The Platform

What Priorities.ai is

Priorities.ai is Strategic Commitment Management — the missing layer between strategy and execution. It is not a ranking tool. Ranking is the last of five steps, and every other tool on the market addresses only that last step.

The platform governs the entire process — from the moment a candidate enters the system to the moment a commitment is recorded in your systems of execution with its full rationale intact.

The mechanisms below run as one recurring loop — see Unify, Understand, Act for the detailed walkthrough.

---

The Five Steps

Most organizations believe prioritization is the act of ranking items. They are right — but they have never been shown that ranking is only the final step of a much larger, mostly invisible process.

Every defensible prioritization outcome requires five distinct decisions to be made explicitly, in sequence:

1. 01ParticipantsWho has a voice, who decides, who observes
2. 02ItemsWhat is eligible to be compared at all
3. 03CriteriaWhich dimensions of value apply, and their weight
4. 04ValuesHow each item performs, on a stated scale

Settled explicitly — or settled by default, undocumented

Where most tools begin

05Ranking The order, and the reasoning that produced it

Most tools address only the ranking. Most processes handle the first four steps informally, incompletely, and without documentation. The result is a ranked list that cannot answer why and does not hold under the first encounter with execution pressure.

Priorities.ai enforces the full sequence.

---

The Three Mechanisms

The platform operates on three mechanisms. Every defensible priority requires all three.

01 — The Frame

The Frame is the foundation. It governs the first four steps — the decisions that must be made explicitly before any ranking is legitimate.

The Frame establishes:

• Participation — who is in the decision, who holds authority, who contributes judgment
• Comparability — which items belong in the same comparison (Track Cell = Item Category × Altitude)
• Criteria — what dimensions of value apply and how they are weighted
• Item readiness — whether every item in the pool is evaluable against the defined criteria

Without a locked Frame, a ranking is not a decision. It is a list.

---

02 — The Cascade

The Cascade is the governed process through which every level of the organization contributes what it carries.

Executive leadership carries vision, strategic context, and pattern recognition from years at altitude. Working teams carry operational truth — what is actually hard, what is actually possible, what the customer is actually saying. Neither has complete information. Both must contribute.

The Cascade is implemented through Catchball — a bidirectional priority alignment mechanism in which proposals flow up from teams and requests flow down from leadership, iterated within a Cycle’s authority lanes until alignment is reached.

Executive leadership

Carries vision, strategic context, pattern recognition from years at altitude.

↓Direction, constraints, requests

↑Proposals, evidence, what it will actually take

Repeats within the cycle’s authority lanes until it is settled

Working teams

Carry operational truth: what is hard, what is possible, what customers are actually saying.

The loop stays open until leadership answers

• Accept
• Accept with conditions
• Return
• Reject

The result is a decision that belongs to everyone who needs to execute it — informed by altitude, grounded in proximity, and not made unilaterally by either.

---

03 — The Plan of Record

The Plan of Record is the durable record that holds after the decision is made.

The Plan of Record functions as the organizational instrument panel. It holds:

• What was decided (the Published Priority List)
• How it was decided (the full decision provenance — criteria, values, session results, governance events)
• Why this, and not that (rationale, overrides, trade-offs, on the record)
• What changed and why (signal-informed adjustments, with the signal named)

When a new priority arrives — a competitive move, a market shift, a failed assumption made visible — the Plan of Record gives the organization a baseline to evaluate it against. It can ask: is this a legitimate response to new information, or a departure from a commitment that still holds?

That question, asked and answered explicitly, is what separates organizations that adapt well from organizations that drift.

---

Two Developer Primitives

The three mechanisms are built on two additional primitives that developers integrate directly:

Item Relationship Graph

Other tools treat items as flat — a list of tickets or a backlog row. Priorities.ai maintains a typed, mutation-governed, auditable relationship graph on every item in the system.

The graph supports seven relationship types: Cluster, Related, Bloc, Dependency, Aggregation, Elaboration, and Lineage. Each carries explicit semantics that affect prioritization logic — a hard Dependency blocks commitment-quality decisions until it is addressed; a Bloc enforces all-or-nothing selection atomicity; Aggregation propagates hard dependencies from children when a parent is committed.

Platform migration Billing API Invoice service Pricing rollout blocks until addressed rolls up all or nothing

• DependencyblocksA hard dependency blocks a commitment-quality decision until it is addressed.
• BlocbindsAll-or-nothing: the items are selected together or not at all.
• Aggregationrolls upCommitting the parent propagates the children’s hard dependencies.
• ClustergroupsItems that should be considered together, without forcing selection.
• RelatednotesA connection worth seeing during comparison; no enforcement.
• ElaborationdetailsOne item expands another into more specific work.
• LineagedescendsThis item came from that one — history preserved across cycles.

No other prioritization tool exposes this graph or enforces its semantics.

Decision Traceability

Every decision the platform makes or records is available as a typed, queryable API resource.

Decision classes: selection, range_selection, convergence, force_ranking, review. Every auto-recorded decision carries its source event, its linked objects, and its rationale. The decision stream is consumable via webhook (for real-time integrations) and via the API (for retrospective analysis and audit).

One decision, as the platform stores it

Decision class

selection

Source event

session_result.published

Linked objects

cycle · track · 14 items · 6 participants

Criteria applied

4, weighted, locked before scoring

Rationale

Capacity fell 20%. Narrowed to the top two tiers.

Recorded by

VP Product · authority lane

Available immediately as

• APIQueryable by class, object, participant, or date
• WebhookPushed to your systems the moment it is recorded
• ExportPiped to the warehouse for audit and analysis

Answers “why did item 7 outrank item 3?” — three years later, not just this quarter.

---

Two Lanes, Kept Independent

Judgment and authority run on separate lanes. Item management, track readiness, session execution, and session results form the judgment lane. Cycle design, Catchball, proposal review, and the Published Priority List form the authority lane — which feeds both systems of execution (Jira, ADO, Asana) and the Plan of Record.

Judgment lane

1. Item management
2. Track readiness
3. Session execution
4. Session results

Where the thinking happens: what is ready, what teams actually believe.

Neither lane advances the other. Finishing a session does not move a cycle; accepting a proposal does not either.

Authority lane

1. Cycle design
2. Catchball
3. Proposal review
4. Published priority list

Where commitment happens — and the only lane that can publish one.

→ Systems of execution · Jira, ADO, Asana → Plan of Record

The two lanes are independent. Session completion does not advance cycle phases. Catchball acceptance does not advance cycle phases. Each lane governs its own state machine.