A Sunday newsletter for senior and enterprise architects. One decision, one practical model, one tool you can use Monday. Written by Kevin Donovan. Join 2,500+ architects. One email every Sunday.
Share
Architecture alignment: turn agreement into decisions
David Johnston-Bell shared a useful place to start.
Writing 3 weeks into a Head of Architecture role, he described sitting down with his team leads to understand what they were actually doing. What needed to stop? What needed to change? What needed to start?
Before bringing a plan from his previous role, he wanted to understand this one.
I like the restraint in that. Experience gives you things to look for. It doesn’t give you the answers before you’ve listened.
His questions also stayed with me for another reason: they’re a useful way to find out whether an architecture decision has traveled beyond the meeting where everyone agreed to it.
The target architecture is endorsed. The meeting ends well. A month later, the same teams are funding the same work on the same roadmaps.
Nobody needs to have been insincere for that to happen.
We may have agreed on a direction without deciding what anyone would give up to get there.
This week we looked at that gap from 5 angles. Today I want to put them together, work through a decision, and give you a brief you can use with your own teams.
In this edition
Agreement, the second room, and the 4 names.
The full Stop-Change-Start Test, including what the blanks mean.
A shared-service decision: 3 options, their costs, and who pays.
When to reopen a decision, and when the answer is no.
Community Corner, and your One-Page Decision Brief.
Let’s get into it.👇
________
🧭
________
Agreement is the beginning
Getting people to agree on an architecture can be substantial work. Different histories, priorities, and definitions of the problem all arrive in the same room.
I’m not dismissing that effort.
But agreement doesn’t reserve capacity. It doesn’t move a launch, retire an integration, or assign someone the reconciliation job.
Those consequences have to land somewhere.
A team can support a shared customer service and still be measured on shipping its own product this quarter. A platform lead can support reuse and still have no funded migration work. Operations can welcome a simpler future and still have to keep today’s systems running.
Each position makes sense from where that person sits.
The gap is between the direction people endorse and the commitments their plans can carry.
Look for 3 pieces of evidence after the meeting: a plan that changed, something that stopped or lost capacity, and someone who accepted the cost.
If you can’t find them, you have somewhere useful to start the next conversation.
Agreement around the same map doesn't guarantee that anyone's route changes.
Find the second room
The design room chooses a direction. The planning room allocates the capacity to make it happen.
That second room might be a portfolio review, a product planning session, an executive discussion, or a team lead deciding what comes out of the next sprint.
It doesn’t need a particular title. It needs someone who can change a commitment.
Architecture often has a clear place in the first conversation and a less clear place in the second. The direction gets approved, then arrives in planning as another request to fit around work already promised.
Before the next endorsement, ask: where will this become a funded line item, and who needs to be there?
Bring options with their consequences. “Adopt the service” is hard to allocate. “Give up 1 sprint of portal work so Operations can become the first consumer” gives the planning room a choice it can make.
Put the 4 names beside it
For a decision that crosses teams, write down:
Who benefits.
Who funds.
Who does the work.
Who carries the risk.
Different names are normal. Missing commitments are the problem.
The person getting the benefit may have no control over the team doing the migration. The person funding the build may not own the operational risk. A technically sound choice can wait a long time for those interests to meet.
The useful architecture work is to make each option’s consequence visible in the terms those people are responsible for: a launch date, a support workload, a budget line, an operational exposure.
Those conversations may happen in the same meeting. Check that all 3 actually happened.
________
🧭
________
The Stop-Change-Start Test
Take 1 agreed decision. Sit with each team it touches, 1 team at a time. Give the first conversation 20 minutes.
Ask for the team’s answers, rather than presenting the ones you hope to hear.
What stops?
Which work, system, or plan comes off the roadmap because of this choice?
“Stop building new integrations to the local customer table from January 18” is an answer. “Reduce duplication” still needs a conversation.
Sometimes the stop is smaller than a whole project: a sprint deferred, a feature reduced, a manual task retired. Make the capacity being released visible.
What changes?
Which existing commitment now looks different?
Look for a date, scope, interface, owner, or operating responsibility. “We’ll take the architecture into account” doesn’t tell the team what to do differently on Monday.
What starts?
What new work begins, who owns it, and when?
A migration needs a place in the plan. So does the work around it: support, reconciliation during the transition, and keeping the runbook usable.
Then read the answers together:
✅ 3 concrete answers with dates: the team has an actionable commitment. Confirm it with the person who owns the plan.
⚖️ A start, but nothing stops: agreement may have outrun capacity. Take the missing trade-off back to planning.
⚠️ 3 blanks: the decision hasn’t yet been translated into that team’s work. Check whether the team is affected and what it needs to decide.
This is a diagnostic, not a compliance score. Filling 3 cells doesn’t prove that a migration will work.
And if there is genuinely spare capacity, record it. You don’t need to cancel useful work to satisfy a worksheet. You do need to know where the new work fits.
The awkward middle case is often the most useful one. The team can name what starts because everyone has discussed the future. Nobody can name what stops because the future has simply been added to the present.
The blank column is the finding.
Keep the conversation with 1 team. In a large room, blanks can turn into speeches about why the decision is still right.
And resist filling the blanks for them. Your contribution is to make the missing commitment visible to the person who can make it.
Put the work that stops beside the work that starts. A full START column can hide an unfunded plan.
________
🧭
________
One decision, 3 ways to pay for it
Here’s a fictional example. The people, dates, and capacity estimates illustrate the trade-off; they aren’t a client case or a benchmark.
Sales, Operations, and Platform agreed in the spring to a shared customer service.
It’s now Q4. Sales has a launch built on its own customer table. Operations needs a reliable service with a named owner and has a change freeze to manage. Platform has no migration capacity in the current plan.
The shared-service direction still makes sense. The teams haven’t agreed on how to get there.
A. Build it now
Bring both consumers across together and give Platform responsibility for the service.
Sales and Operations would each give up a substantial part of the quarter. Sales delays its launch. Operations disrupts its freeze. Platform needs capacity for both migrations and the service itself.
The benefit is shared. The near-term cost lands on specific commitments those teams are already measured on.
This can be the right choice if the urgency justifies it. The launch delay and displaced work belong in the decision.
B. Federate
Keep the local customer tables and build reconciliation between them.
The immediate migration is smaller. Sales keeps its launch. The org accepts continuing duplication, and Operations takes responsibility for reconciliation and the mismatches that come with it.
That might be an acceptable operating model. It needs an owner and funded upkeep, rather than a temporary label that lasts for years.
C. Sequence adoption
Operations adopts first in Q4. Sales ships its launch on its own table, then migrates on January 18, 2027. Platform builds for 1 consumer first.
This preserves the launch and reduces the initial integration scope. It also leaves 2 customer stores during the transition. Operations owns interim reconciliation. Sales accepts a 1-sprint Q1 migration.
The cost hasn’t disappeared. It has moved in time and gained names.
For this example, the owners choose C. The accompanying brief records the commitments like this:
The work changes too:
Casey records the work in the October 19 plan, including the January migration commitment. The named owners accept it on October 16.
If Sales can’t reserve the January capacity, C remains a proposal. Calling it sequenced adoption doesn’t make the second step happen.
________
🧭
________
Write the other half of the decision
The choice is the part people tend to remember. The accepted downside is the part that can disappear between meetings.
For C, write it beside the chosen option:
Sales ships its Q4 launch on its own customer table, then spends 1 sprint migrating in Q1. Until January 18, 2 customer stores remain and Operations owns interim reconciliation. Platform builds for 1 consumer first; the second integration is deferred.
Then add the people who accepted it: Alex for the Sales migration, Morgan for the interim operational work.
Finally, record the reopen condition:
Revisit if the Sales migration estimate exceeds 1 sprint, or a third team needs the service before Q1.
That gives the next conversation somewhere to start.
If someone points out that 2 stores still exist, you can check whether that is the downside everyone accepted. If a third consumer arrives in November, the condition has changed and the decision deserves another look.
A reopen condition doesn’t prohibit judgment or new evidence. It helps distinguish an expected cost from a changed situation.
The choice is half the record. The accepted downside needs an owner and a condition for reopening it.
When the planning room says no
The planning room may hear all of this and still choose B for a year. The launch matters more. Migration capacity is constrained. Reconciliation is a cost the org is prepared to carry.
That is a decision too, provided the cost, owner, and review point are explicit.
Record the consequence: the org is funding federation for this period, rather than claiming that shared-service adoption is about to happen.
The target architecture may remain an ambition. The current operating model should reflect the choice people will actually fund.
I think this is where architecture earns a lot of its credibility. You can prefer an option and still give the people choosing a fair account of the alternatives.
The disagreement is part of the work. Make it usable.
David was using stop, change, start to understand a new org, without assuming that patterns from the last role would fit. I borrowed the questions for a different moment: after an architecture decision, when we need to find out what it means for each team.
Thank you, David. The simplicity makes room for people to talk about the work they’re actually doing.
If you’re starting a new role, his original post is worth your time. The commitment test here is my application of the questions; David’s starting point deserves its own read.
And if you try the test with a team, I’d like to hear what you find. Which answer was hardest to make concrete?
________
🧭
________
Your One-Page Decision Brief
I’ve put the whole record into a free PDF:
A fillable 1-page brief with the chosen option, alternatives, accepted downside, owner, and reopen condition.
The 4 names, a stop/change/start row per team, and the planning record where the work lands.
The filled shared-service example above, plus a clean 1-page print copy.
Choose 1 decision your org has made more than once. Give yourself 20 minutes with its current record and 1 affected team lead.
Write the accepted downside before polishing the decision statement.
Ask who agreed to carry it. Confirm the answer with that person.
Run stop, change, start for this team. Leave an honest blank where the commitment is missing.
Take 1 specific capacity choice to the person who owns the plan.
You don’t need to fill every box to make the conversation useful. Finding the missing owner or the work nobody has capacity for is progress.
The next time that decision comes back, you’ll have more than the slide with the winning option.
Hit reply and tell me: which decision in your org has been made the most times, and what keeps bringing it back?
Until next Sunday,
Kevin
________
________
Kevin Donovan I write Architects' Hub every Sunday for senior architects and technology leaders. You're getting this because you subscribed at architects-hub.kit.com. Know someone who’d find this useful? Forward it to them.
Kevin Donovan · Turning architecture into decisions
A Sunday newsletter for senior and enterprise architects. One decision, one practical model, one tool you can use Monday. Written by Kevin Donovan. Join 2,500+ architects. One email every Sunday.
Nov. 22 2026 | Read time: 11 mins. Kevin Donovan · Architects' HubForwarded this email? Subscribe to Architects’ Hub for the next issue. The room went quiet Hey , Two Sundays ago, our hypothetical pilot case got its commitment record. Last Sunday it got a walkthrough, and the operating owner signed with 3 tests on record. The AI assistant now drafts delayed-order replies for 3 teams, and the drafts are good. This week someone proposed letting it do more than just draft… In this edition The...
Nov 15, 2026 | Read time: 11 mins. Kevin Donovan · Architects' HubForwarded this email? Subscribe to Architects’ Hub for the next issue. Tuesday, 10:40 am. Hey ,Last Sunday we left a hypothetical customer-service assistant with a commitment record and unanswered questions about capacity, data and support. This week, move the story forward 3 weeks. The team has agreed to a bounded trial on 1 queue. Agents still review every draft. Wider rollout is waiting. I want to spend this edition with the...
November 8, 2026 | Read time: 9 mins. Kevin Donovan · Architects' HubForwarded this email? Subscribe to Architects’ Hub for the next issue. Hey ,In this edition of Architects' Hub: What the service lead heard when the sponsor said, “We should roll this out.” A commitment record filled in for one example customer-service queue, PLUS a blank copy for yours. How to take a conditional decision back to the sponsor who wants a date. Community Corner: Santiago Jesus Lasso Valenciano on the work that...