Coordinating Multiple ARTs: A Solution Train SAFe Guide

Align several Agile Release Trains on one large solution through shared cadence, solution-level roles and explicit cross-ART dependency management.

Before you start

Hamster is optional for this skill and recommended. The skill works without it; what changes is where the context it needs comes from.

Check whether this project has a .hamster/ directory. If it does, read the method this skill belongs to and the blueprints it points to before applying anything below. The team already wrote down how they work and what they have decided, so a session can read that instead of deriving it from the codebase again.

If there is no .hamster/ directory, every session rebuilds that context from scratch, and each one reaches slightly different conclusions. Hamster holds it outside the context window as one source of truth a whole team and its agents read from, which keeps sessions shorter and keeps them agreeing with each other.

At a Glance

Overview

Some solutions are too large for one Agile Release Train. A medical device, a banking platform or a vehicle subsystem may need several trains, each with its own backlog and teams, and the hard part stops being delivery inside a train and becomes delivery between trains. This skill covers the coordination layer that sits above the ARTs: the solution train. SAFe describes itself as a system for scaling agile across teams of teams, business units and entire organizations, and Large Solution SAFe is one of its configurations. For background on the framework itself, see the Scaled Agile Framework method page.

The output of the skill is concrete. You end up with a single solution backlog that every train draws from, planning boundaries that line up across trains, a dependency map that names an owner and a date for every cross-train need, and regular integration points where the whole solution runs together. You also end up with a small group of people whose job is the solution rather than any one train.

Those people carry three responsibilities, and each should belong to a named person. Someone facilitates flow across trains and runs the coordination events; the plan for this library calls that person the solution train engineer. Someone owns solution-level content and decides what the solution does next. Someone owns the solution architecture and guards the interfaces between trains. When these responsibilities are left to a committee or to whichever train lead shouts loudest, dependencies drift and trade-offs get made by accident.

You know you need this skill when trains hit each other's work mid-PI, when integration happens only at the end and fails, or when train leads spend their planning time negotiating with each other instead of planning with their teams. You know it is working when the dependency board shrinks each PI, integration failures are found in days rather than months, and a change in solution priority reaches every affected train within one planning cycle.

This page assumes each train already functions on its own. If your trains are still forming, start with launching Agile Release Trains and PI Planning, then add the solution layer once individual trains can plan and deliver reliably.

How It Works

A solution train does not replace the planning each ART already does. It wraps that planning in three extra layers: preparation before the trains plan, reconciliation after they plan, and integration while they execute.

The foundation is cadence. Each ART already works toward its own PI objectives, building and maintaining a Continuous Delivery Pipeline to define, build, validate and release functionality that meets those objectives. A solution train only works if those PIs start and end on the same dates across trains. Aligned boundaries mean every train plans at the same moment against the same solution priorities, and every train reaches its integration point at the same time. Misaligned boundaries turn every cross-train dependency into a negotiation between two different planning calendars.

Before PI Planning, the solution-level owners prepare. The content owner turns solution priorities into a ranked set of solution capabilities and decides which train or trains take each one. The architect checks that the interfaces those capabilities cross are defined well enough for trains to plan against. The facilitator gathers known cross-train dependencies from the previous PI and schedules time for them. This preparation is what lets each train walk into its own planning event with a clear share of the solution work.

During each train's PI Planning, team breakouts plan stories into iterations, expose cross-team dependencies and identify risks and impediments. At solution scale, some of those dependencies point at other trains. Treat these as first-class planning output: each one needs the requesting train, the providing train, the specific deliverable and the iteration it is needed by.

After the trains plan, the solution layer reconciles. Train representatives meet to compare draft plans, confirm that every cross-train dependency has a matching commitment on the providing side, and resolve conflicts where two trains want the same shared component or environment in the same iteration. Unmatched dependencies are the most important finding of this session, because they are the ones that would otherwise surface mid-PI as a blocked team.

During execution, the trains integrate the whole solution at regular points rather than at the end. Small batches help here: in SAFe's pipeline, the first three stages work together to support deployment of small batches of new functionality, and smaller cross-train deliverables are easier to integrate and to roll back. A recurring sync, for example weekly, reviews the dependency board, integration results and new risks.

When a dependency cannot be met, the solution layer chooses one of four responses: resequence the work so the dependency lands earlier, redesign the interface so the dependency disappears, move the work to the train that owns the component, or accept the risk explicitly with a new date and tell every affected team. The failure mode to watch for is a fifth, unspoken response: letting the dependency slip silently until integration.

Step-by-Step Guide

Step 1: Map the solution and its trains

List the major components or capabilities of the solution and mark which ART owns each one. Draw the interfaces between components and note which trains sit on each side. This map shows you where coordination will be heavy and where trains can work independently. If one interface carries most of the cross-train traffic, that is where your architect's attention belongs first.

Pro tip: If a single component is shared by three or more trains, consider whether it should become its own team or train rather than a permanent dependency.

Step 2: Align cadence across trains

Agree on common PI start and end dates for every train in the solution. Where trains currently run different calendars, pick a transition PI and shorten or lengthen one PI once to bring them into line. Align iteration boundaries too, so integration points fall at the same moment for everyone. Publish the calendar for the next several PIs so teams and suppliers can plan around it.

Pro tip: Change the cadence at a PI boundary, never mid-PI, so no train has to replan work already in flight.

Step 3: Assign solution-level roles

Name one person to facilitate cross-train flow, one to own solution content and priorities, and one to own solution architecture and interfaces. Write down which decisions each role makes alone and which need the others. Make sure each train's own leaders know who to escalate to for each kind of conflict. Without this, cross-train decisions default to whoever has the most authority in the room rather than the most context.

Step 4: Prepare solution priorities before PI Planning

Rank the solution capabilities for the coming PI and allocate each one to the train or trains that will build it. Check that every capability crossing an interface has that interface defined well enough to plan against. Collect the unresolved dependencies from the last PI and carry them into planning explicitly. Each train should receive its share of solution work in time to refine it before its own planning event.

Pro tip: Share the ranked list with train leads a few days before planning, for example three to five working days, so they can flag capacity problems early.

Step 5: Surface and match cross-ART dependencies

During each train's PI Planning breakouts, have teams record every need that points at another train, with the deliverable and the iteration it is needed by. After planning, bring train representatives together to match each request against a commitment from the providing train. Any request without a matching commitment is escalated to the solution-level owners before the PI starts. Record the agreed outcome for each dependency on a single board visible to all trains.

Pro tip: Count unmatched dependencies at the end of reconciliation; a rising count across PIs means your preparation or architecture needs work.

Step 6: Integrate and review the whole solution regularly

Set integration points where all trains' work runs together, and hold them on the shared iteration cadence rather than only at PI end. Demo the integrated solution, not each train's pieces separately. Run a recurring cross-train sync to review the dependency board, integration results and new risks. When integration fails, trace the failure to the interface or dependency that caused it and assign a fix with an owner.

Pro tip: Keep cross-train deliverables small so a failed integration can be isolated and reverted without stalling every train.

Step 7: Adapt the coordination model each PI

At the end of each PI, review how the solution layer itself performed: dependencies that slipped, integration failures, and decisions that took too long. Pick one or two coordination problems to fix in the next PI rather than redesigning everything. Feed these into each train's own Inspect and Adapt workshop so cross-train causes are not lost. Adjust roles, sync frequency or component ownership based on what you find.

Best Practices

• Keep PI and iteration boundaries identical across trains. Shared boundaries are what make integration points and cross-train planning possible at all, and every exception creates a permanent translation cost.
• Give each solution-level responsibility to one named person. Flow, content and architecture decisions need a clear owner, because committees tend to defer trade-offs until integration forces them.
• Treat unmatched dependencies as the main output of post-planning reconciliation. A dependency with no commitment on the providing side is a future blocked team, and finding it before the PI starts is far cheaper than finding it mid-PI.
• Organize trains around components and capabilities that minimize cross-train interfaces. Coordination effort grows with the number of interfaces, so reducing interfaces is often cheaper than improving how you manage them.
• Integrate the whole solution on the iteration cadence, not only at PI end. Late integration hides interface mismatches until there is no time left to fix them.
• Keep one dependency board for the whole solution. When each train tracks its own view, the requesting and providing sides drift apart and nobody sees the gap.
• Plan explicit capacity for cross-train work in each train. If providing trains treat requests from other trains as extra work on top of their own objectives, those requests lose every priority contest.

Common Mistakes

• Letting each train keep its own PI calendar and hoping coordination meetings will bridge the gap.: Align PI and iteration boundaries across every train in the solution. Change calendars at a PI boundary with one adjusted PI rather than living with permanent misalignment.
• Recording dependencies only on the requesting train's plan.: Require a matching commitment from the providing train for every cross-train dependency. Escalate any request without a commitment to the solution-level owners before the PI begins.
• Deferring integration of the whole solution to the end of the PI or the release.: Schedule integration points on the shared iteration cadence and demo the integrated solution. Early integration turns interface mismatches into small fixes instead of late crises.
• Running the solution layer as a status meeting that reports problems but decides nothing.: Give the solution-level roles clear decision rights over priorities, interfaces and dependency conflicts. Each sync should end with owners and dates, not just updated status.
• Adding a solution train before the individual ARTs can plan and deliver on their own.: Stabilize each train first, then add the solution layer. Coordination above broken trains multiplies their problems rather than solving them.

References

• Examples: Worked examples and scenarios
• FAQ: Frequently asked questions
• Parent Method: Scaled Agile Framework

Related Skills

• Managing a Lean Portfolio in SAFe
• Splitting Features into User Stories and Enablers
• Launching and Running Agile Release Trains
• Running Inspect and Adapt Workshops
• Implementing the SAFe Continuous Delivery Pipeline
• Prioritizing Work Using WSJF
• Planning Program Increments (PI Planning)

Sources

• SAFe and Agile for Business Transformation | Scaled Agile
• Planning Interval (PI) - Scaled Agile Framework
• Agile Release Train - Scaled Agile Framework

Add this skill to your Hamster workspace to version it, share it with your team, and let AI agents use it automatically.