| Data trust |
|---|
| DATA-0Pipeline sentinelDaily 08:00 | Trust in every dashboard starts with someone checking the pipes.In practice: The nightly NetSuite → warehouse sync — did every table load, with no errors? | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Last run succeeded: 16 Jul 2026, 08:05 Dubai.
- Next run: 17 Jul 2026, 08:05 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Continue supervised monitoring and investigate any later non-ok run.
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| Sales & CRM |
|---|
| CRM-1Deal hygiene watchdogDaily 08:20 (weekly extras Sun) | The CRM's early-warning system for neglect.In practice: Open deals with no scheduled next step — 88% of them, at first count | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Last run succeeded: 16 Jul 2026, 08:20 Dubai.
- Next run: 17 Jul 2026, 08:20 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Continue supervised monitoring and investigate any later non-ok run.
- Rajiv and the sales leads review the next supervised weekly digest against their view of the pipeline
- Agree the recipient list and final alert thresholds before unlocking delivery
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| CRM-2Pipeline pulseWeekly Monday | Week-over-week movement, one screen per business unit.In practice: Each BU's open pipeline: deal counts and value per stage | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Awaiting its first scheduled run.
- Next run: 20 Jul 2026, 08:15 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Verify the first scheduled run at 20 Jul 2026, 08:15 Dubai.
- ≥2 weeks of CRM-1 cleanup so the pipeline numbers are honest
- ≥2 weekly snapshots of its own for the first week-over-week comparison
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| CRM-3Lead intelligence — hospitality firstWeekly | A tireless prospector that reads everything, so Ifham doesn't have to.In practice: Public GCC hospitality-development news, tender portals, project announcements | Built & tested — pilot- One-time lead-research deliverable; it is not a recurring loop.
- Built and first-run complete: a scoring rubric plus a first shortlist — 12 qualified signals, 5 A-band, with urgency flags where competitor exits create a timing window.
| - 2–3 rubric iterations with Ifham's feedback on real shortlists
- Paid data sources only on conversion evidence — with fresh licences, never shared logins
- Schedule only after every prerequisite and a supervised run are approved.
|
| Inventory & purchasing |
|---|
| INV-1Dead-stock watchdogWeekly Monday 08:30 | Nobody could see the sleeping stock. Now someone checks weekly.In practice: Every item with stock on hand, against its last movement date | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Awaiting its first scheduled run.
- Next run: 20 Jul 2026, 08:10 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Verify the first scheduled run at 20 Jul 2026, 08:10 Dubai.
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| INV-2Stock-settings builderGoal-runs, batch of 50 | From 9% lead-time coverage to full — one approved batch at a time.In practice: Items ranked by usage value, 50 per batch | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Awaiting its first scheduled run.
- Next run: 21 Jul 2026, 08:25 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Verify the first scheduled run at 21 Jul 2026, 08:25 Dubai.
- Lerris's sign-off pattern on the first batches (the CSV sanity check is an open question with him)
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| INV-3Reorder-risk watchdogDaily, speaks when needed | The warning we never get: out-of-stock, seen weeks early.In practice: Available stock + open purchase orders vs lead time × demand rate per item | Built & tested — pilot- Built and tested, but not scheduled.
- Built and live-tested — provisional.
| - INV-2 approved lead-time coverage >60% to drop the PROVISIONAL label
- Nish's answer on the risk thresholds (open question from the Wave-4 results email)
- Schedule only after every prerequisite and a supervised run are approved.
|
| INV-4Supplier & commodity intelligence1st & 15th monthly | Wool, foam, Tencel — someone should be reading the market for us.In practice: A standing supplier watchlist — wool, cellulosic fibre, foam and cashmere suppliers we depend on | Built & tested — pilot- Built and tested, but not scheduled.
- Supervised day-one attempt failed the research gate on 16 Jul 2026.
| - Run the prepared manual-only retry no earlier than 17 Jul 2026, with source coverage and claim verification
- Review the retry manifest and QA before any Proof handoff or scheduling decision
- Schedule only after every prerequisite and a supervised run are approved.
|
| Finance |
|---|
| FIN-1Receivables watchdogDaily 08:35 | Money owed to us should never age quietly.In practice: Every open customer invoice, bucketed 0–30 / 31–60 / 61–90 / 90+ days | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Last run succeeded: 16 Jul 2026, 08:35 Dubai.
- Next run: 17 Jul 2026, 08:35 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Continue supervised monitoring and investigate any later non-ok run.
- Shahbaz's thresholds + recipient list (open questions from the Wave-3 results email)
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| FIN-2Bank compliance calendarMonthly, 1st business day | Compliance by calendar, evidence by ledger.In practice: Date-driven obligations: audited-financials deadlines, half-yearly projections, facility review dates | Built & tested — pilot- Built and tested, but not scheduled.
- Tested, not deployed.
| - Install and verify the real covenant configuration from Agent Vault or Infisical
- Finance confirms the fiscal-year end, monthly attestation workflow, recipients and escalation path
- Schedule only after every prerequisite and a supervised run are approved.
|
| FIN-3Ad-spend reconciliationMonthly | Nobody reconciled ad spend. That's how 14 months slipped by.In practice: Platform spend — pulled read-only from the ad accounts | Built — awaiting credentials- Tested with fixtures or read-only data, but not scheduled.
- Built against fixtures and threshold-proven, but not deployed.
| - Complete Google OAuth, production client-account linking, and target-customer selection
- Install the isolated read-only runtime and review budgets and policy
- Add Meta access and wire its token into credential-expiry monitoring
- Schedule only after every prerequisite and a supervised run are approved.
|
| FIN-4Expense-queue nudgeWeekly, speaks when stuck | A small loop with one job: keep the claims queue moving.In practice: Claims stuck pending review > 7 days | Built & tested — pilot- Built and tested, but not scheduled.
- Built and live-tested against the pilot app's database (its claims live in a single JSON array — the loop reads it faithfully).
| - The expense app's production sign-off
- App fix: store submission timestamps as real dates (feature request pending)
- Schedule only after every prerequisite and a supervised run are approved.
|
| Operations & construction |
|---|
| OPS-1Retention & DLP money trackerWeekly | Retention release dates, out of heads and into a watchdog.In practice: Every tracked project's retention % and milestone dates | Built & tested — pilot- Built and tested, but not scheduled.
- Built in registry/campaign mode and tested.
| - Handover dates for the registered projects (with Yasar's team)
- Billing-app feature request: a handover-date field, so registry mode can retire
- Schedule only after every prerequisite and a supervised run are approved.
|
| OPS-2Site-milestone speed monitorWeekly (when data exists) | You can't speed up what nobody times.In practice: Stage durations per project: delivered → inspection, installed → inspection sign-off | Built & tested — pilot- Built and tested, but not scheduled.
- Built in degraded mode and tested; the speed math is unit-verified and dormant until the billing app captures the three requested fields.
| - Billing-app feature request: three milestone-date capture fields (filed; awaiting Nish/Yasar's go)
- Schedule only after every prerequisite and a supervised run are approved.
|
| OPS-3Protector replenishment radarMonthly (1st) | Recurring revenue shouldn't depend on remembering.In practice: Hotel customers' protector purchases, by product family | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Awaiting its first scheduled run.
- Next run: 01 Aug 2026, 09:00 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Verify the first scheduled run at 01 Aug 2026, 09:00 Dubai.
- Nish/Ifham's answer on the minimum-history filter (open question e from the Wave-2 email)
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| IT & governance |
|---|
| IT-1Company-knowledge hygieneSunday 08:00 | Our first loop in production — the librarian for the company brain.In practice: The brain's own 4-hourly self-maintenance — did it run clean? | Live — running on schedule- Scheduled on Hermes vpsops; enabled.
- Last run succeeded: 12 Jul 2026, 08:01 Dubai.
- Next run: 19 Jul 2026, 08:00 Dubai.
| - Continue supervised monitoring and investigate any later non-ok run.
|
| IT-2Workspace security postureMonthly | Security posture drifts quietly. This loop notices.In practice: Two-step verification coverage across all accounts | Built & tested — pilot- Built and tested, but not scheduled.
- Built and live-tested.
| - Service-account delegation grant for the forwarding-rules check (open Wave-4 question)
- Schedule only after every prerequisite and a supervised run are approved.
|
| IT-3The loop-of-loopsWeekly Friday 09:00 | The portfolio's own performance review, every week.In practice: Every registered loop's last runs and outcomes — across both agent runtimes and plain cron | Live — running on schedule- Scheduled on Hermes bizops; enabled.
- Awaiting its first scheduled run.
- Next run: 17 Jul 2026, 09:00 Dubai.
- Email delivery is locked; success means the loop ran cleanly, not that mail was sent.
| - Verify the first scheduled run at 17 Jul 2026, 09:00 Dubai.
- Review the first scheduled Friday scorecard and act on any failures it finds
- Continue the supervised weekly soak; Nish decides when delivery may be unlocked
- Review no-delivery outputs; Nish decides when email may be unlocked.
|
| Marketing — Huxberry.com |
|---|
| HUX-L0Tracking foundationOne-time project | Before anything watches the ads, the ads must be measurable.In practice: Nothing — it's plumbing. It installs : the Meta pixel + server-side events, GA4 analytics, Google Ads conversion import, and Search Console verification | One-time project- One-time project, not a recurring scheduled loop.
- Waiting on one decision: who installs the tracking, given the ongoing agency dispute.
| - Nish's decision on who installs (agency dispute context)
|
| HUX-L1Ad creative factoryOn brief (goal-loop) | Creative volume without brand drift.In practice: Nothing on a schedule — it's a goal-loop : it runs when given a brief and stops when the brief is satisfied | Ready to build- Ready to build; no production runtime exists yet.
- Design complete, not yet built — deliberately queued behind the tracking-foundation owner decision, and one review cycle with Sherif is planned in the first month (this loop leaned hardest on the retired designer model's taste; Opus + the brand-rulebook verifier carry it now).
| - Build the brand-verifier scoring skill
- Sherif confirmed as the approval gate (pending his onboarding conversation)
|
| HUX-L2Search-ranking & competitor watchDaily | The search battlefield, checked before breakfast.In practice: Our rankings for the money keywords (we're #2 for “organic mattress Dubai” ) | Ready to build- Ready to build; no production runtime exists yet.
- Design ready, not yet built — the blocker is credentials (SEO tool API keys into the vault) and the Search Console side of the tracking foundation.
| - SEO tool API keys into the vault
- Search Console verification (part of HUX-L0)
|
| HUX-L3Ads account watchdogDaily + weekly ritual | No spend without results — ever again.In practice: Daily: pacing over/under, disapproved ads, and the coral rule: any ad set over AED 500 spend with zero sales in 7 days | Built — awaiting credentials- Tested with fixtures or read-only data, but not scheduled.
- Built and fixture-proven, but not deployed.
| - Complete HUX-L0 so results mean verified sales
- Complete Google OAuth, production client-account linking, and target-customer selection
- Install the isolated read-only runtime, approve the policy, and add Meta access with expiry monitoring
- Schedule only after every prerequisite and a supervised run are approved.
|
| HUX-L4Content pillar loopWeekly | The content gap no GCC competitor has filled — filled weekly.In practice: The content backlog — 9 planned pieces, Gulf-climate sleep guide first (the biggest search gap in our market) | Built & tested — pilot- Built and tested, but not scheduled.
- Supervised day-one draft preserved, but QA and Proof handoff failed on 16 Jul 2026.
| - Correct the unsupported medical-language rule hits
- Derive source-access timestamps from runtime evidence, not model prose
- Run a separately supervised retry and review QA before Proof routing or scheduling
- Schedule only after every prerequisite and a supervised run are approved.
|
| HUX-L5Creative refreshEvent-driven | Detection to draft with zero human effort. Publishing still needs a tap.In practice: The ads watchdog's fatigue flags : frequency creep + CTR decline crossing thresholds together | Built — awaiting credentials- Tested with fixtures or read-only data, but not scheduled.
- Built and fixture-proven as part of the ads quartet: L3's hand-off file drives the refresh trigger, thresholds tested including deliberate non-triggers, and a degraded upstream (no data) propagates honestly instead of guessing.
| - L1 and L3 live with a clean month each
- Then: wire the trigger (it's wiring, not new machinery)
- Schedule only after every prerequisite and a supervised run are approved.
|
| HUX-L6Negative-keyword janitorWeekly | Our whole portfolio is read-only. This is the one, carefully chosen exception.In practice: The ads watchdog's weekly zero-conversion search terms — e.g. any query over AED 100 spend with no sales in 30 days | Built — awaiting credentials- Tested with fixtures or read-only data, but not scheduled.
- Built and fixture-proven; the write is deliberately unimplemented.
| - HUX-L3 live and matching Sherif's judgment for 4 consecutive weeks
- Then, and only then: build the scoped write
- Schedule only after every prerequisite and a supervised run are approved.
|
| Marketing — B2B pipeline |
|---|
| B2B-L1LinkedIn account watchdog & auditWeekly | If we ever spend on LinkedIn, it starts healthy and stays watched.In practice: Spend pacing and zero-converting audiences | Design on the shelf- Design only; no production runtime exists.
- Design on the shelf, activation gated on two decisions: a B2B marketing budget exists, and someone owns it.
| - B2B marketing budget + named owner (org decision)
- LinkedIn ads account + Advertising API approval (free — request now)
|
| B2B-L3Lead-form capturePer lead | Paid leads with their origin story attached.In practice: Incoming LinkedIn lead-form submissions | Design on the shelf- Design only; no production runtime exists.
- Design on the shelf — activates with the B2B ladder once B2B-L1 exists (it needs campaigns to exist before forms do).
| - B2B-L1 live (campaigns before forms)
- BU-lead approval workflow agreed
|
| B2B-L2Audience builder from our CRMPer batch | The most powerful B2B targeting is our own customer knowledge — handled with the most care.In practice: Nothing continuously — it runs per campaign request : build audience → review → upload | Design on the shelf- Design only; no production runtime exists.
- Design on the shelf, last in the activation queue.
| - CRM-1 hygiene sustained (dup rate <5%)
- B2B-L1 proving itself
- PDPL consent posture confirmed with Rajiv/legal
|
| B2B-L4Outbound assistant — human-sentWeekly packages | The signal engine — everything automated except the one thing that shouldn't be.In practice: Qualified buying signals from the lead-intelligence loop (a hotel entering early planning, a tender award) | Design on the shelf- Design only; no production runtime exists.
- Design complete (the absorbed go-to-market project, loop-ified) with vendor verdicts and phased costs on the go-to-market page .
| - CRM-3 shortlists showing conversion (the free test)
- Phase-1 budget (~$450–650/mo) on that evidence
|