Build something new
Start a thread from what the business needs. Your owners agree the acceptance criteria in plain language, and agents build only to them.
AI agents build it. Every line traces back to a spec you approved.
CrewPE runs each piece of work as a thread, from intent to merged code. Agents lead every stage. Your people lock each one before the next begins.
Thread kind: new feature. Started 14 Oct by the product ownerPolicyholders can track a claim online
Buildmerge approved
Stages
Stage 5 of 5, complete
Build, check and merge each story
Coding agents build each story in your repositories and check it against every acceptance criterion. A person approves the merge.
Built on Intent BriefFunctional SpecADRStories
CLM-118Status service reads claims without changing them
Specs, criteria, architecture, repo impact, agent execution, validationDone
AC1Reads claim status without any write accessPassed
AC2Stage changes appear within 5 minutesPassed
AC3Adjuster notes never leave the servicePassed
AC4Payout results match the claims systemPassed
Validation passed in 2 repositoriesMerge approved by the engineering lead
Technology partners
Clients and network
A recent migration
A specialized debt collection and resolution suite moved from a legacy .NET stack to a modern web stack: 45+ screens in 8 days, for 60% less than the vendor quote, with the existing system running the whole time.
- System
- Debt collection and resolution
- Scope
- 45+ screens
- Team
- EnWithAI on CrewPE
Two ways in. The same proof at the end.
Whether you are building something new or moving an old system, the work runs as a thread with the same locks and the same evidence.
Modernize what you run
A modernization thread runs preflight, assessment and a plan. Business rules are captured with evidence before anything is rewritten, then the system moves in waves while the old one keeps running.
Agents do the work. People hold the keys.
The crew writes the spec, plans the change, writes the code and checks it. People approve what matters, and every approval is kept on record.
Writes the spec
Turns the story into acceptance criteria for every repo it touches.
DraftsSpec and acceptance criteria
Plans the change
Maps the repos, APIs, events and tables the change touches, and scopes the work to each repo.
DraftsA plan per repo
Writes the code
Code and tests in every repo the story touches, each pull request linked to the approved criteria.
DeliversLinked pull requests
Checks every criterion
A separate check that cannot change the code. Failures go back with the reason, up to the loop limit you set.
DeliversCheck report
Only people can
Each approval names who gave it, in what role, and when.
- Set the intentProduct manager
- Approve the specEngineering lead
- Decide at the loop limitEngineering lead
- Approve the mergeEngineering lead
- Approve the tracker updateEngineering lead
Audit logLogged
- Who
- Engineering lead
- Action
- Approved spec v2
- When
- 15 Oct, 09:20
- Role
- Refinement approver
Run it yourself, with us beside you, or hand it over.
One platform under all three. Choose who does the delivery, and change it as your team grows into it.
Your teams use CrewPE directly. Start self-service, then move to an enterprise subscription with your own deployment and security terms.
- Best for
- Teams ready to run governed AI delivery themselves
- Engagement model
- Sign up and start on your own, then move to an enterprise plan when you are ready
It works inside your rules.
Built for your security review. CrewPE starts in a test environment, and new code ships only through your own process.
Read-only to start
CrewPE reads the code and records behavior in a test environment. Any write access needs a separate sign-off.
Your release process
Code lands in your repository and ships through your own pipeline. Nothing goes live outside it.
Agreed data handling
Where recordings are kept, what is masked and for how long is agreed before capture begins.
What you can take to your board.
Every engagement measures five outcomes on your own system: first a baseline, then the same work with CrewPE.
Proof-of-value reportBaseline, then with CrewPE, on your system
NumberOutcomeSourceMeasured as
01ReworkChanges sent back because context was missing.Reviews and ticketsShare of changes sent back
02Review cyclesReview rounds before a change is accepted.Pull request historyRounds per accepted change
03Spec to acceptedDays from spec approval to the accepted change.Timed by CrewPEDays, spec approved to change accepted
04Rules preservedBusiness rules that still behave the same after the change.Parity testsShare of rules passing parity
05Cost per accepted changeEngineering time and AI spend, divided by accepted changes.Time and usage recordsHours and AI spend per change
Bring one real change. See it traced from rule to release.
Start with a 45-minute call about your system: how it is built, where its business rules live, and a sensible first step.
CLM-118Traced from rule to release
- Intent BriefLocked by the product owner
- Functional SpecLocked by you
- Architecture decisionLocked by the engineering lead
- Four storiesChosen by you
- MergeApproved by the engineering lead
From the example thread at the top of this page.