FACTORY AI / AN INTERACTIVE GTM EXPERIMENT
FACTORY
FLOOR.
I gave a software factory
a fictional backlog.
Pick a ticket. Follow the work. See why Factory’s ambition reaches well beyond writing code.
Start productionBy Stephen Stilwell. Independent, unsponsored spec work. Illustrative interfaces, not a live Factory demo.
A job that keeps returning.
Factory documents agents that plan, build and validate changes across the software lifecycle. Sources below
Factory reports hundreds of thousands of developers using its platform. Company statements, not independently audited usage. Funding announcement · Automations release
01 / THE BACKLOG
Give the work
somewhere to go.
Select a fictional ticket to start its workflow. This compressed simulation explains the idea; its timing and results are invented.
- Read
- Plan
- Build
- Test
- PR
00 / INTENTRefresh an expired token without losing the signed-in session.
Select a ticket or run this example to follow the work.
A task needs acceptance criteria.
For this example: refresh once, preserve the session, and reject revoked credentials.
Ready to explore. No code runs and no repository is connected.
Where would shipping fit?
A pull request is a proposed change. Your team still needs review rules, passing checks and a release policy. This demo stops at that boundary. Large migrations also need scope, rollback plans and domain expertise.
MEET THE DROID
A task goes in.
Engineering work comes out.
Droid is Factory’s software-development agent. With the right context, tools and permissions, it can work through a task and return changes for review. Factory documentation
Find the context.
Inspect code, instructions and connected project information.
Choose an approach.
Break the request into work that can be checked.
Make the change.
Create and modify code inside the configured environment.
Check the result.
Run relevant tests and inspect what fails.
Follow the failure.
Investigate errors and CI problems using available evidence.
Make it reviewable.
Return a diff or pull request and verification notes.
02 / AUTOMATIONS · GENERALLY AVAILABLE 30 SEP 2026
The work returns.
The instructions shouldn’t have to.
Factory Automations start Droid workflows on schedules or events. Choose the instructions, model and execution environment. Read the release
Illustrative configuration explorer. Nothing is created in Factory.
Execution identity and permissions matter. Webhook triggers are currently marked Private Preview in the documentation; availability depends on your organization. This is not a claim that every illustrated workflow is a one-click template.
ONE PLATFORM / MANY SURFACES
Meet the work
where it happens.
Sessions, delegation and connectors serve different jobs. Select a surface to see the distinction.
The Factory App brings sessions and repository context to your machine.
Official documentationTHE SOFTWARE FACTORY / A CHANGE IN OPERATING MODEL
From asking again
to designing the system.
- Developer asks a tool for help
- Developer moves the output forward
- The next task needs another prompt
- Humans set intent, scope and policies
- Configured workflows execute appropriate work
- People inspect outcomes and handle exceptions
My interpretation: the engineering job expands to designing and supervising the production system. Factory describes autonomy as gradual, shaped by the organization’s readiness and control requirements. Its vision of self-improving software is an ambition to evaluate, not a guarantee that every repository can run itself. Factory 2.0 announcement
03 / THE ENTERPRISE CASE
Capacity matters.
Control does, too.
The buying conversation should include the work an agent can take on and the operating discipline around it. These are evaluation priorities, not promised results.
Velocity
Move suitable work toward review without manually initiating every step. Measure accepted changes and rework, not generated lines.
Fewer interruptions
Start with recurring maintenance and triage. Check whether automation actually gives engineers longer stretches of focused time.
Legacy modernization
Factory describes migrations and debugging among enterprise use cases. Start with a bounded change and explicit regression checks. Customer project examples
Governance
Business and Enterprise plans document access, model and autonomy controls, plus audit trails. Confirm the policy and deployment fit for your organization. Organization controls
Model flexibility
Factory supports model choice; Router can select a model for a session and escalate if it struggles. Evaluate quality and total cost on your workload. Factory Router
WHO IS ALREADY BUILDING THIS WAY?
Enterprise names.
A serious operating question.
Factory identifies these organizations in its September 15 funding announcement. Names are editorial references, not endorsements of this page. No customer outcome or deployment scope is inferred.
WHY FACTORY NOW / VERIFIED MILESTONES
The story is moving.
$50M Series B
A $300M valuation, with Droids available across multiple interfaces.
Series B announcementFactory 2.0 publication
The dedicated announcement sets out the software-factory vision.
Dated product announcement$200M financing / $5B valuation
Total funding exceeds $400M, according to Factory.
Financing announcementOpenAI marketplace
Eligible enterprise customers can apply part of their OpenAI commitment toward Factory.
Marketplace announcementAutomations general availability
Recurring and event-triggered work becomes the timely campaign hook.
Release announcement
Date note: the September financing post refers to Factory 2.0 in April; its dedicated announcement is dated June 15. This timeline uses the publication date and does not assert an exact first-release date.
THE GTM IDEA / MY INDEPENDENT POSITIONING
The product is technical.
The story can be clear.
Engineers understand repositories, CI and model routing. Buyers still need to connect those mechanics to capacity, reliability, cost and control. I would lead with a familiar job, show the handoff, then earn the deeper technical conversation.
Autonomous software-development agents
Assign the outcome.
Review the work.
Event-triggered agent automation
The recurring job
already has instructions.
Multi-model architecture
Change the model.
Keep the workflow.
Original spec copy by Stephen Stilwell, not Factory’s official slogans or performance claims.
04 / CAMPAIGN LAB
Three ideas.
One job: make it click.
Proposed creative tests. No ads have run and no campaign results are claimed.
Your team
signed off.
The queue
didn’t.
What if your software factory had a night shift?
Audience, distribution & test
Audience: CTOs, VPs of Engineering, engineering managers.
Distribution: a short LinkedIn film moves from the evening queue to a morning review. Link to one scoped maintenance pilot.
Measure: qualified pilot requests and reviewable outputs. Ask whether “night shift” sets realistic expectations about supervision.
Same work.
Every week.
New plan.
Describe it once.
Trigger it when it returns.
Audience, distribution & test
Audience: engineering managers, DevOps and platform teams.
Distribution: a carousel shows the repeated ticket, the instructions and the review step. Follow with a creator walkthrough.
Measure: workflow-example completion and qualified automation conversations. Check whether the trigger is understood.
Less waiting
to start.
More ready
to review.
Build a path for the work
you already know is coming.
Audience, distribution & test
Audience: CTOs, VPs of Engineering and developer productivity teams.
Distribution: an interactive backlog in a targeted account campaign, followed by a workshop on task eligibility.
Measure: qualified discovery meetings and scoped pilots. Watch for a review bottleneck before expanding volume.
05 / THE SALES DESK
Same platform.
Different first question.
I would prospect into an identifiable workflow problem, then qualify its scope. Choose the person on the other side of the conversation.
- PAIN
- The roadmap competes with maintenance, migrations and reliability work.
- HOOK
- Which work could move toward review without waiting for someone to pick it up?
- DISCOVERY
- Which recurring jobs consume capacity, and which have clear acceptance criteria?
- VALUE HYPOTHESIS
- An execution layer for scoped engineering work, with your team setting policies and judging outcomes.
Independent prospecting concepts. Validate fit in discovery; no promised savings or headcount reduction.
MY TAKE
Writing code was
the opening act.
What interests me is the work between prompts. The recurring issue. The neglected migration. The failed build someone still has to investigate.
When AI becomes infrastructure that can keep those jobs moving, the adoption question gets bigger.
DEVELOPER + COPILOTORGANIZATION +
AUTONOMOUS SOFTWARE SYSTEM
That is a much bigger product story. It also demands better tests, clearer ownership and honest measurement. That’s the story I’d want to tell.
ABOUT THIS EXPERIMENT
Great software still
needs distribution.
I’m Stephen Stilwell. I’ve worked in engineering, product ownership and business development. Useful in Practice is where I turn that experience into product explanations, useful demonstrations and marketing experiments.
This page is my independent spec project. Factory did not commission, sponsor or endorse it. I am not affiliated with Factory.
Building something complicated? Let’s talk.SOURCE DESK / CHECKED 02 OCT 2026
Evidence behind the experiment.
Product descriptions come from Factory’s official pages and documentation. I did not test a Factory account, run these tickets in Droid, or measure customer results. The simulations, sales messages and campaign concepts are original editorial work.
Design reference: Factory’s official site uses restrained monochrome surfaces, orange accents, strong sans-serif hierarchy, monospace labels and functional diagrams. This page borrows those principles using the publication’s existing system typography. All diagrams are illustrative, not Factory UI screenshots. No benchmark, ROI, pricing or security-certification claims are made here. Editorial policy.