Noeri Raises $110M for Company-Owned, Self-Improving AI
The knowledge a company puts into Noeri is supposed to stay useful after the next model arrives. That ownership proposition gives the Palo Alto AI lab's $110M launch a commercial question worth following: how much of the intelligence developed through daily work can an organization actually carry forward?
Noeri announced the funding on November 16, 2026, led by Menlo Ventures, with Madrona Venture Group, Kindred Ventures, Wing Venture Capital and NEA participating. Cisco Investments, NVIDIA and Snowflake joined as strategic investors. Stanford computer science professors Yejin Choi and Carlos Guestrin founded Noeri in 2026 and serve as co-CEOs.
The company is developing AI that specializes around an organization's own work and keeps learning as that work changes. For enterprise buyers, the attraction is the possibility of retaining accumulated expertise while gaining access to successive generations of models. The financing stage and valuation were not disclosed, leaving the ownership and operating design as more useful subjects than an invented round label.
What Noeri means by company-owned intelligence
In their founding vision, Choi and Guestrin argue for many specialized intelligences shaped by the organizations and people using them. Their position puts the craft of making AI, and the ownership of what emerges, closer to the source of the knowledge. It is a business proposition with a philosophical foundation: a company should be able to develop expertise in its own systems rather than depend entirely on changes made elsewhere.
That proposition becomes concrete when a workflow improves through repeated use. The company supplying the operational context has an interest in keeping the resulting instructions, methods and specialized models available. An enterprise buyer can reasonably ask how those assets move between providers, how they are maintained and what remains usable when an underlying model changes. Those are procurement questions suggested by Noeri's design, rather than capabilities established by a customer deployment described here.
The ownership language has an important boundary. Noeri says customers retain their data and the workflows and models they develop, subject to applicable third-party model licenses. A buyer evaluating that promise needs to understand the rights attached to each component. Ownership of a workflow and unrestricted rights to every model inside it are separate contractual questions, and the announcement preserves that distinction.
The person reviewing a self-improving system
Noeri describes an improvement process that reaches across instructions, tools, the surrounding agent software, model selection and sometimes the models themselves. The company says changes are explained and sent to a person for approval. That gives human review a specific place in a system whose central selling proposition is its ability to keep changing.
For an operating team, this is a consequential design choice. A proposal can look better on one measure while changing behavior someone else depends on. The person accepting it needs to understand what was measured and which part of the workflow is being altered. Noeri's website says quality and cost are assessed against the customer's own traffic and metrics, placing the company's definition of useful work inside the evaluation.
The broader implication is that buying continual improvement also creates continuing review work. Teams need a way to decide which proposed changes deserve adoption and which lessons should influence later behavior. This is an interpretation of the approval mechanism, not evidence that Noeri has already resolved every governance problem. It explains why the reviewer belongs in the commercial discussion alongside the model and the price.
Noeri says its approach works with closed and open-weight models. That flexibility makes the work around the model important: an organization's instructions, tools and evaluation criteria can have a life across several model choices. Buyers will want that continuity to survive ordinary changes in software suppliers. DevCuration's coverage of Dreamteam's human-approved sales workflows follows a related question about keeping people responsible as AI takes on more work.
Why investors are financing the work around models
Menlo Ventures' 2026 enterprise AI report reports that post-training adoption among AI builders increased from 6% in 2025 to 10% in 2026. Of builders using post-training, 91% relied on outside partners. Menlo is Noeri's lead investor, so these findings belong in the article as an interested investor's research, with the underlying population made explicit.
The report draws on a Morning Consult survey of nearly 400 US enterprise decision-makers conducted in September 2026. Its post-training findings describe a market in which access to models and the expertise to adapt them remain different things. For Noeri, the opening is in that adaptation work and in the possibility that useful improvements can continue after initial deployment.
The founders bring relevant histories to that assignment. Choi's background includes research at Ai2 and NVIDIA, while Guestrin co-created XGBoost and co-founded Turi, which Apple acquired in 2016. He later led Apple's central machine learning and AI organization. Their experience helps explain the research direction; the commercial evidence will come from the systems customers use and the work those systems improve.
Noeri's research agenda includes learning at test time, self-distillation, synthetic data and tool orchestration. These lines of work offer technical context for the company's ambitions. Results described in research publications should be evaluated within their particular tasks and methods before being carried into claims about enterprise deployments. Corma's investment in its own defensive AI models offers another example of research and customer deployment developing inside one business, with a different technical strategy.
Research, customer work and the next handoff
Noeri describes early proof-of-concept partnerships across enterprise and scientific domains. Its funding announcement says some strategic partners are also investors, but it does not establish a roster of named commercial customers. That stage matters when evaluating what the $110M announcement tells an enterprise buyer about the product today.
Scientific discovery extends the same idea into work where the next useful result may require many experiments. Noeri's founders want specialized systems to keep learning through those attempts, alongside scientists. The economic ambition is to make persistent investigation useful for more organizations; no universal savings percentage or deployment outcome is established by the launch.
The lab is hiring researchers and engineers in Palo Alto and Zurich through its official careers page. Research and customer-facing engineering therefore sit inside the same developing business. The announcement gives no detailed allocation of the financing, and the article does not assign one.
For the enterprise team considering Noeri, the purchasing conversation will eventually arrive at the accumulated work: the instructions revised, the tools improved, the models specialized and the changes someone approved. The ownership promise becomes tangible when that team can identify what it has built, understand the rights attached to it and decide how to keep it working through another generation of AI.
Enterprise AI funding, last 30 days
DevCuration's funding database tracked 16 Enterprise AI rounds totaling $1.4B in disclosed capital over the past 30 days. Recent deals we covered:
- Dreamteam Raises $53M for Shared-Context AI Sales CRMSeed and Series A · $53M · Nov 17
- Avanade Agrees to Buy Robiquity for UK AI AutomationNov 14
- Hatz AI Raises $24M Series A for MSP AI BuildersSeries A · $24M · Nov 13
- Nebuly Raises $21M Series A to Measure Enterprise AI ROISeries A · $21M · Nov 13
- Mercor Raises $750M Series D for Enterprise AI EvalsSeries D · $750M · Nov 11
Frequently Asked Questions
What would a company own when using Noeri?
Noeri says customers retain their data and the workflows and models they develop. Applicable third-party model licenses still apply, so buyers need to examine the rights attached to each component.
How does human review fit into Noeri’s self-improving AI?
Noeri says proposed changes are explained and submitted to a person for approval, much like a code review. The described changes can involve instructions, tools, model selection and model parameters.
Why is Noeri relevant to enterprise AI buyers?
Noeri is developing specialized AI around an organization’s own work, with an emphasis on continual improvement and customer ownership. Buyers can evaluate how accumulated workflows and models remain useful as underlying model choices change.
How mature is the commercial evidence behind Noeri’s launch?
Noeri describes early enterprise and scientific proof-of-concept partnerships, including strategic relationships with some investors. The announcement does not establish named commercial customers, revenue or generally applicable customer performance results.
Does the $110M announcement identify a financing stage or valuation?
The November 16, 2026 announcement identifies Menlo Ventures as the lead investor and names participating firms. It does not disclose a Seed or Series label or a valuation.
Where the Money Moved
The intelligence briefing of the innovation economy. Funding, M&A, debt and fund closes, read as market signal rather than deal announcements.
Subscribe to Where the Money Moved

