Internal pilot · NodeOS used by its own team
How we gave our AI development workflow a memory
For about three months we have used NodeOS in our own software development with KIRO, GitHub and a customer-facing workflow. This is what we observed.
Internal product-use case. Not an independent customer testimonial and not a measured productivity study.
The workflow
- 01
A request or bug is registered
Business stakeholders register it where they can follow it.
- 02
Development work is prepared
The task is placed in the project hierarchy with related work and decisions.
- 03
KIRO retrieves historical context
Previous tasks, decisions and documentation are fetched before investigation.
- 04
The agent investigates and implements
Within its authorized environment, verifying current code.
- 05
Progress and evidence are recorded
Work sessions, checkpoints and evidence stay with the task.
- 06
The owner retrieves previous decisions
Earlier choices remain searchable instead of living in someone's memory.
- 07
Deferred ideas stay available
Improvements found along the way are kept for future planning.
Example company · Active development
Example company – Development project
Overview of work items, priorities and deliveries.
Latest release
Demo 0.2 · Welcome and overview
Next planned release
Demo 0.3 · Notifications
Current overview
12
Open items
3
Planned
4
In development
2
Awaiting customer
1
Ready for release
Status right now
8
Delivered in the last 30 days
2
Awaiting you
3
Ready for test
4
Critical & high
Active initiatives
A better customer experience
2 of 6 subtasks completed
Reports and export
1 of 4 subtasks completed
Latest release
Demo 0.2 · Welcome and overview
A new welcome page and a unified overview have been delivered in this fictional project.
Latest activity
New comment on DEMO-104: Notifications
20 minutes ago
DEMO-103 moved to Ready for test
1 hour ago
Demo 0.2 · Welcome and overview recorded
Yesterday
Ongoing work
In development
- In development
Notifications when cases change
Next on the roadmap
- Planned
Choose which messages you receive
Architecture decisions
Notifications follow user preferences
DEC-DEMO-01
Reports use the same terminology as the portal
DEC-DEMO-02
The welcome page shows the next agreed step
DEC-DEMO-03
New items
- In development
DEMO-104 · Customer portal
Notifications when cases change
- Ready for test
DEMO-103 · Reports
Export the project overview
- Planned
DEMO-102 · Customer portal
Choose which messages you receive
- Released
DEMO-101 · Customer portal
Unified welcome page
How an AI development agent starts its next task with previous context — illustrative workflow.
01 · New task
DEMO-104 · Notifications when cases change
The agent receives the task with the context NodeOS has recorded for it.
- Project: Example company – Development project
- Related task: DEMO-102 Choose which messages you receive
- Component: Notifications (recorded)
- Documentation: Notification note in the project
- Current decision: DEC-DEMO-01 Notifications follow user preferences
One record, three perspectives
- Customer: understands development progress.
- Owner: retains control and organizational history.
- Developer / agent: uses previous work to guide new development.
The workflow illustration uses synthetic data, not real development evidence.
What each audience gained
Customer perspective
Stakeholders could follow incoming bugs, development requests, task progress and deliveries.
Developer / AI agent perspective
KIRO used previous history, decisions and documentation to inform subsequent work, making investigations more focused instead of repeatedly reconstructing context.
Development owner perspective
Implementation work, decisions, investigations, outstanding tasks, deferred ideas and progress history were preserved.
Observed vs. expected
Observed in internal use
- Stakeholders followed progress without asking developers for status
- Agent sessions started from recorded tasks, decisions and documentation
- Deferred ideas remained findable later
Expected, not measured
- Shorter onboarding of new agent sessions
- Less repeated investigation over time
- Clearer accountability for delivered work
We have not measured time savings or productivity, and we do not claim any figures.
Build on what came before.
Give your people a shared overview. Give your agents a better starting point.