Xpanderがエンタープライズ向けAIエージェントを本番導入するために$7.5Mを調達

Why it matters

Xpander is turning David Twizer's AWS migration experience into a governed runtime for enterprise agents, with Omni serving as the wedge into production workflows.

David Twizer, Ran Sheinberg, and Moriel Pahima have raised a 750万ドルのシードラウンド for Xpander, their attempt to turn the difficult work of deploying enterprise AI agents into a repeatable software product.

Xpander announced the financing on August 17, 2026, alongside the release of Omni, an agent that Xpander describes as an "agentic Forward Deployed Engineer." Pico Venture Partners led the round, with Emerge Ventures, Samsung Next, and SeedIL participating. Xpander says the money will support market expansion. The announcement gives no valuation, ownership terms, or operating metrics such as revenue, customer count, or growth.

The founders established Xpander in 2024 and serve as CEO, chief product officer, and chief technology officer, respectively. Xpander is headquartered in San Francisco, according to its funding announcement.

Twizer and Sheinberg each spent years at AWS. Twizer worked with large enterprises on complex cloud migrations、while Sheinberg worked on large-scale compute and big-data workloads。Twizer spent six years at AWS as a principal solutions architect and led its go-to-market generative AI solutions architecture team. Sheinberg spent five years there as a principal solutions architecture leader, including work involving EC2 Spot. Xpander identifies Pahima as a former AWS principal engineer.

Twizer has said his work helping large enterprises through complex, multiyear cloud migrations shaped Xpander's thesis: AI-agent deployments need a shared operational layer that replaces isolated pilots.

Omni is the founders' software version of a deployment engineer

Omni accepts a plain-language description of a business process and, according to Xpander, can connect tools, create and test an agent, repair integrations when APIs change, compare models, and move the resulting agent into production. One example on Xpander's Omni page asks the system to build an IT access agent connected to ServiceNow and Jira, automatically approve requests that match policy, and send exceptions for review. Both the platform and Omni are available at chat.xpander.ai.

Xpander says its platform handles runtime infrastructure, permissions, secrets, execution logs, model selection, spend attribution, and deployment across customer-controlled environments. Its documentation describes hosted and self-managed Kubernetes installations. In an official company blog post, Xpander says its VPC-native deployment keeps agent execution, tool calls, and memory operations inside the customer's security perimeter. Its self-hosted documentation says task execution, agent memory, connector credentials, model API keys, and user-facing services run in the customer's VPC. Xpander also claims support for private networks, on-premises environments, and air-gapped systems, and says it is SOC 2 Type II certified and GDPR compliant.

Xpander also says developers can bring agents built with frameworks such as LangChain, LangGraph, Agno, or the OpenAI Agents SDK. According to its deployment documentation, agents can be exposed through Slack, webhooks, APIs, scheduled tasks, chat interfaces, and the Model Context Protocol. Xpander says its model support spans commercial and open-weight systems.

Vendor neutrality is central to Xpander's sales pitch. Xpander says a customer can change the model, framework, or cloud underneath an agent while keeping its operating and governance layer. That proposition speaks directly to buyers wary of committing production workflows to a single model provider while model quality, pricing, and availability continue to change.

It also gives Xpander a demanding product scope. Each additional framework, model, connector, deployment target, and authentication method increases the number of failure paths Xpander must support. Enterprise buyers will judge the platform on incident response, auditability, upgrade safety, and the behavior of agents when connected systems change.

The pilot bottleneck is real

A 2025 McKinsey survey found that 88% of organizations regularly used AI in at least one business function. Roughly one-third had begun scaling AI across their enterprises, leaving nearly two-thirds in experimentation or pilot stages, and only around 1% described their deployments as mature.

The founders are building for the work between those states. Xpander says its product design centers on bounded permissions, durable integrations, monitoring, approval paths, cost controls, credential management, and execution logs for agents operating across company systems.

Xpander's pitch is that these controls should be shared across agents instead of rebuilt for each project. If the platform works as described, the first deployment creates reusable infrastructure for later ones. That would let Xpander expand inside an account as teams move from isolated assistants toward agents that can act across internal systems.

Xpander positions Omni as a clearer commercial wedge for that infrastructure. Platform layers can be difficult to sell before a buyer has deployed enough agents to feel the operational pain. Xpander promises an immediate outcome: give Omni a workflow, let it assemble the agent, and keep the resulting workload on Xpander. In effect, Xpander is packaging part of the professional-services labor commonly needed for enterprise AI projects into the product itself.

A benchmark claim with important limits

Xpander says Omni scored 90.9% on the GAIA validation benchmark and has published its results and methodology in a GitHub repository. That score measures general assistant performance; it does not establish production reliability, enterprise governance, or safe recovery from failed actions.

The result is a company-published claim, without an independent benchmark audit.

GAIA evaluates general AI assistants. It does not test the enterprise controls at the heart of Xpander's pitch, including permission boundaries, credential handling, regulatory compliance, integration durability, or safe recovery from a failed action. Evidence for those capabilities would need to come from customer deployments or independent testing.

Xpander says it works with organizations across retail, manufacturing, financial services, technology, and government. Its website displays customer or associated-company logos including Lenovo, Intel, Wix, Siemens, NVIDIA, SAP, Salesforce, and Workday, without specifying deployment size, contract value, or whether each relationship is commercial. Xpander has not disclosed customer count, revenue, annual recurring revenue, or retention. No independent customer or deployment data in the available materials establishes commercial traction or production reliability.

The seed round funds a distribution test

Pico Venture Partners led Xpander's seed round, backing infrastructure engineers who are turning their implementation experience into a platform.

Xpanderには、エージェントの開発と運用のさまざまな部分をカバーする選択肢がある。LangGraphとLangSmithはエージェントの構築、評価、可観測性、およびデプロイを扱う。Amazon Bedrock AgentCore、Microsoft Foundry、Google Vertex AI Agent Builderはそれぞれのクラウドプラットフォームにエージェントツールを接続する。Salesforce AgentforceとIBM watsonx Orchestrateは、より広範なソフトウェアポートフォリオの中で企業向けエージェント、ワークフロー、およびガバナンスに注力している。対象範囲が異なるため、いずれも機能ごとの直接比較にならない。Xpanderは、自社の差別化点はクラウド、モデル、フレームワークを横断したポータビリティと、デプロイ、ガバナンス、および運用制御の組み合わせにあると主張している。

Xpanderは、企業が既存のクラウドスタックとエージェントの間にもう一つのコントロールプレーンを望んでいるかどうかを見極める必要がある。特に大手クラウドやエンタープライズソフトウェアのサプライヤーが、顧客が既に結んでいる契約にエージェントツールをバンドルできる場合において、ポータビリティと集中化されたガバナンスが別個のベンダーを導入する正当性を示す必要がある。

Twizer、Sheinberg、Pahimaはその主張を立証するために実践的な方法を選んだ。Omniは一つのワークフローから導入でき、Xpanderはその下のインフラストラクチャになる。各エージェントが成功するごとに、権限、統合、デプロイパターン、および運用データを再利用することで、アカウント内でプラットフォームの価値が高まる可能性がある。

750万ドルのラウンドは、創業者たちにその進展が実際の組織で成立するかどうかを試す余地を与える。Twizerは、複雑で数年にわたるクラウド移行において企業を支援した自身のAWSでの仕事がXpanderのプラットフォームに着想を与えたと述べている。Xpanderは、AIエージェントを本番環境に導入するために必要な実装作業のより大きな割合をソフトウェアが担えるという考えを中心に構築されている。