Enterprise AI
Software Factory
What enterprise software teams build with Factory
September 24, 2026 - 2 minute read
Enterprise software work includes the migration nobody wants to restart, the package that must keep building, and the integration failure that crosses team boundaries. A convincing agent demonstration should resemble one of those jobs.
Factory’s published customer stories describe specific engineering work alongside the outcomes customers report. They provide a starting point for choosing a pilot, without treating different workloads or measurement methods as a shared benchmark.
Look at the enterprise software being built
Comarch describes XSLT modernization, database migration, and customer-facing builds. A scoped XSLT project estimated at 10,000 person-days completed in 45 days.
Chainguard builds secure open-source packages across repositories. One two-week Droid session spanned six repositories and built 80 packages.
Nav describes work on financial products, including a credit-monitoring update across integrations. It reports 60% less context-switching time and twice-as-fast feature development cycles.
Empower builds cash-advance and credit-building software. Its account reports up to 40% less incident-response time and up to 50% less PR-to-approval time.
You.com describes search infrastructure and open-source integration work. In one investigation, Droid reproduced a cross-stack MCP timeout and traced it to the client.
Groq describes day-0 launches, telemetry, configuration, and CI work. Its reported 3x medium-complexity and 5x quick-turn task results compare Droid with Kimi K2 on Groq against GPT-5 on Codex and Claude Code, on groq/openbench.
The most useful detail is often the work surrounding the patch. Nav’s engineers needed approved access to code, documentation, tickets, and prior decisions. Chainguard needed continuity between design discussions, package implementation, and engineering review.
Nav CTO Gian Perrone describes that boundary in the linked case study:
“Factory provided us with a secure, controlled way to unify our engineering context without compromising our compliance requirements.”
Chainguard Staff Engineer Josh Wolf describes the continuity:
“When you don't have to think about context windows, you can treat Droid like a colleague that just remembers what you've been talking about.”
Choose an enterprise software pilot from the bottleneck
A team blocked by cross-repository context should evaluate a task that requires that context. A team waiting on review should preserve its existing review criteria and measure the wait. Copying another company’s percentage into a business case skips the engineering conditions that produced it.
You.com’s account adds a different workload: recurring research and integration work that competes with new product development. Senior Software Engineer Edward Irby explains:
“With Droid handling the research and the long-running tasks, our limited time goes to the new things we have planned.”
Factory’s Missions support complex, multi-step work. Its automated code review addresses an existing review stage. Choose the mechanism after identifying the bottleneck, then record acceptance criteria, reviewer effort, elapsed time, and operating cost.
Keep the project boundary explicit. A customer’s scoped migration is not evidence that every legacy estate can move at the same speed. An MVP is not equivalent to a complete production rollout.
Discuss a comparable project with Factory and the evidence your team would need before expanding adoption.
