Business Requirements Document v0.1

Miro Brain

A centralised context and data intelligence layer providing trusted, validated data to AI agents across Miro's business functions.

Author
Harry Castle
Date
20 Feb 2026
Status
Draft
Phase 1
Sales Brain
Scroll

02 / Business Objectives

What Miro Brain delivers

Five measurable outcomes that define success for this initiative.

O1

Single trusted data layer

Agents across all four domains pull from Miro Brain rather than querying raw sources directly.

O2

Eliminate data hallucinations

Validation mechanisms catch >95% of format, filter, and logic errors before they reach end users.

O3

Accelerate time-to-value

New agent use cases deployed in <2 weeks by composing existing data products, skills, and integrations.

O4

Cross-functional insights

Agents reason across domain boundaries — e.g. correlating support ticket volume with ARR churn.

O5

Governed engagement model

Every persona has a defined interaction pattern with clear escalation and feedback loops.

04 / Architecture

Four layers, one brain

From user interaction to data infrastructure — every layer has a distinct responsibility.

Miro Brain Architecture Diagram
👤

Engagement Layer

Business personas interact via Slack, Web App, Workato-Go, or dashboards. AEs, CSMs, analysts, and leadership each have tailored interaction patterns.

🔀

Agent Orchestration

Workato handles agent routing, context assembly, response validation, callback protocol, and session management.

🧠

Miro Brain Core

Four domain sub-brains (Sales, CS, Finance, Marketing) unified by a shared context and validation layer.

💾

Data & Integration

Snowflake, Databricks, Workato, Salesforce, Stripe, and the RAG Repository built with Connected Systems.

05 / Domain Sub-Brains

Four brains, one architecture

Each sub-brain follows an identical anatomy: Data Products, Integrations, Skills, Metrics, and Use Cases — all mapped to real Snowflake tables.

📈

Sales Brain

Empower sales teams (AEs, SDRs, Sales Leadership) with trusted pipeline, account, and deal intelligence. Phase 1 priority — targeting Q1 2026.

Integration Architecture

End-to-end Sales Brain architecture — from engagement surfaces (chatbot, micro site, dashboard) through RBAC-validated access, into Workato's orchestration hub. Workato both ingests data and takes action: updating SFDC, drafting Outreach sequences, triggering enrichment. Gong transcripts and Slack channel history flow directly into the RAG repository.

Sales Brain Integration Architecture — Workato hub connecting SFDC, Gong, LinkedIn Sales Nav, Outreach, Loopio, Snowflake, Common Room, Clay, ZoomInfo to Sales Brain and RAG Repository
Arrow Types
Data ingestion (read data in) Actions & write-back (take action out) Unstructured content → RAG (direct) RBAC-validated user requests
Engagement & Access
Chatbot — Conversational agent interaction Micro Site — Lightweight web app for sales tools Dashboard — Analytics & reporting views RBAC — Role-based access validated via Workato
CRM & Pipeline
SFDC — Read pipeline + Update records & fields
Prospecting & Outbound
LinkedIn Sales Nav — Prospect insights Outreach — Read engagement + Draft sequences & messages
Enrichment & Signals
Common Room — Community signals Clay — Read enrichment + Trigger enrichment jobs ZoomInfo — Read contacts + Enrich contacts
Conversation Intelligence → RAG
Gong — Call transcripts → RAG (direct) Slack — Account channel history → RAG (direct)
Product & Content
Snowflake — Product usage data, analytics Loopio — RFP responses, proposal content

Data Products

  • Pipeline AnalyticsOpen pipeline by stage, segment, forecast category. S2+ funnel analysis.
  • Account IntelligenceEnriched account view: segment, ICP, territory, health scores, whitespace.
  • Deal VelocityStage-to-stage velocity, time-in-stage, win rate trends.
  • Territory & OwnershipTerritory mapping, rep coverage, account assignments.
  • GTM User ProfilesSales-enriched user view with contact data and lifecycle scoring.

Skills & Metrics

  • query_pipelineQuery open pipeline by segment, stage, forecast, owner
  • account_briefEnriched account summary (ARR, health, whitespace)
  • deal_risk_assessmentFlag stalled stages, declining forecast, missing next steps
  • territory_analysisTerritory coverage, pipeline density, rep performance
  • forecast_roll_upAggregate forecast by category vs. target

Use Cases

S-1Pre-meeting Account BriefAE

Agent generates a one-page brief before a customer call — ARR, health, open opps, recent tickets, product usage.

📚 RAG Required — Connected Systems
Gong call transcripts Slack account channels CSM handoff notes Previous meeting notes Account plan documents
S-2Pipeline Risk AlertsSales Manager

Daily digest of deals that moved backward, stalled >14 days, or have forecast mismatches.

📚 RAG Required — Connected Systems
Sales playbook (risk criteria) Deal stage definitions & SLAs Forecast methodology docs Escalation procedures
S-3Forecast PreparationVP Sales

Auto-roll-up of commit / best-case / pipeline by region and segment for QBR.

📚 RAG Required — Connected Systems
Quarterly targets by region/segment QBR templates & narrative guidelines Previous QBR presentations Forecast category definitions
S-4Account Whitespace AnalysisAE / AM

Identify expansion opportunities based on seat utilisation, product adoption gaps.

📚 RAG Required — Connected Systems
Product capability matrix Pricing & packaging docs Expansion playbook Competitive battle cards Customer success stories by vertical
💚

Customer Success Brain

Equip CS teams with health, retention, and engagement intelligence to proactively manage the customer lifecycle.

Data Products

  • Customer HealthAccount-level health combining product usage, support, and commercial signals.
  • Retention & RenewalGRR, NRR, churn/contraction at deal and account level.
  • Support TicketsTicket volume, resolution time, CSAT, ticket-to-ARR correlation.
  • Product AdoptionDAU/MAU, activation %, feature adoption by organisation.
  • User EngagementActivation funnel, engagement scoring, reactivation tracking.

Skills & Metrics

  • health_checkCustomer health summary (scores, trends, risk signals)
  • churn_risk_assessmentFlag declining engagement + rising tickets + contraction signals
  • renewal_prepUsage trends, ROI metrics, expansion opportunities
  • ticket_analysisSupport trends per account (volume, severity, resolution)
  • adoption_reportActivation %, feature usage, DAU/MAU per org

Use Cases

CS-1Churn Early WarningCSM

Agent alerts CSM when an account shows compounding risk signals — health drop + ticket spike + usage decline.

📚 RAG Required — Connected Systems
Churn playbook & intervention SOPs Health score methodology docs Slack account channel history CSM account notes Executive sponsor contact records
CS-2Renewal PreparationCSM

Agent generates a renewal brief with usage stats, ROI narrative, and expansion recommendations.

📚 RAG Required — Connected Systems
Renewal playbook ROI narrative templates Product release notes (last 12 months) Pricing & packaging updates Customer success stories by use case
CS-3QBR Content GenerationCSM / CS Leader

Auto-generate quarterly business review slides with account metrics.

📚 RAG Required — Connected Systems
QBR slide templates Previous QBR decks per account Product roadmap highlights Best practice / adoption guides Account plan objectives
CS-4Support Escalation TriageCS Leader

Correlate ticket patterns with ARR to prioritise escalations.

📚 RAG Required — Connected Systems
Escalation policy & SLAs Known issues / bug tracker Support runbooks by issue type Engineering incident postmortems
💰

Finance Brain

Trusted revenue, billing, and forecasting intelligence for finance teams, FP&A, and Rev Ops.

Data Products

  • ARR AnalyticsSubscription-level monthly ARR by segment (SS/HT), change type, product.
  • Product RevenueARR broken down by 13 add-on products (AI, EKM, Enterprise Guard, etc.)
  • Customer RevenueEnriched customer-level revenue view with org/account/domain data.
  • Renewal EconomicsGRR/NRR per renewal opportunity at deal level.
  • Pricing AnalyticsQuoted ARR, price-per-seat, contract terms for enterprise deals.

Skills & Metrics

  • arr_snapshotCurrent ARR by segment, change type, product
  • revenue_waterfallMoM ARR bridge (new + expansion - contraction - churn)
  • product_arr_breakdownARR contribution by add-on product
  • cohort_retentionCohort-based retention curves
  • pricing_analysisPPS trends, ELA economics, tier distribution

Use Cases

F-1ARR ReportingFinance Analyst

Agent generates current ARR snapshot with segment, product, and geo breakdowns.

📚 RAG Required — Connected Systems
Metric definitions (ARR, MRR, NRR, GRR) Finance reporting standards Segment classification rules Board presentation templates
F-2Revenue WaterfallCFO / FP&A

Automated ARR bridge showing new, expansion, contraction, churn month-over-month.

📚 RAG Required — Connected Systems
Waterfall methodology docs Change type hierarchy rules FP&A commentary guidelines Budget/plan targets by quarter
F-3Cohort AnalysisFinance Analyst

Retention curves by acquisition cohort.

📚 RAG Required — Connected Systems
Cohort definition standards Benchmark retention targets Investor reporting templates
F-4Pricing Impact AssessmentRev Ops

Measure ARR impact of pricing changes — isolating organic growth from repricing events.

📚 RAG Required — Connected Systems
Pricing change log (dates, tiers, products) Pricing strategy documents ELA pricing curve & tier definitions Competitive pricing intelligence
📣

Marketing Brain

Attribution, campaign, and funnel intelligence to optimise demand generation and ABM strategy.

Data Products

  • Pipeline AttributionTouch-based and conversation-based attribution for enterprise pipeline.
  • Campaign PerformanceCampaign member engagement aggregated to account level.
  • User AcquisitionRegistration trends by marketing channel and attribution bucket.
  • Mktg-Influenced Pipeline% of open pipeline touched by marketing activities.

Skills & Metrics

  • attribution_reportInfluenced vs sourced pipeline by quarter
  • campaign_roiCampaign performance: spend vs attributed pipeline
  • funnel_analysisRegistration → activation → engagement funnel
  • channel_mixAttribution bucket distribution for new registrations

Use Cases

M-1Attribution DashboardMktg Ops

Marketing attribution breakdown (influenced vs sourced) for current quarter pipeline.

📚 RAG Required — Connected Systems
Attribution model documentation Touch vs sourced definitions Marketing channel taxonomy Pipeline target allocations by channel
M-2Campaign Performance SummaryDemand Gen

Campaign ROI, engagement rates, pipeline contribution.

📚 RAG Required — Connected Systems
Campaign briefs & objectives Spend data by campaign ROI calculation methodology Benchmark performance by campaign type
M-3Funnel Health CheckGrowth Mktg

Registration-to-engagement funnel by channel with drop-off analysis.

📚 RAG Required — Connected Systems
Funnel stage definitions Activation & engagement criteria docs Growth experiment history & results Onboarding flow documentation
M-4ABM Target EnrichmentABM Manager

Enrich target account list with product usage, whitespace, and engagement data.

📚 RAG Required — Connected Systems
ICP definitions & scoring criteria ABM tier strategy docs Target account selection methodology Vertical-specific value propositions Competitive intelligence reports

06 / Data Validation Framework

Trust through validation

Six categories of validation rules ensure agents never return incorrect, misleading, or hallucinated data.

🛡️

Mandatory Filters

Required WHERE clauses auto-applied. E.g. USER: exclude Miro employees and test users.

🔤

Format Validation

Enforce correct casing. lifecycle_state is lowercase, user_top_license is UPPERCASE.

📊

Sanity Checks

Result-range validation. DAU should be ~1.4M — if >10M, likely missing DISTINCT.

🔗

Join Safety

Product PKs are numeric; CRM PKs are 18-char text. Never join without CAST.

⚠️

Null Awareness

organization.arr is 98.8% NULL — always route to MONTHLY_ARR_HYBRID instead.

🎯

Table Selection

Route queries to the correct table. Org-specific DAU → GROWTH.ORGANIZATION_USER_DAILY_ACTIVITY.

07 / Engagement Layer

Who interacts, and how

Every persona has a mapped sub-brain, interaction mode, and set of use cases.

Persona Sub-Brain(s) Interaction Key Use Cases
Account ExecutiveSales, CSSlack / Web AppS-1, S-4
SDRSalesSlack / Web AppOutbound targeting
Sales Manager / VPSalesDashboard / Workato-GoS-2, S-3
CSMCS, SalesSlack / Web AppCS-1, CS-2, CS-3
CS LeaderCSDashboard / Workato-GoCS-4
Finance AnalystFinanceWeb App / DashboardF-1, F-3
CFO / FP&AFinanceDashboard / Workato-GoF-2, F-4
Marketing OpsMarketingWeb App / DashboardM-1, M-2
ABM ManagerMarketing, SalesWeb App / Workato-GoM-4
Leadership / ExecAllDashboard / BriefingCross-domain

Agent-to-Brain callback

1
User Query — Agent interprets intent and identifies required sub-brain(s)
↓
2
Brain API Call — Domain, data product, skill, and parameters sent to Brain
↓
3
Validation — Mandatory filters applied? Formats correct? Right table selected?
↓
4
Execution — Brain executes query or retrieves context from RAG
↓
5
Response — Data payload + freshness timestamp + confidence score + caveats returned
↓
6
Assembly — Agent composes user-facing response from validated data

08 / Technology Stack

Built on proven foundations

Snowflake
Primary data warehouse. All sub-brain data products query here.
Databricks
Advanced analytics, ML models, revenue forecasting.
Workato
Agent orchestration, workflow automation, integration recipes.
Salesforce
CRM data — accounts, opportunities, contacts, campaigns.
Stripe
Self-serve billing and subscription data.
RAG Repository
Playbooks, SOPs, product docs, Slack, transcripts, policy docs. Built with Connected Systems.

10 / Implementation Phases

From foundation to full brain

Q1 2026

Phase 0 — Foundation

Core architecture, Workato agent orchestration, validation framework, RAG repo setup with Connected Systems, data contracts for top 5 entities.

Q1 2026

Phase 1 — Sales Brain

Sales data products, 5 skills, 4 use cases (S-1 through S-4), AE and Manager engagement.

Q2 2026

Phase 2 — CS Brain

CS data products, health scoring, churn detection, CSM engagement workflows.

Q2 2026

Phase 3 — Finance Brain

ARR analytics, product revenue breakdowns, retention analysis, revenue forecasting.

H2 2026

Phase 4 — Marketing Brain

Attribution models, campaign performance, funnel analysis, ABM target enrichment.

H2 2026

Phase 5 — Cross-Domain

Cross-brain queries, executive dashboards, advanced composite skills spanning multiple domains.

11 / Dependencies & Risks

Known risks, planned mitigations

High Connected Systems team capacity for RAG repository Early alignment on scope and timeline
Medium Data freshness SLAs vary by table (some lag 1-2 days) Freshness metadata in every brain response
Medium MONTHLY_ARR_HYBRID is 1,447 lines of SQL — most complex model Finance analytics team must be stakeholder
Medium Workato agent capabilities need validation for routing/callback requirements Spike in Phase 0 to confirm fit
Medium Cross-domain joins require careful PK type handling Built into validation layer from day one
Medium Some marketing data sources still TBD Phase 4 scoping to confirm connectors