Agile4AIBlog

Structuring the first question

How you frame the opening question in an SCI session shapes everything that follows. The structure is more important than the length.

In a Structured Collaborative Intelligence session, you’re not prompting a single AI system. You’re opening a structured dialogue between multiple models, with one AI — chosen by you — that synthesizes the collective response in the voice and format that you prefer. The first question and your AI voice selection don’t just determine what the first responses will be — they set the frame that every subsequent contribution works within.

Getting that frame right is worth the time and effort.

What the first question does

The opening prompt establishes the problem space. It signals what kind of thinking is wanted: analytical or generative, convergent or exploratory, evaluative or creative. It sets the context for the AIs to align their independent perspectives and, based on their training data, establish their position relative to the question and other AIs’ contributions as the conversation progresses.

A poorly framed first question runs the risk of over-biasing some or all of the AIs, potentially pitting one against another instead of creating the shared context where all can collaborate in genuine agreement and disagreement. The goal is for the AIs to reach consensus, not agreement. Consensus here is the worked-through collective position that keeps the disagreements visible inside it; agreement is the flatter thing that sands them away. That is what lets your chosen AI synthesize the responses — agreements and disagreements included — in a way that delivers both breadth and depth.

A well-framed first question produces responses that are aware of each other — where an AI pushes back on an assumption, or builds on a prior response, or identifies a tension in the framing that the session needs to resolve.

Elements of a strong opening question

The most productive SCI opening questions contain three things: the context (what situation are we actually in), the constraint (what are we working within), and the specific question (what exactly do we need to figure out), which points in the direction of our desired outcome from the SCI session.

Context without constraint produces abstract answers. Constraint without context produces recommendations the models have no way to evaluate against your reality. Neither without the specific question produces anything actionable.

“We’re a twelve-person product team, we’ve been running Scrum for two years, and our retrospectives have stopped producing changes. What are the most likely root causes and how would you investigate them?” — that’s a question with all three elements. The context grounds the answers. The constraint (retrospectives specifically, not process in general) focuses them. The specific question (root causes and how to investigate) directs the kind of thinking you want.

There is nothing new about this approach. This is exactly what Agile culture has taught us to do to improve our team communications, align towards a common goal, and remain focused on the thinking that produces the best quality answer.

What to leave out

The opening question doesn’t need to be long to be well-structured. Unnecessary context actually dilutes the focus — the models will try to incorporate everything you’ve given them, and excessive setup can make the session slower to get to the useful territory. It can also derail the conversation by offering too many details that unnecessarily constrain the response to fit all those details; wisely choosing “just enough” is important to get the best answer.

Avoid front-loading every piece of background that might be relevant. Give the models what they need to answer the question you’re asking, not everything you know about the domain. If additional context turns out to matter, it can come in the follow-up turns or as interjections while the conversation is going and you spot divergence away from what is important to you.

When the first question is wrong

Sometimes you discover mid-session that the first question framed the problem incorrectly. Maybe the AIs’ responses revealed that you were solving the wrong problem, or that a key assumption you built the question around is contestable.

This isn’t a failure — it’s often the most valuable thing an SCI session produces. The right response is to reframe explicitly: “Based on what you’ve raised, I think the more important question is…” and start from there. The session has already done useful work by surfacing the reframe.

The first question is a starting point, not a contract. But the quality of your start determines how quickly you get to the territory that matters. Keep in mind that when multiple AIs engage in this type of exchange they produce deep and meaningful answers. You want your first question to be just enough to set them in the right heading; that’s all that’s needed for the AIs to carry the conversation forward.

Comments 0

Sort
Checking your sessionThe article is ready. Comment options will appear once your session check finishes.