Operating With Decisions

Decisions are recorded commitments about how to advance an outcome. Each one carries a choice statement, rationale, owner, owning outcome, lifecycle status, alignment posture, and source trail.

This is where Ask The W acts as the judgement layer for AI-native product teams: humans and AI agents share the same record of why a product decision was made, what outcome it serves, and whether it is still aligned to the North Star.

The daily loop (signals -> decisions -> outcomes)

Start each day in /next:

1. Confirm the current North Star and operating target.
2. Review workspace insights and alerts for what changed.
3. Open the outcomes with the highest urgency or risk.
4. Review new signals and linked decisions.
5. Apply exactly one correction, note, or next move before continuing.

A decision is not a task, status update, recurring routine, or target state. It should say what changed and why the team should remember it.

Product reference

Find it in Ask The W by opening Decisions from the left navigation. The top of the page combines North Star context, outcome filters, outcome cards, and the decision feed so you can review signal-backed choices in one pass.

Next moves and notes

Next moves are generated prompts based on the current outcomes, linked signals, decision coverage, source health, and review backlog. They are not stored as standalone tasks. When the underlying signals, decisions, outcomes, or dismissed state changes, Ask The W rebuilds the list.

The /next route opens the Next Moves view because it is the operating front door. Decisions remain a first-class primitive at /decisions.

Use Add note on a Needs outcome evidence next-move card when you want to attach source material or decision context to that outcome, such as a customer note, product metric, source, or decision rationale. Saving the note stores it as a linked Signal for the generated move.

Use Make a note at the bottom when you want to add new material to the workspace. The selector controls what gets created:

• Signal: captures source context or supporting material.
• Decision: creates a canonical human decision.
• Outcome: creates a new outcome.

The composer may infer Decision when the text starts with decision language such as "we decided", "chose", or "committed". If you only want context for the operating map, leave it as Signal.

How decisions are created

Signals arrive from connected sources such as GitHub, pasted context, uploads, URL watchers, hosted connectors, and the Ask The W Plugin. The ingestion pipeline analyzes each signal, filters out low-information or purely technical noise, and looks for explicit decision language with business impact.

When the pipeline finds a valid decision, it creates or updates a canonical decision and links the signal to it. Similar decision headlines can merge into the same canonical decision so the product judgement layer does not fill with duplicates.

You can also create decisions from W chat or promote source signals from the Signals page. Use manual creation when the source trail is clear but the automation did not have enough confidence to publish a decision on its own.

Decision lifecycle

Each decision has one lifecycle status:

• Proposed: a tentative or exploratory direction.
• Committed: the default for accepted plans and newly created decisions.
• Shipped: the work has launched, merged, deployed, or gone live.
• Abandoned: the direction was cancelled, dropped, sunset, or marked as not a decision.

Incoming decision text can set this status from its language. If the text does not clearly say otherwise, Ask The W treats the decision as committed.

Decision alignment

Each decision also has an alignment label relative to its linked outcome:

• Aligned: the decision moves the outcome forward.
• Orthogonal: the decision is related but does not directly advance or block the outcome.
• Conflicts: the decision works against the outcome or contradicts a prior committed decision.
• Ambiguous: there are not enough signals yet to classify alignment.

Outcome mapping sets the initial link and alignment. Decisions with no linked outcome, ambiguous alignment, conflicting alignment, or an orthogonal link are treated as needing attention in the operating view.

Reviewing signals

Signals are timestamped source records from traceable sources. Each signal carries source, time, raw content, and inferred topic. Ask The W stores them as structured signals first, then maps them to outcomes and decision history.

Signal lifecycle:

1. Ingest from source connectors and upload or paste flows.
2. Parse decision-like facts, rationale, references, and timestamps.
3. Classify confidence, source labels, and semantic kind.
4. Link signals to outcomes, makers, and known workflow threads.
5. Surface signals in the signal stream, decision feed, and outcome cards.

Use Signals to inspect source type, source label, reference links, file context, timestamp, and review actions. Mark high-impact signals for review, attach them to the right outcome, and add missing owners.

Keep signals specific, attributable, and reviewed. Prefer specific events over summaries, keep authors or teams visible, and review frequently so ambiguous signals get clearer follow-up sources.

The decision feed and filters

Use decision-feed filters to control operational noise. Narrow filters for active teams, urgent reviews, and loud sources. Broaden them for onboarding, context gaps, and signal-quality review.

Common filters include outcome, functional area, growth stage, maker, source, status, and freshness.

Use tighter filters when multiple teams are active at once, when you need the highest-impact decision this sprint, or when noisy sources create alert fatigue. Use broader filters when onboarding a workspace, investigating a context gap, or training a new team on signal standards.

Alerts and what to do with them

Alerts collect operational interruptions the product uses for review prompts:

• unresolved conflicts
• approval or governance prompts
• onboarding or source-quality issues
• priority nudges from trend shifts

Open each alert, review linked context, take the suggested action, then acknowledge it. Acknowledged alerts are muted but retained.

North Star: set it, edit it

The North Star is the workspace-level aim for the company, shared workspace, or personal workspace right now. Outcomes are derived from it.

Open North Star in Decisions and click Edit. You can update the metric, current value, target, target date, owner, and rationale. Updates reconcile against active signals.

North Star changes directly affect freshness scoring, outcome triage suggestions, and the way W frames risk. Refresh failures can happen when signals are still loading or source sync is slow; if that happens, retry after confirming connector health.

Workspace insights

Insights are short workspace observations generated from recent signals and decisions.

Use insights as prompts, not verdicts. If an insight points to clustered conflict, open the linked outcome or decision and inspect the source trail before changing direction.

Analytics hub (metrics, pins, composer, leaderboards)

Open /analytics for operating health beyond one outcome.

Filter by team, stage, maker, outcome, source, and status. Pin important metrics. Leaderboards rank makers and outcomes by activity, signal volume, and decision leverage.

Watch for stalling outcomes, high unreviewed agent volume, repeated source conflicts, and North Star changes by maker or category.

Maker profiles

Maker profiles show who is making decisions and how those decisions affect outcomes.

Each maker profile can show decisions seen, outcomes touched, files referenced, trust score and trend, pairing and approval signals, and ownership details.

Use profiles to identify source concentration, ownership gaps, over-reliance on one source, and agent-origin decisions that need approval.

Maker types:

• Human: direct human-origin decisions.
• Agent: model or automation-origin decisions.
• Collaborative: human and agent decisions combined.

Corrections

The decision feed lets you change a decision's linked outcome from the decision card. If something is not really a decision, expand the card and choose Not a decision? to demote it back out of the feed.

Corrections also support decision text edits, status changes, alignment edits, rationale edits, outcome remaps, and demotions. Corrections are stored as user edits and applied back to the canonical decision record.

Notifications

Notifications cover approvals, conflicts, health changes, and refresh nudges. They are workspace-scoped.

Use Project, team, and settings to set notification categories to High, Normal, or Off.