Problem

Patched addresses the slow, expensive and difficult-to-control manual work inside regulated operations. Its target workflows involve high-volume fraud and risk investigations with false positives and SLA pressure, customer-operations cases involving ambiguous processes and sensitive data, and reconciliation work where cross-system references and near-zero error tolerance make automation difficult. The principal use case appears to be fraud and risk case resolution: Patched cites more than 30,000 cases per month, an 82% resolution rate and 3.4x ROI.

The company also emerged from a related software-maintenance problem: development teams spend substantial time on post-code work such as reviews, bug fixes, documentation and security patches rather than shipping new functionality. In both settings, the pain is not simply generating an answer; it is completing an operational task reliably, within an SLA, and with evidence that can withstand audit or human review.

Product / Service

Patched presents itself as a managed, outcome-oriented AI service for regulated operations. Customers show it an existing process, and Patched executes the work on dedicated infrastructure with compliance controls, auditability and human control; it claims responsibility for outcomes and SLAs rather than merely providing workflow-orchestration software. Its public examples report average turnaround under an hour versus two days previously, with more than 500,000 workflows executed monthly and over 99% completed within SLA. Pricing scales with operational volume.

The underlying delivery model retains an open-source and self-hosted option. The Patchwork framework lets developers define customizable “patchflows” that combine atomic actions with AI, inspect or extend the resulting Python code, connect preferred LLMs, and run workflows locally through a CLI or within CI/CD; Patched also offers a hosted version. This gives regulated enterprises a choice between a managed service and greater deployment, model and privacy control, while the company’s evaluation tooling is designed to filter for more consistent and reliable outputs.

Market

Patched competes in enterprise AI workflow automation, with a sharper focus on regulated operational work in banking and other high-sensitivity enterprises. It is adjacent to fraud and risk platforms such as ComplyAdvantage, Featurespace, Feedzai, Sardine and Hawk:AI, although the available evidence does not establish that each is a direct head-to-head competitor. Its earlier developer-automation product also places it alongside AI code-review and software-maintenance tools such as CodeRabbit, Greptile, Qodo, Graphite and GitHub Copilot. Patched’s differentiation is the combination of autonomous task execution, configurable code-native workflows, auditability and claimed SLA ownership rather than an IDE copilot or a generic workflow builder.

The public evidence indicates that Patched is active and has meaningful operating traction rather than being merely pre-revenue: YC lists it as an active Summer 2024 company, while its own site reports 500,000-plus monthly workflows, sub-hour average turnaround and 99%-plus SLA completion, and says it is trusted by publicly listed enterprises. The site also gives substantial operating examples for customer operations and reconciliation, including 400,000-plus monthly requests and two million-plus daily transactions. However, the retrieved materials do not disclose revenue, contract values or customer names, so the scale and quality of commercial traction cannot be independently assessed from the public record.