The dilemma
Inside a team, SoulCode was quick. Between teams, the work stalled.
Every team hit its own plan. The customer still waited. Dependencies hid in chats. SLAs were soft enough that nobody was ever late. Changes landed without notice. Escalations bounced across org lines until somebody senior enough got tired and made a call.
Busy looked like progress. Throughput said otherwise.
What was broken
Visibility. A dependency mentioned in a thread is not a dependency anyone owns. Most seams existed only in someone's memory of a conversation.
Commitment. Soft SLAs are not commitments, they are hopes. Without a clock, nothing at a seam is ever late, and nothing is ever prioritised either.
Notice. Changes shipped on one side and arrived on the other as breakage. The receiving team paid for a decision it never saw coming.
Escalation. With no single path, an urgent seam problem generated three conversations, no owner, and a resolution that depended on who happened to be free.
Quinn's playbook: write the pact
Interface contracts. Each seam documented once: inputs, outputs, SLAs, acceptance tests, rollback steps. If it cannot be written down, the seam is not understood well enough to depend on.
Driver and steward. One owner per seam, and a named partner on the other side. Two people, both accountable, neither able to point across the line.
Request to commit flow. Ask, estimate, agree, deliver, verify. Five steps, timestamped. The estimate is not the commitment and the delivery is not the acceptance.
Change windows. Published calendars and notice periods. Anyone can change anything, on the schedule everyone can see.
Integration tests. Pre-merge checks and live monitors sitting on the seam itself, not inside either team. The contract is tested the way the code is.
Single-path escalation. A co-written one-pager, one room, one call. If a seam problem cannot be solved by the driver and the steward, it goes up once, together.
The shift at SoulCode
The first contract took an afternoon to write and settled an argument that had been running for two months. Both sides had been right about what they were owed and wrong about what they had promised.
Within a cycle, escalations dropped because most of what had been escalating was ambiguity, not conflict. Rework fell because changes stopped arriving as surprises. The teams did not get faster individually. The work between them stopped losing days.
Leadership takeaways
Clean seams cut rework, cool escalations, and make speed repeatable.
Autonomy only moves fast when handoffs are visible, owned, and testable.
An undocumented dependency is not a dependency. It is a future incident.
Teaser for Episode 27: The Capacity Ledger
Seams are documented and the handoffs hold. Then Quinn looks at the week itself. The plan says green and the calendar says crowded, and nobody has counted what actually happens between them.
