Configure the system
What data, technology and controls do we actually need?
Translate the selected problem into a capability, a decision, required information and then technology. One reference architecture serves both small and advanced organisations; the difference is which responsibilities require shared services. sourcesSYS §§3–9,29–34AI Augmented GTM Systems Architecture — Systems architecture; proposed patterns, not an audited deployment.RT §§25–32AI Augmented GTM Red Team Verdict — Red Team; claim audit, investment challenge and simplification.Evidence for this chapter · 4 claim records
Customer, market, product and commercial signals
↓
Relevant trusted records and agreed economic definitions
↓
Analysis and interpretation → feasible intervention card
↓
Accountable decision and current authority checks
↓
Manual or native workflow execution through existing systems
↓
Delivery records, outcomes, costs and proportionate evaluation
↓
Reviewed evidence → next decision
Across the path: customer safeguards, named ownership, traceability and stop
One concrete minimum deployment
ModelM07ModelThe business owner selects an onboarding problem. Product and Finance reconcile a permitted export of account, activation and contribution records. An analyst records the diagnosis and comparison in a controlled document or spreadsheet. Product approves a repair or narrow test. The existing product release process or lifecycle platform delivers it. The analyst joins assignment and outcome records, Finance checks the value, and the owner writes a continue/change/stop decision.
An optional assistant can summarise already authorised feedback. It has no permission to change the customer record or publish a treatment. The workflow continues if the assistant is unavailable. This is a deployment pattern, not a claim of observed results.
Advanced additions need separate justification. Use this model with the conditions and limits described above.
What each responsibility needs
| Responsibility | Minimum | Add only when justified |
|---|---|---|
| Data and economics | Authoritative source records, relevant joins, agreed metrics and freshness | Shared analytical services when several funded workflows reuse them |
| Intelligence | Query, cohort analysis and customer research | Calibrated predictions, uplift or optional generative interpretation |
| Decisions | Named owner, alternatives, authority and a linked record | Shared decision service for recurring cross-system conflicts |
| Execution | Existing manual/native workflow with appropriate checks | Durable adapters for repeated machine writes and reconciliation |
| Measurement | Feasible comparison, outcome join and Finance review | Shared assignment infrastructure or specialist aggregate models |
| Learning | Searchable decision and evidence records | Specialised retrieval only when ordinary search is inadequate |
Preserve different entities and authoritative records: person, account, subscription and consent are not interchangeable. Unknown identity or stale permission blocks a dependent customer-specific action. Validate only the data the chosen decision needs; do not require an enterprise migration for a safe, useful drafting step. sourcesSYS §§4–7AI Augmented GTM Systems Architecture — Systems architecture; proposed patterns, not an audited deployment.RTB JAI Augmented GTM Playbook Editor Brief — Red Team summary; same workstream as RT, not corroboration.Evidence for this chapter · 4 claim records
Text version
- Data and economics: Minimum — Authoritative source records, relevant joins, agreed metrics and freshness; Add only when justified — Shared analytical services when several funded workflows reuse them.
- Intelligence: Minimum — Query, cohort analysis and customer research; Add only when justified — Calibrated predictions, uplift or optional generative interpretation.
- Decisions: Minimum — Named owner, alternatives, authority and a linked record; Add only when justified — Shared decision service for recurring cross-system conflicts.
- Execution: Minimum — Existing manual/native workflow with appropriate checks; Add only when justified — Durable adapters for repeated machine writes and reconciliation.
- Measurement: Minimum — Feasible comparison, outcome join and Finance review; Add only when justified — Shared assignment infrastructure or specialist aggregate models.
- Learning: Minimum — Searchable decision and evidence records; Add only when justified — Specialised retrieval only when ordinary search is inadequate.
Advanced additions and their prerequisites
| Addition | Fix first | Entry evidence | When to defer |
|---|---|---|---|
| Predicted customer value | Customer linkage, contribution definitions, cohort history | Temporal calibration, useful cohort discrimination and prospective decision benefit | Forecast is unstable or does not improve a choice |
| Uplift or next-best-action model | Available treatments, assignment, sufficient outcomes | Valid treatment evidence and out-of-sample policy benefit | Only a few treatments or sparse effects; rules suffice |
| Portfolio models and MMM | Comparable horizons, variation, cost and budget rights | Identifiable response ranges and feasible decisions | No independent variation or resources cannot move |
| Shared coordination service | Clear local rights and existing workflow baseline | Recurring friction or policy inconsistency exceeds ownership cost | A responsibility table and native tools suffice |
| Knowledge graph | Clean linked records, valuable recurring questions, curator | Valuable queries outperform SQL/full-text baseline after maintenance cost | Relationships exist but no query advantage is demonstrated |
| Vector retrieval | Permissioned corpus and source truth | Better retrieval against ordinary search with scope and stale-data tests | Exact records or financial calculations are the need |
| New real-time infrastructure | A latency requirement and reliable native event handling | Measured delay loss exceeds the additional cost and risk | Decision is daily, weekly or slower |
| Bounded agent | A fixed-workflow baseline and evaluation | Dynamic investigation improves accepted work net of controls | Tools and steps are already known |
The Control Plane is optional technical shorthand for shared policies, approvals, monitoring and coordination. It does not prescribe a custom portal. Build only a valuable residual gap after testing existing/native capability. Compare control, portability, integration, maintenance, vendor maturity and exit alongside licence price. sourcesRT §§26,29–32AI Augmented GTM Red Team Verdict — Red Team; claim audit, investment challenge and simplification.Evidence for this chapter · 4 claim records
Related
Evidence for this chapter4claims reviewed, with confidence ratings, limits and sources.
Claim classifications
Accepted means a principle or mechanism survived critical review, not that it guarantees a return. Accepted with conditions means prerequisites apply. Qualified means the original claim has been narrowed. Advanced needs substantial readiness. Experimental needs a capped comparison. Unresolved means the evidence cannot settle the claim. Rejected means the recommendation is excluded.
High, Medium and Low describe confidence in the stated target, not the chance of a positive ROI. For a Qualified or Rejected historical claim, confidence applies to the decision to narrow or reject it and to the supported limit—not to the original headline. Read the headline and classification together. An accounting identity or design principle is not empirical evidence of a commercial outcome. Follow the linked chapters, Plays and tools to see where each claim applies.