# 70 agents. One operator.

BrainstormGTM compresses a 5-person marketing team into one operator + 70 specialized agents running inside the OODA loop. Outbound, content, intelligence, pipeline — each a swarm. Continuous, evidence-chained, ChangeSet-gated.

14:23:41.082 outbound EMAIL_SEND · seq=icp-04 step=2/4· msg:8f3a2c…
14:23:39.451 intelligence INTENT_SIGNAL · cobalt.systems · score=0.81· sig:b7c1d9…
14:23:36.108 content ARTICLE_PUBLISH · "agentic ops in MSPs" · 1850 words· art:4d8f17…
14:23:33.762 pipeline OPP_SCORE · dana.hartwell → 0.74 · stage=discovery· opp:a52b1e…
14:23:30.224 outbound REPLY_CLASS · positive · book=Wed-1400-PT· rpl:c419fa…
14:23:27.984 content SOCIAL_DRAFT · linkedin · 3 variants· soc:e2b48d…
14:23:25.451 intelligence COMPETE_DELTA · vendor X · pricing page edit· sig:5740c1…
14:23:22.108 pipeline NEXT_BEST · acct=cobalt → schedule disco call· opp:9a8d3f…
14:23:18.762 outbound LIST_ENRICH · 240 contacts · 86% match· lst:31b274…
14:23:15.451 content NEWSLETTER_ASSEMBLE · segment=msp-ops · 6 sections· nwl:7e51c2…

## 01 — the flow

### From signal to closed deal.

Every signal — a vendor mention, a hiring change, a funding event, a research click — enters the same five-phase flow. The swarm reasons, drafts, sends, classifies, routes. Operator touches the high-leverage moments.

#### Signal
Vendor news, hiring change, funding event, intent signal.

```plaintext
intent: deep_research_visit · article: "agentic ops" · score: 0.74
```

#### Persona
PersonaResearcher synthesizes role, bio, recent activity.

```plaintext
name: Dana Hartwell · title: VP RevOps · co: Cobalt Systems
```

#### Outreach
EmailComposer + LinkedInOps draft + sequence in operator voice.

```plaintext
channel: linkedin → email · step 1/4 · subject: "noticed the cobalt + RevOps angle"
```

#### Reply
ReplyClassifier routes positive/negative/OOO; CalendarBroker books on positive.

```plaintext
class: positive · intent: discovery · slot: 2026-05-23 14:00 PT
```

#### Pipeline
OpportunityScorer creates a deal; DealRoomBuilder spins up collaboration space.

```plaintext
stage: discovery · score: 0.81 · room: dlrm-cobalt-q2
```

70 specialized agents

4 suites

1 operator

24/7 continuous run

5× team multiplier

OODA reasoning loop

∞ evidence chain

ICP scoped continuously

## 02 — the swarm

### 70 specialized agents. Four suites.

Outbound (18) + Content (20) + Intelligence (14) + Pipeline (18). Filter by suite to see the specialists; click any agent to read its role. All MCP-exposed — drive any one from Claude Code or the operator console.

#### PersonaResearcher
Per-prospect bio + role + recent activity synthesis

#### EmailComposer
Multi-step sequence drafting per ICP voice

#### LinkedInOps
Connection requests + InMail orchestration

#### WarmCaller
Outbound dial sequencing with context briefs

#### ReplyClassifier
Auto-routes positive/negative/OOO replies

#### ComplianceGate
CAN-SPAM, GDPR, jurisdiction-specific pre-send check

#### DeliverabilityGuard
Per-domain sender reputation + warmup

#### SubjectOptimizer
A/B testing across ICP segments

#### CalendarBroker
Meeting time negotiation + booking

#### HandoffOrchestrator
Qualified-lead handoff to sales operator

#### TimingLearner
Per-prospect best-time-to-reach inference

#### ListBuilder
ICP-scoped list construction + enrichment

#### EnrichmentBroker
Multi-source data merge for prospect records

#### UnsubscribeKeeper
Suppression list integrity + audit log

#### TemplateGuardian
Per-channel template versioning + retirement

#### SequenceController
Multi-touch cadence orchestration

#### BounceTriage
Hard/soft bounce classification + list hygiene

#### PrintBroker
Direct-mail integration for high-value prospects

#### LongFormWriter
2000-word article drafts in operator voice

#### SocialDrafter
Per-platform short-form variants

#### VideoScripter
Hook + body + CTA scripts for short video

#### PodcastPrep
Guest brief + episode outline + question list

#### NewsletterEditor
Per-segment weekly newsletter assembly

#### CaseStudyBuilder
Customer-story drafting from CRM data

#### StyleTransfer
Voice consistency across artist + operator

#### SeoOptimizer
Per-keyword article tuning + linking

#### ImagePrompter
Hero image briefs for downstream image gen

#### CrossPoster
Multi-platform syndication with attribution

#### EditorialCalendar
Topic queue + cadence + dependency tracking

#### FactChecker
Per-claim citation verification before publish

#### LegalReviewer
Trademark + libel + claims-substantiation pass

#### HookFinder
Headline + opening-line optimization

#### PullQuoteExtractor
Surface tweetable lines from long-form drafts

#### BlogPublisher
Headless-CMS deploy + sitemap update

#### ContentRetriever
Vector-memory recall for evergreen content

#### TranslationBroker
Multi-locale variants for global content

#### AccessibilityScribe
Alt-text + transcripts + reading-level

#### AnalyticsAttribution
Content-to-pipeline attribution at piece level

#### CompetitiveMonitor
Per-competitor activity + product changes

#### IndustrySignals
Vertical-news ingestion + clustering

#### CustomerVoice
Support + review + interview synthesis

#### DealRoomBrief
Per-opportunity pre-call brief

#### CRMObserver
Stage-transition signal extraction

#### ProspectIntent
Buying-signal scoring from public + 1P data

#### TrendForecaster
Topic-momentum prediction for content scheduling

#### FundingTracker
TAM segment funding events + buyer activation

#### HiringMonitor
Org-chart change signal + persona shifts

#### ChurnAnalyst
Customer-churn pattern extraction

#### EventScout
Conference + meetup ICP-relevance scoring

#### NewsAggregator
Multi-source news clustering + summarization

#### AnalystSpotter
Per-analyst report monitoring + extraction

#### BoardroomNote
Quarterly business review draft assembly

#### OpportunityScorer
Per-deal probability + stage-fit scoring

#### StageProgressor
CRM stage auto-advance on signal threshold

#### NextBestAction
Per-opportunity recommended next step

#### ForecastEngine
Quarterly forecast accuracy regression

#### RouteController
Lead routing to operator + handoff context

#### DealRoomBuilder
Per-deal collaboration space spin-up

#### ProposalDrafter
Per-opportunity proposal generation

#### ContractStager
Per-stage contract draft + redline tracking

#### PricingAdvisor
Per-opportunity pricing optimization

#### CRMOps
Cross-system data sync + cleanup

#### PipelineSentry
Stale-opportunity surfacing + escalation

#### ChannelAttribution
Per-opportunity multi-touch attribution

#### WinLossAnalyzer
Per-deal post-mortem with pattern extraction

#### TerritoryMapper
Per-rep territory health + capacity

#### QuotaProjector
Per-rep + per-segment quota attainment forecast

#### RenewalKeeper
Existing-customer renewal pipeline orchestration

#### UpsellSpotter
Per-customer expansion-signal detection

#### NoiseFilter
Deal-flow noise filtering for operator attention

## 03 — the math

### Hours in, output multiplied.

Drag the slider to see what one operator with the swarm produces vs a traditional 5-person marketing team. Numbers are conservative; real customers hit higher.

### How much does one operator hour produce?

#### Operator hours per week
15h

The swarm runs continuously; the constraint is operator review time. At about 25 hours per week the operator becomes the bottleneck — beyond that, throughput grows more slowly.

#### Weekly output

- Outbound msgs: 3,240  
- Long-form: 12  
- Intel briefs: 21  
- CRM updates: 90

## 04 — the spectrum

### Tune the swarm to your operating model.

Founder-led, sales-led, demand-gen-led — each adjusts which actions auto-execute vs which the operator drives. Move between models as the business evolves; the audit chain captures every shift.

### Model 01
#### Founder-led
Operator drives every nuance.

**Swarm auto-executes**
- Persona research
- Intent signal monitoring
- Content suggestions to operator
- CRM data hygiene

**Operator drives**
- Every outbound send
- Every published article
- Every deal-room creation
- Pricing decisions

### Model 02
#### Sales-led
Swarm runs outreach; sales drives close.

**Swarm auto-executes**
- Persona research
- Intent signal scoring
- Outbound sequences (within policy)
- Reply classification + routing
- Calendar booking on positive reply
- Content publishing on cadence

**Operator drives**
- Multi-touch deal-room curation
- Proposal drafting
- Pricing negotiation
- Renewal conversations

### Model 03
#### Demand-gen-led
Full autopilot for top-of-funnel; humans on warm leads.

**Swarm auto-executes**
- Persona research
- Intent signal scoring
- Outbound sequences (full cadence)
- Reply classification + routing
- Calendar booking
- Content publishing
- Newsletter assembly + send
- Social cross-posting
- Opportunity scoring + stage progression

**Operator drives**
- Sales-stage human handoffs (warm)
- Pricing negotiation
- Custom enterprise proposals

## 05 — vs traditional marketing stack

### Different shape. Not different tools.

Operating shape
- Traditional stack: 5-person team — SDR, content writer, growth marketer, designer, ops manager.
- Brainstorm GTM: One operator + 70-agent swarm. Same domain expertise, multiplied throughput.

### Cadence
- Traditional stack: Weekly stand-up, quarterly planning, content calendar in a spreadsheet.
- Brainstorm GTM: Continuous OODA loop. Cadence is the operator setting compliance gates, not booking meetings.

### Voice consistency
- Traditional stack: Style guide PDF; drift across team members; review cycles slow.
- Brainstorm GTM: StyleTransfer agent ensures every published asset matches operator voice — measured, not asserted.

### Audit trail
- Traditional stack: CRM activity log; email tool report; manual cross-reference.
- Brainstorm GTM: Every send, publish, score, route is an evidence-chained ChangeSet. Replayable end-to-end.

### Compliance posture
- Traditional stack: Manual CAN-SPAM/GDPR check per campaign; risk of misconfiguration.
- Brainstorm GTM: ComplianceGate runs pre-send on every message. Jurisdiction-aware. Audit log retained.