AI 热点

Islo 获 Insight Partners 领投 1500 万美元,推出企业软件工厂平台

Unite.AI更新于 11月18日 01:50

Islo Raises $15M to Scale Its Enterprise Software Factory

AI 导读

Islo 宣布获得 Insight Partners 领投的 1500 万美元投资,并发布 Enterprise Software Factory 平台,用 AI 智能体自动化重复性工程工作,同时保留安全、验证和部署方面的企业控制。该平台以 Factory Lines 把工程流程拆成可重复的智能体与非智能体步骤,构建、lint 和测试由脚本执行,失败时回退重试、高风险动作可等待人工审批;每台 Machine 提供硬件隔离的持久环境,默认拒绝出站连接并记录命令与文件改动。Databricks 将成为其后端推理合作伙伴,团队可在 Claude Code、Codex 等工具间切换模型,并用历史工单在模拟的 GitHub、Slack 和 Linear 环境中测试模型或提示词改动。

推荐理由

AI 评分 57

融资与产品发布同时落地,文中对工厂流水线、隔离执行环境和审批控制的描述可供评估智能体自动化的落地边界。

主题

标签

正文

Islo announced a $15 million investment led by Insight Partners and the launch of its Enterprise Software Factory platform on November 17, 2026. Part of Incredibuild’s product family, Islo designed the system to automate recurring engineering work with AI agents while preserving enterprise controls over security, verification and deployment.

The money will support product development, adoption and expansion of Islo as a dedicated product line within Incredibuild. Incredibuild acquired Kypso, the startup founded by Islo’s leadership team, in 2025. The platform incorporates Incredibuild’s build, test and continuous-integration acceleration, connecting the company’s existing pipeline technology with controls for agent-generated work.

Factory Lines and Isolated Execution

Factory Lines turn an engineering procedure into a repeatable sequence of agent and non-agent steps. A team can reserve implementation or review for an AI agent, then use scripts for builds, linters and tests so those checks do not depend on model judgment. Runs may be triggered by a new issue, a Slack message, a webhook or a timetable. When a check fails, the process sends work back for another attempt; when an action is considered risky, it can wait for human approval.

The execution environment is also part of the product. Each Machine supplies a persistent, hardware-isolated environment with the repositories, services and dependencies required to build and run the customer’s software. This allows agents to verify changes against the actual stack, while cloud hosting lets multiple Machines continue running without an engineer staying logged in.

Security controls sit within that environment. Outbound connections are denied unless permitted, gateway injection keeps raw credentials out of the sandbox, and commands, file modifications and network activity are recorded. Customers can also run Islo in their own cloud, set spending limits for each line and attach sessions, actions, outbound calls and usage costs to an auditable run record, according to the product page.

Early customer uses listed by Islo include refactoring, code review, quality assurance, bug fixing, issue triage and long-tail feature delivery. Teams can set gates for safety, testing, functionality, security, performance and cost, and decide which actions require human sign-off.

“AI can dramatically increase the amount of code an engineering organization produces, but that doesn’t automatically translate into more software being delivered safely,” said Adam Gold, Islo’s chief product officer. Islo’s response is to make the checks and approvals surrounding a change as repeatable as its generation.

Databricks Partnership and Model Testing

Databricks is also slated to become Islo’s backend inference partner. Islo said the integration will give customers access to proprietary and open AI models and let joint customers run factories against their own data without giving up privacy or control. Model execution and access to Databricks data will be available inside the Software Factory, placing those resources within the same controlled environment as the engineering workflow.

Model independence is another part of the design. Teams can use Claude Code, Codex or other agent tools and select different models for individual stages, without rebuilding an entire line when they change one component.

Islo says proposed changes to a model, prompt or agent harness can first be tested against historical tickets in the customer’s environment. Simulated, state-preserving versions of GitHub, Slack and Linear prevent the agent from discovering the eventual resolution. The evaluation then examines the result, hidden tests, side effects, attempts and cost, producing evidence based on the customer’s work rather than a general vendor benchmark.

Islo reports that Rite uses the platform for feature delivery and bug fixing, while Incredibuild expanded from automated pull-request review to bug fixing, feature delivery and continuous-integration triage. An ARIMLABS pilot grew to roughly 500,000 evaluation executions per day. Those examples point to the platform’s initial workload: continuous, recurring tasks whose inputs and acceptance checks can be specified in advance.

The immediate opportunity is the backlog of maintenance, testing and triage work with objective checks. The central question for the new product line is whether Islo’s controls can keep review and verification from becoming the next constraint as customers increase the volume and autonomy of agent-generated changes.

来源

Unite.AI