Factory, the enterprise coding-agent startup that counts NVIDIA, Adobe, and Palo Alto Networks among its customers, announced a product repositioning on June 15 that reframes its pitch from autonomous coding to what it calls the “software factory”: a continuous, AI-driven loop spanning bug triage, code review, security analysis, deployment, and incident monitoring.
The announcement, published directly on factory.ai, describes a system where AI agents handle the full development lifecycle rather than assist individual engineers. Inbound signals from bug reports, customer feedback, and business requirements feed an automated triage layer. That layer generates planned changes, which pass through build, test, review, and security stages before deployment. Post-deployment monitoring then generates new signals, restarting the loop. Factory calls this architecture “agent-native” and says it must be capable of improving by observing its own behavior over time.
Three design principles structure the offering. First, model independence: the platform routes tasks to whichever model offers the best cost-performance-speed tradeoff for that specific job, using a component Factory calls the Router. Second, sovereign intelligence: enterprises can deploy fully cloud-hosted, bring-your-own-key, self-hosted, EU-specific, or completely air-gapped configurations, with all learning from agent sessions, code reviews, and incident responses staying inside the customer’s environment. Third, continual self-improvement: because code review, QA, security, and incident response share the same agent core, a security finding can inform subsequent reviews and a deployment can trigger automatic documentation updates.
Factory says its software factories are already running in production at a roster of large enterprises. The company names NVIDIA, EY, Adobe, Palo Alto Networks, Adyen, Blackstone, Wipro, and Comarch as customers. That list is notable for its breadth; the absence of disclosed autonomy metrics is equally notable. The announcement does not include independent benchmarks, third-party adoption data, or any figure measuring how much of the development lifecycle is actually automated at any of those customers versus how much remains human-supervised. A vendor naming eight marquee logos without stating what those customers are actually running autonomously is describing pipeline, not outcome.
The company also positions this shift as a redefinition of the engineering role. In its framing, engineers stop being individual producers of software and become designers of the factories that produce software, taking on governance, safety, and business-outcome responsibilities that extend well beyond the codebase. This is a reasonable description of where the industry is heading; it is also a standard vendor narrative for every enterprise platform that wants to justify broad deployment, so it warrants scrutiny before acceptance.
The announcement covers a spectrum of autonomy levels, from simple Droid agents handling well-defined tasks, to Automations coordinating recurring workflows, to multi-agent Missions that decompose complex work over hours or days. This gradation suggests Factory has accepted that full automation is not the near-term default, even for sophisticated customers. The desktop app update ships alongside the announcement, adding a management view for monitoring the factory’s state directly.
Teams currently evaluating enterprise coding-agent platforms should benchmark Factory’s Router against the task-routing approaches from GitHub Copilot Workspace and Cursor’s background agents before committing to a long-term contract, since model-routing architecture is the differentiator Factory is explicitly staking its platform strategy on.
Based on a company announcement published by Factory on June 15, 2026, at factory.ai/news/software-factory.
