Decide on rollout with evidence from your own delivery.
The Factory Pilot runs a Software Factory with AI agents on one real value stream for about three months and measures what changes in your delivery.
- For
- Organizations whose assessment recommends a pilot
- Outcome
- A running Factory on one real value stream, with delivery evidence
- Timing
- About three months
- Price
- Often starts around EUR 150,000 as a guide. Your assessment sets scope and price.
- Next step
- A rollout recommendation and a named internal owner
- About three months
- scope set by your assessment
How each pilot run learns
- 01
Select the value stream
a real product or engineering flow
- 02
Build the slice
intent, context, agents and human steering
- 03
Run and observe
in real work, with delivery and quality signals
- 04
Codify and decide
manual fixes become rules or tests
You get a working slice of the Factory and the data from it
The pilot puts your readiness roadmap to work before you scale it: small enough to learn from, real enough that the results count.
Running Factory
Orchestration, repository integration, work isolation, quality gates, review and human handoff, working on a real value stream.
Delivery data
Baseline and pilot signals for DORA with lead time from idea to first users, rework rate, use of what ships and cost per change, plus the outcome that matters in your context.
Operating constraints
The technical and organizational limits that real use uncovers, documented before you roll out further.
Scale recommendation
What to scale, what to fix first, who should own it and how the rollout should run.
Choosing the right pilot project
Choosing the right pilot project
The pilot is only as useful as the project it runs on. On the wrong one, you learn little you can use.
-
A real, active project
Work already on the roadmap that would happen anyway. A sandbox mainly proves the sandbox works.
-
Urgency and a mandate from management
Real urgency behind the work, and management that has backed the pilot with its mandate. Without that backing, the pilot tends to lose priority to whatever escalates next.
-
Volunteers, junior to senior
People who opt in, at every experience level. The mix brings in perspectives from across the team.
-
Access to codebase and CI
The team works in the real repository with the real pipeline. Anything built alongside is hard to carry back.
-
What we rule out
Toy sandboxes, forced participation, all-junior teams and work under crisis pressure. Each makes the result hard to read and hard to transfer.
How the pilot runs: four steps
-
01
Select the value stream
Pick a real product or engineering flow with enough value, evidence and organizational backing to learn from.
-
02
Build the slice
Wire up intent, context, agents, repositories, CI, review, deployment and human steering.
-
03
Run and observe
Put the Factory to work on real tasks. Record delivery, rework, use, cost, failure and handoff signals.
-
04
Codify and decide
Turn manual fixes into rules or tests, document the limits and recommend the next rollout step.
What the pilot builds for your value stream
What the pilot builds for your value stream
We cut the slice to fit your value stream, so the pilot builds what that stream needs.
Work tracking and state
Where work lives, what state it's in, and how an agent picks it up and hands it back.
Context and orchestration
The layered context an agent reads before starting, and the rules that assign work to agents.
Repository and CI integration
Agents work in branches, open pull requests and go through the same pipeline as everyone else.
Tests and review
Executable checks and the review path every change takes before merging. Agent and human review happen on the same change.
Tracing and cost visibility
What a run did, what it cost and where it failed. Without traces, the pilot ends in opinions.
Human and agent handoff
Where a person decides, where an agent carries on alone, and how handover works both ways.
How we measure the pilot: delivery, rework, use, cost
A Factory earns its name by improving the delivery system and learning from every run.
- Lead time from idea to first users, failures and recovery
- DORA
- Changes redone before they reach users
- Rework rate
- Target users active with what shipped after 30 days
- Use
- Tokens, agent time and review time
- Cost per change
Prove it in one value stream first
Big-bang Factory program
Infrastructure and governance are designed before there is a working reference.
- Assumptions harden into architecture.
- Teams wait for a central platform.
- Business outcomes stay abstract.
- Lessons arrive late and cost a lot.
One real value stream
A bounded slice builds the case for wider investment.
- Architecture is tested against real constraints.
- The internal owner emerges through use.
- Quality and handoff are tested early.
- Rollout scope follows the evidence.
What you have at the end of the pilot
What you have at the end of the pilot
Something running, and a decision you can defend inside your own organization.
A running reference for your context
A working slice in your codebase, plus the artifacts agreed at the start. It stays with you, and your team can keep developing it.
Evidence against an agreed baseline
The baseline is agreed before the pilot starts; afterwards we read the same delivery and quality signals.
Scale, adjust, or stop
A reasoned recommendation, and a named person in your company who carries whichever path you choose. Scaling without an internal owner doesn't hold.
From 2 engineers to 20 who work with agents daily
Pilot
We went from 2 engineers playing with agentic coding to 20 using it daily. We created 17 skills and 9 agents available to every engineer. Skeptics now embrace it, with a Slack channel dedicated to sharing learnings and complaints when Claude goes down.
Chief Technology Officer
Pricing follows the evidence
A pilot often starts around EUR 150,000. Use that figure as a guide for your budget: the assessment defines the value stream, baseline and measurement plan, and your offer prices the pilot from them.
Questions about the Factory Pilot
Can we skip the Factory Readiness Assessment?
The standard route scopes the pilot from the assessment. If you already have equivalent evidence, we evaluate it. Either way the pilot needs a named outcome, baseline, constraints and measurement plan.
Which value stream works best?
Real work with visible value, a mix of AI-curious and sceptical people, enough technical complexity to expose constraints, and leadership support. In a large organization, one pilot per representative area, such as SAP, e-commerce or infrastructure, shows what holds overall.
A value stream on a legacy system can double as the pilot of a legacy modernisation.
Legacy ModernisationWho owns the Factory after the pilot?
Internal ownership is part of the design. The pilot identifies the people, platform-team setup, operating model and handover work a wider rollout needs.
Who delivers the pilot?
Stefan Munz and Benedikt Stemmildt own your pilot and remain your contact throughout. Senior engineers from our delivery partner re:cinq build it with your team, otherwise engineers from our network of more than 70 practitioners. We agree the team with you up front.
Where are the procurement and legal details?
Everything your purchasing and legal teams need is on one page.
AboutProcurement and legalBuild one working reference before you scale
Use the readiness assessment to set the right value stream, outcome, evidence and scope for your Factory Pilot.