Govern swarm · Governance-decision-router agent · Five-destination routing skill · Build pillar · Published July 13, 2026

How to build five-destination routing for AI-decision approval

Every AI-drafted artifact (marketing claim, creative, product description, review response, social post, ad copy, email, SMS, push, landing-page copy, blog post, FAQ, GBP post, call-transcript summary) reaches one of exactly five terminal destinations: auto- publish, batch-review, send-to-FBC, escalate-to-team-lead, or reject-with-feedback. This is precisely the human-oversight architecture EU AI Act Article 14 requires for high-risk AI systems plus the transparency-of-automated-decisions architecture Article 22 requires and the deployer-obligations Article 26 requires. The Score + Route + Gate + Audit skill bundle on the governance-decision-router agent sits above your existing Asana + Monday + ClickUp + Wrike + Workfront + Salesforce Approvals + Workato + n8n approval-workflow surface and writes the per-draft routing decision with named regulatory anchors preserved in every audit record.

Start with a Tier 1 AI Readiness Assessment Tier 3 Fractional CMO with AI Swarm Take the 3-question quiz

The 4-skill bundle on the governance-decision-router agent

Score

Five coordinated dimensions per draft: brand-voice score (tone + vocabulary + cadence + syntactic fingerprint + emotional register + forbidden / required phrase) grounded in the runtime brand-voice gate sibling; sentiment (positive / negative / neutral + 5-emotion + sarcasm + aggression vs empathy); factual confidence (claims-allowlist substantiation + named-entity disambiguation + numeric-claim validation + LLM self- consistency + multi-LLM ensemble agreement across OpenAI + Anthropic + Google + Mistral + Cohere + Meta under per-vendor zero-retention); reach magnitude (impression projection + audience size + per-platform amplification + virality coefficient); compliance confidence (per-anchor pre-evaluation across the regulatory stack). Per-dimension confidence tier + explainability written to Audit.

Route

Five terminal destinations. Auto-publish (every dimension cleared threshold; publish with per-draft autonomy-profile pointer + EU AI Act Article 50 disclosure stamp). Batch-review (per-reviewer queue + per-batch window + SLA + priority). Send-to-FBC (Forbidden-Borderline-Confirmed reviewer for confirm / reject / return with rationale that feeds threshold recalibration + pattern-learning). Escalate-to-team-lead (gate that requires senior review — regulated-vertical exposure, Annex III high-risk classification, per-vendor LLM zero- retention failure). Reject-with-feedback (hard gate failure; explainer written back to drafting agent for next-attempt learning). Per-destination escalation timers fire fallback actions on SLA expiry. FBC override capture writes reason + reviewer + timestamp + rationale + threshold-recalibration target + cross-draft / cross-banner / cross-location correlation into Audit; learnings flow into Score thresholds and the FBC-override-learning sibling build-pillar.

Gate

Five anchors before destination commit. EU AI Act Article 14 human oversight (this is precisely what the five-destination pattern implements) + Article 13 + 15 + 22 + 26 + 50 + Annex III + Article 9 + 10 + 11 + 12 + NIST AI RMF + ISO 42001 + 23894 + 5338. FTC Section 5 + Pfizer 1972 + Endorsement Guides 16 CFR Part 255 (2023 AI-content) + Fake Review Rule 16 CFR Part 465 (October 2024) + MARS + Made-in-USA + Green Guides + Health Products Compliance Guide + Negative-Option + Lanham. Per-ad-platform policy (Meta + TikTok + Snap + Pinterest + Reddit + LinkedIn + YouTube + X + Google Ads + Amazon + Walmart Connect). Per-vertical overlays (HIPAA + FINRA 2210 + SEC Reg S-K + FDA 21 CFR Part 11 + DSCSA + CPSC CPSIA + EPA FIFRA + DEA + CFPB UDAAP + FDD Item 19 per FTC Franchise Rule 16 CFR 436). ADA Title III + WCAG 2.2 AA (Robles 2019 + Gil 2021) + California Unruh + Colorado SB21-176 + NYC Local Law 12 + CCPA opt-out + CPRA + 17-state + GDPR Article 22 + Digital Services Act + per-vendor LLM zero-retention. Policy- as-code via OPA Rego + AWS Cedar + Casbin + Cerbos + Oso + Styra DAS + Permit.io.

Audit

Per-draft WORM record: routing-decision ID + per-banner + per-location + per-AI-agent + draft snapshot + Score snapshot + Route decision + per-anchor Gate decision with evidence + per-vendor LLM zero-retention verification + escalation timer state + FBC override (reason + reviewer + timestamp + rationale + threshold-recalibration target + pattern-learning seed + cross-draft / cross-banner / cross-location correlation) + sibling-handoff pointers. Storage: AWS S3 Object Lock + Azure Blob immutable + Google Cloud Storage Bucket Lock + Wasabi WORM. Retention stacks: 7-year FTC + 7- year IRS + 7-year FDD + per-state franchise + GDPR Article 30 + EU AI Act Article 12 record-keeping + SOC 2 CC7 / CC8. End-to-end replay rewinds every stage.

The real vendor ecosystem this sits above

Approval workflow + reviewer surface

Asana, Monday.com, ClickUp, Wrike, Adobe Workfront, Smartsheet, Kissflow, ProcessMaker, Nintex, Camunda, Pega, Appian, ServiceNow, Salesforce Approvals, Workato, Tray.io, n8n, Make, and Zapier remain the assignee-based reviewer-handoff substrate where the batch-review + send-to-FBC + escalate-to-team-lead destinations materialize as queue items.

LLM ensemble + observability + safety + evaluation

OpenAI + Anthropic + Google + Mistral + Cohere + Meta + AWS Bedrock + Azure OpenAI + Vertex AI LLM providers under per-vendor zero-retention. LangSmith + Weights & Biases + Arize + WhyLabs + Helicone + Langfuse + PromptLayer + Galileo observability. Lakera Guard + Robust Intelligence + HiddenLayer + CalypsoAI + Protect AI + Garak AI safety. DeepEval + Ragas + TruLens + Phoenix + UpTrain + Inspect AI + Promptfoo LLM evaluation back the Score multi-LLM ensemble agreement computation.

Policy-as-code + WORM + sibling skills

OPA Rego + AWS Cedar + Casbin + Cerbos + Oso + Styra DAS + Permit.io policy-as-code expresses every Gate rule including the EU AI Act Article 14 / 22 / 26 / 50 anchors, per-ad- platform policy, per-vertical overlays, and CCPA / GDPR automated-decision opt-out. AWS S3 Object Lock + Azure Blob immutable + Google Cloud Storage Bucket Lock + Wasabi compliance WORM holds the per-draft audit substrate. Sibling skills feed Score and consume Route: brand-voice runtime gate, claims-allowlist substantiation, forbidden-phrase library, borderline routing, multi-stream severity routing, routing- audit-trail, nested-autonomy profile-inheritance, marketing-AI- autonomy-profile-configuration, FBC-override-learning.

The 6-workstream reporting cycle

Numeric uplift commitments are not made up-front. The engagement ships a pre-engagement baseline across six workstreams; the cycle tracks delta against that baseline. Reporting is the substrate, not the promise.

Score coverage. Per-dimension coverage across brand voice + sentiment + factual + reach + compliance confidence; per-vendor LLM zero-retention verification per ensemble call; sibling-skill grounding completeness (brand-voice + claims-allowlist + forbidden-phrase).
Route quality. Per-destination distribution across the five terminal destinations; auto-publish share vs human-review share (the Article 14 lever); FBC override rate + recalibration cadence; escalation timer expiry rate; per-destination SLA adherence.
Gate quality. Per-anchor evaluation completeness (EU AI Act + NIST AI RMF + ISO 42001 + FTC + per-ad-platform + per-vertical + ADA / WCAG + CCPA opt-out + GDPR Article 22 + Digital Services Act); per-anchor pass / fail / route-to-counsel distribution; remediation-loop turnaround.
Audit quality. Per-draft WORM record completeness; retention-window coverage (longest of 7-year FTC + 7-year IRS + 7-year FDD + per-state franchise + GDPR Article 30 + EU AI Act Article 12 + SOC 2 CC7 / CC8); end-to-end replay success rate.
Compliance posture. Article 14 human-oversight coverage across destination distribution; Article 50 disclosure stamp coverage on auto-publish; Annex III high-risk classification detection rate; per-ad-platform policy adherence by platform; per-vertical overlay coverage; CCPA right-to-opt- out + GDPR Article 22 transparency stamp coverage.
Audit-trail completeness. Per-anchor regulatory citation completeness; sibling-handoff pointer completeness into the governance-decision-router bundle (content-approval- workflow + multi-stakeholder approval routing + multi- dimensional threshold routing + ai-routing-decision audit trail + tiered content filtering + auto-publish threshold gating + severity routing + borderline-routing build-pillar + multi- stream severity routing build-pillar + routing-audit-trail build-pillar + versioned-history regulatory-defense + nested- autonomy profile-inheritance + marketing-AI-autonomy-profile- configuration).

Frequently asked questions

What is five-destination routing for AI-decision approval — and why is it the EU AI Act Article 14 human-oversight pattern?

Five-destination routing is the architectural pattern by which every AI-drafted artifact (marketing claim, creative, product description, review response, social post, ad copy, email, SMS, push, landing-page copy, blog post, FAQ, GBP post, call-transcript summary) reaches one of exactly five terminal destinations: auto-publish (the draft cleared every threshold and gate), batch-review (the draft routed to a human reviewer queue with SLA), send-to-FBC (the draft routed to the Forbidden-Borderline-Confirmed reviewer for explicit confirm or reject with rationale that feeds threshold recalibration), escalate-to-team-lead (the draft tripped a gate that requires a senior reviewer), or reject-with-feedback (the draft is rejected with an explainer that feeds future drafts). This is precisely the human-oversight architecture EU AI Act Article 14 requires for high-risk AI systems plus the transparency-of-automated-decisions architecture Article 22 requires and the deployer-obligations Article 26 requires. The four-skill bundle on the governance-decision-router agent — Score, Route, Gate, Audit — sits above your existing Asana + Monday + ClickUp + Wrike + Workfront + Salesforce Approvals + Workato + n8n approval-workflow surface and writes the per-draft routing decision with named regulatory citations preserved in the audit trail.

Why do Asana + Monday + ClickUp + Wrike + Workfront + Salesforce Approvals + Workato + n8n break at multi-location AI-decision-approval scale?

Asana, Monday.com, ClickUp, Wrike, Adobe Workfront, Smartsheet, Kissflow, ProcessMaker, Nintex, Camunda, Pega, Appian, ServiceNow, Salesforce Approvals, Workato, Tray.io, n8n, Make, and Zapier ship per-tenant flat approval-workflow primitives — assignee-based handoff with status transitions. None composes a multi-dimensional threshold (brand-voice score + sentiment + factual confidence + reach magnitude + compliance confidence) and routes on the composition. None implements the five-destination pattern with the FBC reviewer queue that feeds threshold recalibration. None enforces the per-anchor compliance gate (EU AI Act Article 50 disclosure + FTC Endorsement Guides + Fake Review Rule + per-ad-platform policy + per-vertical overlays + CCPA right to opt out of automated decision-making + GDPR Article 22) before publish. None writes a WORM record of every routing decision with per-anchor evidence retained for the longest applicable regulatory window. The four-skill bundle Score + Route + Gate + Audit sits above the workflow surface — it does not replace it. Score computes the multi-dimensional threshold. Route picks the destination from the five. Gate evaluates the per-anchor compliance overlay. Audit writes the per-draft WORM record and captures FBC overrides that feed the next round of threshold calibration.

What does Score do — brand voice + sentiment + factual + reach + compliance confidence composition?

Score computes five coordinated dimensions per draft. Brand-voice score (tone of voice + vocabulary + cadence + syntactic fingerprint + emotional register + forbidden-phrase presence + required-phrase presence) grounded in the runtime brand-voice gate sibling skill. Sentiment score (positive / negative / neutral + emotion across anger / joy / sadness / fear / surprise + sarcasm detection + emotional-tone aggression vs empathy). Factual confidence score (claim substantiation against the claims-allowlist sibling skill + source citation + named-entity disambiguation + numeric-claim validation + LLM self-consistency cross-check + multi-LLM ensemble agreement across OpenAI GPT-4o + Anthropic Claude Opus + Google Gemini Pro 2 + Mistral Large 2 + Cohere Command R+ + Meta Llama 3 70B + Qwen 2 + DeepSeek V3 under per-vendor zero-retention). Reach magnitude score (impression projection + audience size + per-platform amplification factor + virality coefficient). Compliance confidence score (per-anchor pre-evaluation across the regulatory stack — see Gate). Each dimension carries a per-draft confidence tier and explainability trace written into Audit. The five scores compose into a routing-decision vector consumed by Route.

What does Route do — the five-destination decision + escalation timer + FBC override capture?

Route arbitrates the routing-decision vector into one of five terminal destinations. Auto-publish — every dimension cleared its threshold; the draft publishes with a per-draft autonomy profile pointer (sibling skill: nested-autonomy profile inheritance) and an Article 50 disclosure stamp written into the publish substrate. Batch-review — at least one dimension landed in the batch-review band; the draft routes to the per-reviewer queue with per-batch window + per-batch SLA + per-priority. Send-to-FBC — at least one dimension landed in the Forbidden-Borderline-Confirmed band (sibling build-pillar: borderline routing); the draft routes to the FBC reviewer who confirms, rejects, or returns with rationale that feeds threshold recalibration and pattern-learning. Escalate-to-team-lead — a gate tripped that requires senior review (regulated-vertical exposure, per-vendor LLM zero-retention failure, EU AI Act Annex III high-risk classification triggered); the draft routes to the team-lead pointer with SLA. Reject-with-feedback — at least one gate hard-failed; the draft is rejected with an explainer written back to the drafting agent for next-attempt learning. Per-destination escalation timers fire timer-elapsed fallback actions (re-route, escalation, auto-reject, audit snapshot) when SLA passes without resolution. FBC override capture writes per-override reason + reviewer + timestamp + rationale + explainer + threshold-recalibration target + pattern-learning seed + cross-draft correlation + cross-banner correlation into the audit substrate; the learnings flow back into Score thresholds and the sibling FBC-override-learning build-pillar.

What does Gate do — EU AI Act Article 14 + 13 + 15 + 22 + 26 + 50 + Annex III + NIST AI RMF + ISO 42001 + FTC + per-ad-platform + per-vertical anchors?

Gate evaluates five operationally distinctive anchors before any destination is committed. Anchor 1 (the most operationally distinctive): EU AI Act Article 14 human oversight (this is precisely what the five-destination routing pattern implements) + Article 13 transparency to deployers + Article 15 accuracy + robustness + cybersecurity + Article 22 transparency of automated decisions + Article 26 deployer obligations + Article 50 transparency for AI-generated content + Annex III high-risk classification (employment + creditworthiness + essential services + law enforcement + migration + administration of justice + democratic processes) + Article 9 risk-management system + Article 10 data and data governance + Article 11 technical documentation + Article 12 record-keeping; NIST AI Risk Management Framework Govern + Map + Measure + Manage functions; ISO 42001 AI Management System + ISO 23894 risk management + ISO 5338 AI system life cycle. Anchor 2: FTC Section 5 + FTC substantiation doctrine (Pfizer 1972 reasonable-basis) + FTC Endorsement Guides 16 CFR Part 255 (2023 update covering AI-generated content + algorithmic disclosure) + FTC Fake Review Rule 16 CFR Part 465 (effective October 21, 2024) + FTC MARS + FTC Made-in-USA Labeling Rule + FTC Green Guides + FTC Health Products Compliance Guide + FTC Negative-Option Rule + Lanham Act 15 USC 1125(a) + per-state UDAP + per-state Deceptive Trade Practices Act. Anchor 3 (per-ad-platform policy stack): Meta Advertising Standards + TikTok Ads Policy + Snap Advertising Policies + Pinterest Advertising Guidelines + Reddit Advertising Policy + LinkedIn Advertising Policies + YouTube Advertising Policies + X / Twitter Ads Policies + Google Ads Policies + Amazon Advertising Policy + Walmart Connect Advertising Policy. Anchor 4 (per-vertical regulatory overlays when the draft crosses regulated territory): HIPAA 45 CFR 164 when healthcare; FINRA Rule 2210 + SEC Reg S-K when financial services; FDA 21 CFR Part 11 + DSCSA + 21 CFR Part 101 + DSHEA + medical device when FDA-regulated; CPSC CPSIA when child products; EPA FIFRA when pesticides; DEA when controlled substances; CFPB UDAAP when consumer finance; FDD Item 19 per FTC Franchise Rule 16 CFR 436 + 15-state franchise registration when franchise system. Anchor 5: ADA Title III + WCAG 2.2 AA accessibility (Robles v Dominos Pizza 9th Cir 2019 + Gil v Winn-Dixie 11th Cir 2021) + California Unruh + Colorado SB21-176 + NYC Local Law 12; CCPA right to opt out of automated decision-making + CPRA + 17-state comprehensive privacy + GDPR Article 22 automated decisions + Digital Services Act Article 30; per-vendor LLM zero-retention verified per call. Policy-as-code expression via OPA Rego + AWS Cedar + Casbin + Cerbos + Oso + Styra DAS + Permit.io.

What does Audit do — per-draft WORM + FBC override capture + end-to-end replay?

Audit writes a per-draft WORM record at every routing decision: per-draft routing-decision ID + per-banner pointer + per-location pointer + per-AI-agent pointer + canonical-draft snapshot + Score snapshot (brand-voice + sentiment + factual + reach + compliance confidence with per-dimension confidence tier and explainability trace) + Route decision (one of the five destinations with rationale) + per-anchor Gate decision with evidence (EU AI Act Article 14 / 13 / 15 / 22 / 26 / 50 / Annex III / 9 / 10 / 11 / 12; NIST AI RMF; ISO 42001 + 23894 + 5338; FTC Section 5 + Endorsement Guides + Fake Review Rule + MARS + Made-in-USA + Green Guides + Health Products Compliance Guide + Negative-Option + Lanham; per-ad-platform policy stack; per-vertical overlays; ADA / WCAG; CCPA opt-out + CPRA + 17-state + GDPR Article 22 + Digital Services Act) + per-vendor LLM zero-retention verification + escalation timer state + FBC override (reason + reviewer + timestamp + rationale + explainer + threshold-recalibration target + pattern-learning seed + cross-draft / cross-banner / cross-location correlation) + sibling-handoff pointers. Storage on AWS S3 Object Lock + Azure Blob immutable + Google Cloud Storage Bucket Lock + Wasabi compliance WORM. Retention stacks (longest applicable wins): 7-year FTC substantiation + 7-year IRS tax + 7-year FDD record + per-state franchise registration + GDPR Article 30 records of processing + EU AI Act Article 12 record-keeping + SOC 2 CC7 / CC8. End-to-end replay rewinds Score + Route + Gate + escalation timer + FBC override with confidence tier and explainability at every stage. Sibling handoffs flow into the content-approval-workflow commercial pillar (parent), multi-stakeholder approval routing, multi-dimensional threshold routing, ai-routing-decision audit trail, tiered content filtering, auto-publish threshold gating, severity routing, the borderline-routing sibling build-pillar, the multi-stream severity routing sibling build-pillar, the routing-audit-trail sibling build-pillar, the versioned-history regulatory-defense sibling, the nested-autonomy profile-inheritance sibling, and the marketing-AI-autonomy-profile-configuration sibling.

Engage Completions on the governance-decision-router bundle

The Score + Route + Gate + Audit four-skill bundle ships as the orchestration layer above your existing approval-workflow surface and your LLM ensemble. The five-destination pattern implements the EU AI Act Article 14 human-oversight requirement directly. FTC Endorsement Guides + Fake Review Rule + per-ad-platform policy + per-vertical overlays + CCPA opt-out + GDPR Article 22 + NIST AI RMF + ISO 42001 anchors are preserved in every per- draft audit record. Tier 1 AI Readiness Assessment scopes the bundle in two to three weeks; Tier 3 Fractional CMO with AI Swarm operates the bundle end-to-end.

Tier 1 AI Readiness Assessment Tier 3 Fractional CMO with AI Swarm Take the 3-question quiz