Labelbox Unveils Recursion to Automate Enterprise AI Workflows

Overview

Labelbox has officially launched Recursion, a new platform designed to automate complex enterprise AI workflows. The release marks a significant expansion of the company’s capabilities, moving beyond its established role as a data infrastructure provider for training frontier models into the realm of autonomous enterprise operations.

Recursion allows businesses to deploy AI agents that operate independently within their existing software ecosystems. Rather than functioning as passive chatbots requiring constant human input, these agents can initiate tasks based on schedules, alerts, or API events and work in the background until objectives are met.

The launch underscores Labelbox’s dual strategy: providing the foundational environments for frontier labs to train next-generation AI, while simultaneously offering enterprises the tools to put those agents to practical work. By integrating with popular business applications, Recursion aims to turn static AI models into dynamic, results-driven collaborators.

How Recursion Transforms Enterprise Operations

At the core of Recursion is the concept of sessions. Each session represents a complete workflow, from initiation to graded outcome. According to Labelbox, the platform recently processed 144 such sessions in a single week, demonstrating its capacity to handle high-volume, repetitive, and critical business tasks.

A key differentiator for Recursion is its learning loop. Every run is graded by a dedicated grader agent, which scores the outcome based on predefined success metrics. These graded sessions leave behind memories, notes, and skills that subsequent sessions read before acting. This continuous feedback mechanism ensures that every run makes the next one better, creating a compounding improvement effect over time.

"Recursion isn’t a chatbot. Its agents start on their own, on a schedule, an alert or an event, and work in the background until the job is done."

This approach allows agents to handle end-to-end processes such as due diligence for mergers and acquisitions. In one example, a coordinator and five specialist agents reviewed a data room, checked court records and audits, and drafted an investment committee memo. The system sourced 11 red flags and achieved a high outcome score, with the final document approved for review.

The platform connects seamlessly with a wide array of enterprise tools, including Box, Slack, Salesforce, Notion, and GitHub. This integration capability allows agents to read data, draft documents, check for compliance, and notify stakeholders without manual intervention.

Use Cases Across Departments

Recursion is designed to serve a diverse range of functions, from engineering and product development to finance, security, and research. By automating routine but vital tasks, the platform aims to free up human professionals for higher-level strategic work.

Engineering and Product

For engineering teams, Recursion offers several specialized agents:

• Feature Builder: Transforms ready tickets into working features with tests in a sandbox environment.
• Bug Fixer: Reproduces new bugs, identifies the root cause, and writes fixes with regression tests.
• Dependency Upgrader: Automatically upgrades packages as new versions ship, fixing breakages and running tests.
• Flaky Test Fixer: Investigates intermittent test failures, identifies race conditions, and applies fixes.
• Incident Investigation: Correlates alerts, logs, and recent deployments to produce root-cause drafts before on-call engineers are even notified.

Product teams benefit from agents like the Prototype Builder, which turns approved specifications into clickable prototypes, and the Spec Writer, which drafts initial product requirements based on customer calls and tickets. Additionally, the Docs Updater ensures documentation stays in sync with merged code changes.

Research and Data Science

Research teams can leverage Recursion to manage experimental workflows at scale. The Experiment Sweep agent runs dozens of configurations simultaneously and compares results against baselines. The Eval Regression Watch monitors every model or prompt change, blocking merges if performance regresses.

For more intensive tasks, the Model Trainer launches fine-tuning runs, monitors performance curves, and stops underperforming experiments early. The Benchmark Runner evaluates candidate models against custom benchmarks overnight, providing leaderboards that include cost and latency metrics.

Security and Compliance

Security operations are enhanced through agents like Threat Investigation, which analyzes identity, mailbox, and endpoint alerts as they fire. The Code Security Review agent examines every pull request for vulnerabilities, reproducing findings to verify real issues. In the event of a critical CVE, the Vulnerability Patcher identifies affected services and generates tested patches, prioritizing the most urgent exposures first.

Finance and Sales

In finance and sales, Recursion assists with data synthesis and customer communication. The Feedback Digest reads weekly sales calls and reviews to rank feature requests by customer value and revenue potential. When a requested feature ships, the Close the Loop agent drafts personalized notes to every customer who requested it, ensuring strong account management without manual effort.

Strategic Implications for the AI Industry

The launch of Recursion highlights a growing trend in the AI industry: the shift from experimental proof-of-concept to production-grade automation. While frontier labs continue to push the boundaries of model capability, enterprises are increasingly focused on reliability, integration, and measurable outcomes.

Labelbox’s decision to offer both Horizon for frontier AI development and Recursion for enterprise workflow automation positions it as a full-stack partner in the AI lifecycle. By grading every session and retaining institutional knowledge, the platform addresses one of the most persistent challenges in enterprise AI: consistency and continuous improvement.

As organizations seek to integrate AI into their daily operations, tools like Recursion provide a structured environment where agents can operate safely, effectively, and autonomously. The result is a workflow where AI does not just assist, but actively drives business processes from start to finish.