Company cases

How companies build AI software factories

Compare how 21 technology companies move work from request to checked result. Every case explains the operating system in plain language, names the AI-engineering patterns, and keeps company claims separate from what remains unknown.

21 company cases · sources checked through September 1, 2026 by a second AI editorial-review agent (Codex) · company claims stay labeled · alphabetical, not ranked

Compare company software factories

The same six fields appear for every company. Reported use tells you whether the public evidence describes an operating system, one bounded production line, an early rollout, or a product the company also uses itself. Access tells you what a reader can actually inspect or buy.

First-party report marks evidence supplied by the company. Observed artifact marks something readers can inspect directly. Neither label is a quality score, and no blank cell means hidden certainty: unknowns are written out.

The table can scroll sideways when the window is narrow. Tab to the comparison, then use Left and Right Arrow. Home and End move to the first and last columns.

Company software factory comparison, source window through September 1, 2026
Company and systemMain jobReported useHuman checkpointAccessSource window
AirbnbLLM migration pipelineMove 3,500 React test files through staged refactoring and validation.Bounded production lineOne completed production migration program.People tuned the line and finished the final 3% of files.Not publicly availableThe migration system is described publicly; its code is internal.
First-party report
AmazonQ Developer code transformationUpgrade production Java applications at fleet scale.Bounded production lineTens of thousands of migrations; the product path is changing.Developers review and accept each transformation before it lands.Existing customersNew Q Developer accounts and subscriptions stopped May 15, 2026. Eligible existing customers retain transformation access through April 30, 2027; AWS directs new modernization work to AWS Transform.
First-party report
AnthropicClaude Code and Claude TagCarry work from repository instructions through layered review and incident response.Product with internal usePublic products plus extensive first-party internal use.Risk determines review, approval, deployment, and incident permissions.MixedClaude products are public; Anthropic's internal Tag and integrations are private.
First-party report
BlockBuilderbot, Goose, and BuzzTurn Slack requests and tickets into tested pull requests and shared agent work.Internal operationBuilderbot is rolled out internally; Buzz remains early-stage dogfood.People steer work, review changes, and approve consequential actions.MixedGoose, Builderbot, and Buzz have public repositories; internal services differ.
First-party reportObserved artifact
CloudflareInternal AI engineering stackGive company agents one identity, context, tool, cost, and review control plane.Internal operationCompany-wide internal system across thousands of repositories.People own approval while risk rules can block serious findings.InternalThe operating stack is internal; some underlying platform parts are public products.
First-party report
CognitionDevinMove bugs, design audits, and routine fixes from event to reviewed pull request.Product with internal useCognition publishes a detailed account of Devin building Devin.People define playbooks, review changes, and own releases.CommercialDevin is commercial; the internal code and operating data are private.
First-party report
CursorCloud agents and security agentsBuild parallel pull requests with proof artifacts, then run event-driven review.Product with internal useCustomer-facing agents and internal review and security lines.People review artifacts and own merge, severity, and production decisions.MixedCursor is commercial; a security-automation reference implementation is public.
First-party reportObserved artifact
DropboxNova and DeflakerRun interactive coding and repeatable repair jobs on one shared agent platform.Internal operationInternal platform used for CI repair, flaky tests, and migrations.The surrounding workflow controls tests, branches, publication, and retry limits.Not publicly availableNova and Deflaker are internal systems described in a public engineering article.
First-party report
GitHubCopilot coding agentTurn assigned issues into draft pull requests inside the normal repository workflow.Product with internal usePublic product used inside the core repository that builds github.com.A different person approves the work; repository and CI rules still apply.CommercialCopilot coding agent is a commercial GitHub capability.
First-party report
LinearCoding SessionsUse issues and team chat as dispatch and review surfaces for coding agents.Product with internal useA recent public product with concrete internal design and triage examples.People review the diff, proof artifacts, and merge decision in Linear.CommercialCoding Sessions is a commercial Linear capability; the implementation is private.
First-party report
LinkedInCAPTTurn company knowledge and internal tools into reusable, executable playbooks.Internal operationInternal system reported in use by more than 1,000 engineers.Domain experts author and maintain the playbooks; engineers own results.Not publicly availableCAPT is internal and built on the open Model Context Protocol.
First-party report
MetaCapacity Efficiency agentsTurn production performance signals into diagnosed, review-ready code changes.Internal operationSpecialized internal line for performance regressions and opportunities.Engineers review, deploy, and remain responsible for the proposed fix.Not publicly availableThe agent platform and its skills are internal.
First-party report
OpenAICodex and SymphonyPull tasks from a project board into isolated agent workspaces and review loops.Internal operationInternal teams plus one fully agent-written experimental product repository.People prioritize work, set acceptance criteria, and validate outcomes; pull-request review may be agent-to-agent.MixedCodex CLI and Symphony are public; internal repositories and operating data are not.
First-party reportObserved artifact
RampInspectRun background coding with full company context in prepared remote environments.Internal operationInternal operating system with multiple specialized factory lines.People review and land changes; automated lines stop at review-ready work.Not publicly availableInspect is internal; an independent inspired implementation is public.
First-party reportObserved artifact
ReplitInternal agent fleetsLaunch parallel agent work across company systems, check results, and learn from feedback.Product with internal useCustomer product plus a first-party internal agent-of-agents system.People set policy, review risk-rated changes, and judge product outcomes.CommercialReplit Agent is public and commercial; internal orchestration and data are private.
First-party report
ShopifyWorld, River, and AquiferRun public-to-the-company agent sessions and specialist lines on one durable substrate.Internal operationLarge internal operation; the shared Aquifer substrate is still rolling out.People join visible threads, redirect work, and review pull requests and findings.MixedRoast is open source; River, Aquifer, and Dispatch are internal.
First-party report
SpotifyHonk and FleetshiftApply agent judgment to repository-wide maintenance and data migrations.Internal operationInternal agent work built on Spotify's existing fleet-management systems.People initiate work and own review, merge, and unresolved edge cases.MixedHonk is internal; Fleetshift is available as a managed Backstage product.
First-party report
StripeMinionsTurn one unattended delegation into a checked branch for two human review gates.Internal operationHigh-volume internal background-agent operation.The initiator inspects the result and another engineer reviews the pull request.Not publicly availableMinions and its Stripe-specific infrastructure are internal.
First-party report
TempoCustom AgentsTurn product and error signals into prioritized, parallel agent work and review.Early rolloutConcrete public examples; organization-wide operating scale is not stated.People prioritize the queue and review the pull requests or media output.CommercialTempo's agent products are commercial; internal operating details are limited.
First-party report
UberSoftware FactoryCoordinate coding agents, shared context, managed jobs, checks, and result economics.Internal operationCompany operating model described across multiple factory layers.People define valuable work and own quality, acceptance, and release.Not publicly availableUber's factory infrastructure is internal.
First-party report
VerceleveGive many production agents durable runs, sandboxes, permissions, traces, and evals.Internal operationShared framework reported across more than 100 internal production agents.Engineers own release controls, executable guardrails, and production outcomes.MixedThe eve framework is open source; Vercel's internal agents and connections are private.
First-party reportObserved artifact
  1. Airbnb · LLM migration pipeline

    Airbnb's LLM migration factory: 3,500 test files with a proof loop

    Airbnb gave every test file a visible state, used ordinary tools to check model-generated conversions, improved common failures in batches, and sent the final 3 percent to people.

    Main job
    Move 3,500 React test files through staged refactoring and validation.
    Compare details
    Reported use
    Bounded production line One completed production migration program.
    Human checkpoint
    People tuned the line and finished the final 3% of files.
    Access
    Not publicly available The migration system is described publicly; its code is internal.
    Source window
    First-party report
  2. Amazon · Q Developer code transformation

    Amazon Q migrations: an AI factory for fleet-wide Java upgrades

    Amazon turned Java upgrades into repeated analyze, plan, transform, build, test, and review jobs, then began moving customers to successor products.

    Main job
    Upgrade production Java applications at fleet scale.
    Compare details
    Reported use
    Bounded production line Tens of thousands of migrations; the product path is changing.
    Human checkpoint
    Developers review and accept each transformation before it lands.
    Access
    Existing customers New Q Developer accounts and subscriptions stopped May 15, 2026. Eligible existing customers retain transformation access through April 30, 2027; AWS directs new modernization work to AWS Transform.
    Source window
    First-party report
  3. Anthropic · Claude Code and Claude Tag

    Anthropic's AI-native SDLC: agents write, review, test, and investigate code

    Anthropic connects Claude Code and Claude Tag to remote environments, specialist review agents, deterministic checks, risk-based approval, and incident response.

    Main job
    Carry work from repository instructions through layered review and incident response.
    Compare details
    Reported use
    Product with internal use Public products plus extensive first-party internal use.
    Human checkpoint
    Risk determines review, approval, deployment, and incident permissions.
    Access
    Mixed Claude products are public; Anthropic's internal Tag and integrations are private.
    Source window
    First-party report
  4. Block · Builderbot, Goose, and Buzz

    Block's AI software factory: Goose, Builderbot, and the Buzz workspace

    Block connects an open agent engine, a shared workflow that turns Slack requests into reviewed code changes, layered tests and review, and an early multiplayer workspace.

    Main job
    Turn Slack requests and tickets into tested pull requests and shared agent work.
    Compare details
    Reported use
    Internal operation Builderbot is rolled out internally; Buzz remains early-stage dogfood.
    Human checkpoint
    People steer work, review changes, and approve consequential actions.
    Access
    Mixed Goose, Builderbot, and Buzz have public repositories; internal services differ.
    Source window
    First-party reportObserved artifact
  5. Cloudflare · Internal AI engineering stack

    Cloudflare's AI engineering stack: a control plane for company-wide agents

    Cloudflare centralizes identity, model access, cost, company context, approved internal tools, write controls, and risk-based AI code review across thousands of repositories.

    Main job
    Give company agents one identity, context, tool, cost, and review control plane.
    Compare details
    Reported use
    Internal operation Company-wide internal system across thousands of repositories.
    Human checkpoint
    People own approval while risk rules can block serious findings.
    Access
    Internal The operating stack is internal; some underlying platform parts are public products.
    Source window
    First-party report
  6. Cognition · Devin

    How Cognition uses Devin to build Devin

    Cognition starts Devin from chats, tickets, alerts, schedules, and APIs, then connects implementation to review, CI repair, bug investigation, and reusable playbooks.

    Main job
    Move bugs, design audits, and routine fixes from event to reviewed pull request.
    Compare details
    Reported use
    Product with internal use Cognition publishes a detailed account of Devin building Devin.
    Human checkpoint
    People define playbooks, review changes, and own releases.
    Access
    Commercial Devin is commercial; the internal code and operating data are private.
    Source window
    First-party report
  7. Cursor · Cloud agents and security agents

    Cursor's agent factory: cloud computers, Bugbot review, and security loops

    Cursor combines isolated cloud coding agents, an eight-pass pull-request reviewer, and event-driven security agents with evidence and human escalation paths.

    Main job
    Build parallel pull requests with proof artifacts, then run event-driven review.
    Compare details
    Reported use
    Product with internal use Customer-facing agents and internal review and security lines.
    Human checkpoint
    People review artifacts and own merge, severity, and production decisions.
    Access
    Mixed Cursor is commercial; a security-automation reference implementation is public.
    Source window
    First-party reportObserved artifact
  8. Dropbox · Nova and Deflaker

    Dropbox Nova: agents propose changes while deterministic systems control proof

    Dropbox runs several coding agents through one internal cloud platform, with exact code snapshots, declared checks, bounded retries, one branch, and publication outside the agent.

    Main job
    Run interactive coding and repeatable repair jobs on one shared agent platform.
    Compare details
    Reported use
    Internal operation Internal platform used for CI repair, flaky tests, and migrations.
    Human checkpoint
    The surrounding workflow controls tests, branches, publication, and retry limits.
    Access
    Not publicly available Nova and Deflaker are internal systems described in a public engineering article.
    Source window
    First-party report
  9. GitHub · Copilot coding agent

    How GitHub uses Copilot coding agent to build github.com

    GitHub assigns issues in its core repository to Copilot, receives first-pass pull requests, and keeps the decision to merge, revise, or close with human engineers.

    Main job
    Turn assigned issues into draft pull requests inside the normal repository workflow.
    Compare details
    Reported use
    Product with internal use Public product used inside the core repository that builds github.com.
    Human checkpoint
    A different person approves the work; repository and CI rules still apply.
    Access
    Commercial Copilot coding agent is a commercial GitHub capability.
    Source window
    First-party report
  10. Linear · Coding Sessions

    Linear coding sessions: from product issue to reviewed pull request

    Linear carries issue and customer context into a managed coding sandbox, then returns code and can include checks and visual evidence in the same workflow for review.

    Main job
    Use issues and team chat as dispatch and review surfaces for coding agents.
    Compare details
    Reported use
    Product with internal use A recent public product with concrete internal design and triage examples.
    Human checkpoint
    People review the diff, proof artifacts, and merge decision in Linear.
    Access
    Commercial Coding Sessions is a commercial Linear capability; the implementation is private.
    Source window
    First-party report
  11. LinkedIn · CAPT

    LinkedIn CAPT: turning company knowledge into executable agent playbooks

    LinkedIn gives existing agents approved internal tools and reusable playbooks, then uses verifier-driven loops for specialized production machine-learning work.

    Main job
    Turn company knowledge and internal tools into reusable, executable playbooks.
    Compare details
    Reported use
    Internal operation Internal system reported in use by more than 1,000 engineers.
    Human checkpoint
    Domain experts author and maintain the playbooks; engineers own results.
    Access
    Not publicly available CAPT is internal and built on the open Model Context Protocol.
    Source window
    First-party report
  12. Meta · Capacity Efficiency agents

    Meta's capacity agents: production signals become review-ready fixes

    Meta combines regression detection, profiling data, code history, reusable expert skills, agent implementation, and human review in one performance-efficiency line.

    Main job
    Turn production performance signals into diagnosed, review-ready code changes.
    Compare details
    Reported use
    Internal operation Specialized internal line for performance regressions and opportunities.
    Human checkpoint
    Engineers review, deploy, and remain responsible for the proposed fix.
    Access
    Not publicly available The agent platform and its skills are internal.
    Source window
    First-party report
  13. OpenAI · Codex and Symphony

    OpenAI's Codex factory: harness engineering and Symphony orchestration

    OpenAI made one repository legible and testable by Codex, then used Symphony to turn project issues into isolated, continuously managed agent work.

    Main job
    Pull tasks from a project board into isolated agent workspaces and review loops.
    Compare details
    Reported use
    Internal operation Internal teams plus one fully agent-written experimental product repository.
    Human checkpoint
    People prioritize work, set acceptance criteria, and validate outcomes; pull-request review may be agent-to-agent.
    Access
    Mixed Codex CLI and Symphony are public; internal repositories and operating data are not.
    Source window
    First-party reportObserved artifact
  14. Ramp · Inspect

    Ramp Inspect: how a background coding agent became factory infrastructure

    Inspect pairs a prepared remote development environment with company context, executable feedback, human acceptance, and specialized production lines.

    Main job
    Run background coding with full company context in prepared remote environments.
    Compare details
    Reported use
    Internal operation Internal operating system with multiple specialized factory lines.
    Human checkpoint
    People review and land changes; automated lines stop at review-ready work.
    Access
    Not publicly available Inspect is internal; an independent inspired implementation is public.
    Source window
    First-party reportObserved artifact
  15. Replit · Internal agent fleets

    Replit's self-driving company: manager agents launch parallel work

    Replit gives employees manager agents that can start several agents in parallel, while the wider system checks results and routes exceptions back to people.

    Main job
    Launch parallel agent work across company systems, check results, and learn from feedback.
    Compare details
    Reported use
    Product with internal use Customer product plus a first-party internal agent-of-agents system.
    Human checkpoint
    People set policy, review risk-rated changes, and judge product outcomes.
    Access
    Commercial Replit Agent is public and commercial; internal orchestration and data are private.
    Source window
    First-party report
  16. Shopify · World, River, and Aquifer

    Shopify River and Aquifer: one durable platform for many software agents

    Shopify runs River in public Slack threads while rolling Aquifer out profile by profile as a shared foundation for sessions, sandboxes, controls, and agent profiles.

    Main job
    Run public-to-the-company agent sessions and specialist lines on one durable substrate.
    Compare details
    Reported use
    Internal operation Large internal operation; the shared Aquifer substrate is still rolling out.
    Human checkpoint
    People join visible threads, redirect work, and review pull requests and findings.
    Access
    Mixed Roast is open source; River, Aquifer, and Dispatch are internal.
    Source window
    First-party report
  17. Spotify · Honk and Fleetshift

    Spotify Honk: adding an agent to a software factory that already worked

    Spotify kept existing systems for choosing repositories, coordinating work, checking results, and merging safe changes, then used Honk for complex transformations.

    Main job
    Apply agent judgment to repository-wide maintenance and data migrations.
    Compare details
    Reported use
    Internal operation Internal agent work built on Spotify's existing fleet-management systems.
    Human checkpoint
    People initiate work and own review, merge, and unresolved edge cases.
    Access
    Mixed Honk is internal; Fleetshift is available as a managed Backstage product.
    Source window
    First-party report
  18. Stripe · Minions

    Stripe Minions: how developer infrastructure became an AI software factory

    Stripe combines ready-to-use development computers, scoped context, curated internal tools, repeatable workflow steps, limited automated build-and-test repair, and two human review gates.

    Main job
    Turn one unattended delegation into a checked branch for two human review gates.
    Compare details
    Reported use
    Internal operation High-volume internal background-agent operation.
    Human checkpoint
    The initiator inspects the result and another engineer reviews the pull request.
    Access
    Not publicly available Minions and its Stripe-specific infrastructure are internal.
    Source window
    First-party report
  19. Tempo · Custom Agents

    Tempo's software factory: turning product signals into review-ready PRs

    Tempo routes product and error signals through human prioritization, then Custom Agents work in parallel, open pull requests, and return decisions to people.

    Main job
    Turn product and error signals into prioritized, parallel agent work and review.
    Compare details
    Reported use
    Early rollout Concrete public examples; organization-wide operating scale is not stated.
    Human checkpoint
    People prioritize the queue and review the pull requests or media output.
    Access
    Commercial Tempo's agent products are commercial; internal operating details are limited.
    Source window
    First-party report
  20. Uber · Software Factory

    Inside Uber's AI software factory: measuring cost and quality at scale

    Uber connects coding-agent adoption with reusable skills, managed workloads, quality signals, and cost and quality per completed outcome.

    Main job
    Coordinate coding agents, shared context, managed jobs, checks, and result economics.
    Compare details
    Reported use
    Internal operation Company operating model described across multiple factory layers.
    Human checkpoint
    People define valuable work and own quality, acceptance, and release.
    Access
    Not publicly available Uber's factory infrastructure is internal.
    Source window
    First-party report
  21. Vercel · eve

    Vercel eve: the shared infrastructure behind more than 100 production agents

    Vercel standardized file-based agent definitions, resumable runs, sandboxes, permissioned connections, activity records, repeatable behavior checks, previews, and rollback in one platform.

    Main job
    Give many production agents durable runs, sandboxes, permissions, traces, and evals.
    Compare details
    Reported use
    Internal operation Shared framework reported across more than 100 internal production agents.
    Human checkpoint
    Engineers own release controls, executable guardrails, and production outcomes.
    Access
    Mixed The eve framework is open source; Vercel's internal agents and connections are private.
    Source window
    First-party reportObserved artifact

Read the company cases

Each case follows the workflow, the proof, the human checkpoint, the public artifact, and the limit of the available evidence.

  1. Cognition · Devin

    How Cognition uses Devin to build Devin

    Cognition starts Devin from chats, tickets, alerts, schedules, and APIs, then connects implementation to review, CI repair, bug investigation, and reusable playbooks.