How to use the ORCA Dynamic Agenda

The agenda answers one question every Monday: where should the next hour go? It is a read of Close, not a second CRM — nothing you do here writes back to Close, and nothing here replaces the skills. It ranks, and it hands off.

The three pages

A lead appears on exactly one page. The sweep keeps the most urgent opportunity per lead, and that opportunity's Close status decides the page — so a lead with both a live deal and a parked follow-up shows up under the deal.

Three scores, not one

Each page scores a different question, and the number in the score column means something different on each. They are not comparable across pages: a readiness of 60 and a health of 60 are answers to different questions. The ? beside each score column gives the full formula, the thresholds, and what that number cannot see.

Priority is a separate read again, and it exists because health and urgency point in opposite directions all the time. A humming deal with a booked next step is high health and low priority. A good-fit deal cooling from neglect can sit at 55 health and be the most urgent thing in the pipeline. Read all of them as a triage aid, not a verdict: none of the three scores reads email bodies, call content or transcripts. They are computed from Close metadata. The ⚡ hand-off is what puts an actual conversation in front of a model — it carries the scores, the week's read and the Close link in one paste, and says which parts rest on metadata alone.

The lanes

Every lane name is an action, not a state — that is why they read as instructions. A record lands in the first lane it matches, so the order below is the order the sweep tests them in.

Two lanes exist to stop a bias rather than to describe a state, and they matter more than they look. 🔍 Triage on Active deals and 🔍 Fit Check on Active leads both catch records whose score is low because the data is thin, not because the lead is weak. Badly logged records systematically score worse, so ranking strictly by score would quietly punish records for being badly logged. Those two lanes catch them instead of burying them, and the job there is a two-minute pursue-or-park decision, not CRM tidying.

🔍 No Record on Follow-up later is the same idea pointed at data debt: a lead parked with no dequalification record can never be triggered intelligently, because nothing says why it was parked or when to come back. It is by far the largest lane on that page, and that is the finding rather than a fault of the page.

Lanes are collapsed, and capped per lane

Every lane starts collapsed. The header carries the count, the progress bar and — when the cap is trimming — a N of M badge, so the page opens as a map of where the week's work sits and you expand the one you are actually working.

Active leads and Follow-up later show 100 rows per lane; Active deals is uncapped. The cap is a display limit, never a filter: every row is swept, scored and stored, the overflow count is always shown, and a trimmed lane carries a Show all control that lifts the cap for that lane. It is capped per lane deliberately — one cap across a whole section let one lane starve another, and did: Fit Check held 91 rows and displayed zero, because rows that cannot be scored sort last by definition.

The weekly rhythm

1. Monday 06:00 UTC — the full sweep pulls Close, scores every record and writes the week. Prior weeks are never overwritten, so the agenda is a record of what you decided as well as what to do.
2. Every 30 minutes — a top-up finishes anything the weekly run could not, and does nothing when there is nothing outstanding. Some work cannot fit in one run: reading ~1,100 dequalification records takes many passes against the platform's per-invocation limits, so it drains in the background instead of blocking a sweep.
3. Work the top lane down. It is ranked, and nothing is hidden to keep it short. If it is empty, that is a real answer.
4. Tick as you go. Checkmarks are shared across the team — the initials beside the box say who ticked it. Your priority and date overrides are private to you.
5. Use the week switcher (top right) to look back. Scores are the read as of that Monday, not a live recomputation.

Sweeps can also be run by hand from the footer, per section. Those controls are visible only to the operators, because a sweep pulls the whole pipeline and spends model credits — widening who may read must not widen who may trigger.

The row buttons

The hitlist

◎ Hitlist, top right of Active deals, is the forecast view: every deal with an expected close date in this month, next month or the month after, filterable by rep and by team, with a chart of weighted ARR per month. Active deals only, because the other two pages carry no close dates.

Weighted ARR is recurring value × the rep's own confidence in Close. Three things are deliberately kept out of that total rather than guessed at, each reported with a count so the gap is visible: a deal with no confidence recorded (unknown is not zero), a one-off value (not recurring revenue), and a deal whose value period Close has not recorded (neither one thing nor the other). Deals with no close date at all are listed at the bottom and are in no month — usually the biggest reason a real deal is missing from a forecast.

Filters and defaults

Each page opens filtered to your own leads. That is a default, not a restriction — the owner menu takes several people at once, so you can read two reps together, and clearing it shows the whole pipeline. Clearing it is not the same as ticking everyone: only the cleared menu also shows leads with no owner, and there are a handful owned by people who have left. The team chips are shortcuts for that menu: 10x ticks Tomas and Pasquale, Bratwurst ticks Charmaine and Helio, and each replaces whatever was ticked rather than adding to it — so the chip and the menu always agree on who is being counted. Owners on no team are reached by clearing the menu. Search matches the lead name, the owner and the next step. Priority and score-band chips stack with everything else.

What it will not do

• Write to Close. Nothing here changes a status, sends a message, or moves a deal.
• Send anything. Even the reads that produce a draft only put it on screen for you to copy.
• Recompute the past. Each week is the read as of that Monday; a rescued deal keeps its old score in its old week, which is the point.
• Read conversations. Bodies, transcripts and sentiment are not inputs to any of the three scores — that is what the skills are for.

What it needs from you in Close

The agenda can only rank what Close records. Four fields do most of the work, and an empty one does not degrade quietly — it lands the record in a lane that says so:

• Expected close date on an opportunity — without it a deal is in no month of the hitlist and no forecast.
• Confidence on an opportunity — without it the deal is left out of weighted ARR rather than counted as zero.
• Lead Segment — without it ICP fit cannot be read, so the lead sits in Fit Check with no warmth score instead of a real lane.
• A dequalification record when you park something — the type, the reason and the re-engage date are what let Follow-up later bring it back at the right moment instead of leaving it in No Record for ever.

When something looks wrong

The sweep reports its own failures rather than hiding them, at the top of every page, and those messages are worth reading rather than dismissing. If a signal could not be resolved — a Close filter that came back unfiltered, a missing API key, a limit reached partway — the result names it and the affected signal is dropped instead of guessed at. A lane that looks surprisingly empty or surprisingly full is worth checking against Close before trusting it.

One signal is switched off on purpose: Close's own needs attention flag is discarded on every sweep, because the request for it returns the same rows with and without the flag, which means it is not being applied. It feeds health, lane and rank, so trusting it unverified would mis-rank the whole pipeline while looking normal. Lanes rest on close date and staleness, both verified.

← Back to the agenda