions from, the information presented in this report. Those reports reflect different assumptions, views and analytical methods of the analysts who prepared them and Sacra is under no obligation to ensure that such other reports are brought to the attention of any recipient of this report.

All rights reserved. All material presented in this report, unless specifically indicated otherwise is under copyright to Sacra. Sacra reserves any and all intellectual property rights in the report. All trademarks, service marks and logos used in this report are trademarks or service marks or registered trademarks or service marks of Sacra. Any modification, copying, displaying, distributing, transmitting, publishing, licensing, creating derivative works from, or selling any report is strictly prohibited. None of the material, nor its content, nor any copy of it, may be altered in any way, transmitted to, copied or distributed to any other party, without the prior express written permission of Sacra. Any unauthorized duplication, redistribution or disclosure of this report will result in prosecution.

View PDF
Details
Headquarters
San Francisco, United States
CEO
Francesco Signorato
Website
Milestones
FOUNDING YEAR
2022