The AI in Priorities.ai
The AI does the work around the decision. You make the decision.
More than thirty agents and assists work the record around the clock. They shape what arrives, red-team the ranking before you publish, watch the plan after you do, and bring the voices that were never in the room. You make the decisions.
Most tools give you an assistant that answers questions about a list. Priorities.ai gives you a workforce. More than thirty agents and assists work inside the record of what your organization decided: they shape what arrives so it can be compared, they stress-test a ranking before you publish it, they watch execution and the outside world for the moment your plan is touched, they remind you when a decision is due to be reconsidered, and they bring the customer and the pipeline into the room as voices. Every one of them proposes. Every decision is yours, and it is on the record.
Before the decision: the record shapes itself
Requests arrive from everywhere, in every form. The agents make them comparable.
ATC reads each new item and files it: what kind of thing it is, and at what altitude it belongs, using your own categories as the reference. Twins sweeps the backlog and finds the same request hiding under three names, so nothing is scored three times. Scribe turns typed notes or a dictated conversation into structured items with their relationships already sketched. Nexus finds the dependencies and breakdowns nobody declared. Talk through the meeting; leave with a pool that is ready to rank.
Before you publish: the ranking gets red-teamed
A ranking that looks settled is where the expensive mistakes hide. Three agents read the numbers and tell you where.
Fracture stress-tests the order: the knife-edge pairs a small shift in weights would flip, the top priority propped up by a single criterion, the popular item that a minority scored near the floor. Compass holds the list up against what leadership said the strategy was and shows the gaps: critical objectives with nothing driving them, effort piling onto a low-priority goal while a critical one starves. Keystone reads the dependency graph and finds the low-profile item that, sequenced first, unblocks a stack of priorities behind it, and the circular dependencies that can never start.
You see all three before the list is published, not after the quarter is spent.
After you publish: the plan gets watched
This is where most tools stop and Priorities.ai keeps going. Change is coming; the question is whether you see it in time and whether the reason gets recorded.
Meridian compares what you ratified with where the team’s energy is actually going, read from the tools they already work in, and speaks only when the two diverge: a top commitment nobody is touching, an unranked item quietly absorbing the team. Condition Watch notices when the rules for changing the plan are bypassed: a material signal on the record that nobody acted on, or a plan change with no signal behind it. Beacon builds its own watchlist from your ratified priorities and assumptions, competitors, technologies, regulations, named customers, and sweeps the outside world daily, surfacing a signal only when it touches something you committed to, with the evidence attached. Review Due keeps the promises the record made to itself: every decision recorded with a review date, every commitment armed with a trigger, comes back to the docket on time.
Together they are how a plan changes once, deliberately, with the reason on the record, instead of drifting.
In the room: the voices that were never in the room
Voices are AI participants with standing in a session, each grounded in a body of evidence your organization already has. A Customer Voice built on the support-ticket corpus speaks for every frustrated customer. A Revenue Voice built on the pipeline speaks for the deals you are losing. They contribute context and argument to the decision. They never cast a vote. You can create your own.
At your side, in your voice
A guide is one click away on every page, in the Navigator’s voice, in Harry Max’s, or in Wes Bright’s, to find the right screen or answer a plain-language question. A Co-Facilitator coaches the method as you use the tool: how to frame this decision, which method fits this session, what a disagreement about weights actually means.
Ask the record for what you want in plain English, “active risks nobody has touched in a month,” and get a precise, shareable query. Get a draft of the Catchball request or proposal from the context already in the record, then edit and throw. Get criteria that will actually separate the candidates before anyone scores. After a session, get the result explained in plain language: why the items ranked as they did, where alignment is weak, what the disagreement was really about.
Your agents, our record
Priorities.ai is built to be read by the agents you already run. Claude, Cursor, a personal agent gateway, or an agent you build yourself can read the live ranked list and the reasoning behind it, and file attributed input: propose an item, add evidence, make a Catchball argument. Decision events are pushed the moment they land, so your agents react to a published list instead of asking for it.
Voice, never authority
Every agent and assist proposes; a named person accepts, edits, or declines. Publishing a list, accepting a proposal, changing the plan: those acts are made by people and written to the record, every time. That is not a limit on the AI. It is what lets you give it this much to do.