Design how your
important decisions
get made.
Every person and every organization that makes consequential decisions, the ones that are costly to reverse, already has a structure for making them. Most of these structures were never designed. Decision architecture is the practice of designing that structure on purpose: setting where you want to arrive before you read the present, carrying a decision from your first thoughts to a recorded commitment and keeping the act of deciding with a person, not with AI.
Decide backward. Set where you want to arrive before you read the present.
The desired future state, a finish line to know you have arrived and the things you refuse to lose on the way.
The two lenses are compared until comparing them changes neither. What would happen if no one decided sits alongside.
The present, examined in light of what the finish line requires, with each finding dated and attributed.
Bias is not the problem.
Unmanaged bias is.
Everyone relies on mental shortcuts, and most of the time they serve us well. But shortcuts can lean judgment in consistent directions, and that leaning is bias. When nothing in the way a decision is made manages it, the result is a frame no one examined, assumptions no one tested and priorities no one chose.
Bias plays out differently from person to person, so people who decide together rarely share the same understanding. AI makes this harder: it builds fluently on whatever frame it is given.
Over time, the cost builds up as decision debt: decisions postponed, owned by no one, never recorded or left in place long after they stopped fitting. Fixing decisions one at a time leaves the structure that produced them unchanged.
The same principle and cycle, for yourself and for an organization.
Which model governs a case depends on how the case is opened, not on the size of the decision. Both are open to anyone, and both set out the part AI may play.
For a decision you open for yourself, on your own authority. Solo means accountable alone, not working alone. Its framework is specified in enough detail to be used, and its paper names the parts it does not yet cover.
For decisions made on behalf of an organization, under a mandate from a board, an owner or a strategy in force. A proposed model: its requirements are stated, and it has not yet been specified in operational detail or demonstrated.
You decide.
AI proposes, sorts, classifies, surfaces and assembles. The decider elects, confirms, adjudicates and rejects. The act of deciding always belongs to the decider. Your own account is recorded before any AI takes part, and every request to AI to develop a route must ask for the strongest argument against it.
The simplest habit: write down where you want to arrive before you ask AI for analysis.
How AI takes part →
An open discipline.
A dated record.
One app.
BC-DS’s work has three parts. Decision architecture and its two models are open to anyone. Convoking4 keeps the record of how the work evolved. Decisiontect puts elements of the individual model, SDA, into an app.
Decision architecture, its principle, its cycle and its two models are open: anyone may apply, evaluate or build on them, with or without software.
BC-DS’s earlier platform for decisions in digital transformation and project management. Convoking4 is no longer offered as a product: its development evolved into Decisiontect, and convoking4.com keeps its history and will carry BC-DS’s work on Business Decision Architecture (BDA).
A decision architecture app that implements elements of the Solo Decision Architecture (SDA) framework. The Decisiontect beta is coming. Decisiontect certification: four planned credentials, DT-S™, DT-A™, DT-C™ and DT-P™.
From your first thoughts
to a recorded commitment.
Every decision case follows the same five stages. The order is fixed; the timing is not. Understand runs while the case is open and Evolve once the commitment is made; Communicate, Align and Decide follow in order, and a case goes back to an earlier stage whenever a condition no longer holds.
Each person describes the situation in their own words before hearing anyone else, and before any AI. The destination is described first, with its finish line; then the present is read against it.
The destination is shared before the present is assessed. This is designed to bring disagreement out early, while it can still change the decision, and to make it about the route rather than about whether the decision is right.
The perspectives the case requires are brought in and recorded. Alignment means being ready to carry out the commitment, whether or not you agree with it.
The destination is fixed at the crossing, as the case leaves Align. At Decide, one accountable decider commits through a recorded act, after a check of conviction, costs and standing. The outcome is a commitment, a decline or a hold.
The commitment is watched after it is made and reopened only at a tripwire or a scheduled review. Findings from the work go back to the case, which must respond.
↺ What is learned opens the next case.
Most of the parts
are not new.
The combination is.
Backcasting, defining values before options, independent judgments before group discussion, structured dissent, the premortem, clear decision roles, tripwires and written decision records: most elements of decision architecture existed before, and many have been tested.
What the papers propose as new, in the literature they examine, is how the parts are combined: one structure for deciding, for individuals and organizations, that covers committing and acting as well as choosing, with explicit rules for the part AI plays.
Whether the combination leads to better decisions than the parts used separately has not been tested yet. Fifteen hypotheses about its effects are stated, with the evidence that would count against them. The authors have used the Dual Lens Principle informally, with friends and family; that is not a study.
The founders.
The authors are practitioners, not researchers in cognition, behavioral strategy or decision analysis. BC-DS uses decision architecture as the foundation of its commercial app, Decisiontect™, and makes the discipline available so that others can examine, apply and test it independently of that interest.
Start with where
you want to arrive.
Write down where you want to arrive before examining the present or asking AI for analysis.
Business Decision Architecture comes after SDA: BC-DS is completing the first prototype of SDA in Decisiontect, and work on BDA follows.
Understand · Communicate · Align · Decide · Evolve