How should an artifact-first distributed team make decisions without turning everything into meetings or consensus?
Reading time: 7 minutes
Summary:
The decision starts as an artifact.
First define what is being decided, why, who decides, the deadline and the constraints.
Thinking enters the artifact before people influence one another.
People independently write their positions rather than starting with a meeting.
The artifact accumulates the team’s intelligence.
Individual input gets structured and synthesized rather than disappearing into Slack, messages or conversation.
The team reviews the artifact, not a presentation of it.
Everyone gets another look and can modify or add arguments there.
A meeting is conditional.
It happens only when synchronous discussion can add something — particularly discovery.
Even if a meeting happens, the process returns to the artifact.
The final decision, rationale, owner, date and reopening condition are recorded. The meeting doesn’t become the authoritative source.
An Artifact-First Decision Process
There are many ways to make decisions. Most can work under the right conditions.
I found the following process particularly useful when:
- I want to spend as little personal time on the decision as possible.
- I want to include enough relevant information to make a good decision while keeping the effort reasonable.
- I want people to contribute when they have something useful to add.
- I don’t want irrelevant opinions to consume everyone’s time.
- I want to use the simplest possible process for the decision.
There is another assumption behind it.
I don’t care who proposed the right answer or the wrong one. I don’t want decision-making to become an ego fight. I want useful opinions to surface, regardless of where they came from.
For this to work, the decision-making process has to be clear and transparent. Everyone should know how a decision will be made and what their role is.
For example, in a small startup:
CEO — makes the final call.
Relevant advisor — provides expert input that carries additional weight.
Other advisors — contribute like other team members unless they have relevant expertise.
Team members — suggest options, information and reasoning.
Or:
Investor — makes the final call.
CEO — provides influential input.
Relevant advisor — provides expert input relevant to the decision.
Other advisors and team members — contribute suggestions and reasoning.
The exact structure can change, but it must be clear in advance how the final decision will be made and who owns it. The decision can be made by majority vote, by the CEO making the call based on the discussion, or in another way. Any mechanism is fine if it is explicitly agreed and shared with the team in advance.
Use only as much process as the decision needs
Not every decision deserves the full process.
If a decision is cheap, local and easily reversible, let the person closest to it decide. If the consequences are meaningful but reversible, written input and a short synthesis may be enough. If the decision is expensive, difficult to reverse or affects many people, use the full process.
The purpose is not to process every decision thoroughly. The purpose is to use the lightest process that can still produce a good decision.
Start with the decision
Before asking for opinions, write down:
- What exactly are we deciding?
- Why does it need to be decided?
- Who makes the final call?
- When does it need to be decided?
- What important constraints already exist?
This keeps people answering the same question.
First: think independently
For a remote distributed team working artifact-first and using AI at the company level, opinions should normally begin in writing.
There is an important benefit to collecting them independently. People first say what they actually think. They are not reacting to the CEO, the loudest person in the room, the most senior expert, or whatever idea happened to be mentioned first.
For fast decisions, the writing can be relatively free-form. AI can structure it afterwards. When time allows, more structure produces better thinking.
Writing forces the person making a suggestion to inspect their own idea. Sometimes an opinion sounds convincing until you have to explain precisely what it means, what problem it solves and why it should work.
Some opinions don’t survive this stage. That’s useful.
Ask for structured input
For decisions worth spending more time on, I like asking people to describe their thinking at two levels.
First, distill it. Then explain it.
For example:
Idea — around 10 words or less.
Idea explained — up to around 50 words.
Problem it solves — around 10 words or less.
Problem explained — up to around 50 words.
Then, where useful, add supporting context, like downsides or risks, evidence or useful information, and other relevant assumptions.
The word counts are not strict limits, they are constraints that force thinking.
If you cannot describe your idea in roughly ten words, there is a good chance you don’t understand it clearly enough yet. Without some constraint, people often start writing before they have finished thinking. You get long explanations containing ideas that contradict themselves. Short writing makes that harder.
The same applies to the problem. Naming the exact problem in ten words or less forces you to decide what you are actually trying to solve.
Ten words are often not enough for somebody else to properly understand the thought. That’s why a slightly longer explanation is also needed.
To put it shortly, the pattern is: distill first, explain second.
Additional fields like risks, evidence and assumptions can provide useful context, but they are not always necessary. Add them when they improve the decision.
Then synthesize
Once the independent opinions are collected, ask AI to organize them to make the team’s thinking easier to inspect.
The result should preferably be a very small table that people can understand quickly.
At this stage, attribution is usually unnecessary. The argument matters more than the person who made it. Removing names also reduces defensiveness and ego fights over whose idea was right, wrong or eventually chosen.
Give everyone one more look
Now share the synthesis so everyone can see what the others think, what options exist, where people agree and where they don’t.
Give people a defined amount of time to think and respond.
If someone discovers another option or changes their opinion, they can add it to the shared artifact. But now the new argument has to fit the structure of the document. That forces the person to explain it even more precisely.
Again: better writing usually means better thinking.
Then decide whether a meeting is useful
Many decisions can finish here. The decision-maker has the information, the options and the disagreements. They can make the call and record it.
For some decisions, however, a short meeting is still valuable.
The purpose of that meeting is discovery.
People have already thought independently. They have already seen everyone else’s arguments. Now they can challenge assumptions, combine ideas and discover an option that nobody found alone.
The table may already contain a clear leading decision. That’s fine. The meeting exists because a better one might still emerge from the conversation.
Keep the discussion bounded
The facilitator should start with the decision itself, the leading options, the important disagreements and the unresolved questions. Then discuss those.
To keep the agreed time, the facilitator has to cut off-topic discussions quickly and strictly manage the length of everyone’s contributions.
If another problem is discovered, capture it as a separate problem. Then check whether the original decision can still be made. If yes, proceed as planned. If not, decide what needs to happen next and re-plan the activity accordingly.
Make the call
At the end, make the decision using the mechanism agreed in advance.
In most cases, this means the identified decision-maker makes the call. It could also mean majority voting, consensus, or another mechanism if that was explicitly chosen beforehand.
Then, to support the artifact-first model, record the decision in the predefined format, including:
- what was decided,
- why,
- who owns the next action,
- when it was decided,
- and, where useful, what new information would justify reopening the decision.
With repetition, the team gets better at the process. People learn what useful input looks like. Writing gets shorter. Noise decreases. Meetings become smaller, some of them disappear entirely.
The process itself should also change when you find a better way to run it.
The goal is not to follow the process perfectly but to make good decisions with as little unnecessary work as possible.