The AI-Augmented GTM Playbook
Start with a customer or commercial problem. Work out what it costs, choose an action someone can deliver, and measure what changes. Use AI when it improves the work enough to justify its cost. Compare it with the best practical alternative before scaling.
- SenseDated signal and source
- UnderstandDiagnosis, alternatives and uncertainty
- DecideNamed owner, rationale, limits
- ActivateAction version, delivery or receipt
- MeasureScoped estimate and guardrails
- LearnFinding, limits, owner and review trigger
- OptimiseNext decision and revised Play
Seven logical responsibilities; not seven products.
Why it matters
More campaigns, content and automation will not fix a weak proposition, broken onboarding or a team that cannot act. Find the problem holding growth back, then choose the simplest reliable way to address it.
The smallest useful system
Choose one important problem. Name one decision owner. Establish credible facts and economics. Compare a feasible intervention with a competent conventional alternative and doing nothing. Deliver it through existing tools. Apply customer safeguards, evaluate the outcome and retain a short decision record.
twelve operating principles
ch01- Fund an intervention with an available action and an accountable owner.
- Use contribution and cash over an explicit horizon, within customer and strategic constraints.
- Repair the binding growth constraint before optimising its downstream symptoms.
- Compare AI with competent process repair; compare integration with the simpler workflow.
- Keep observation, prediction, attribution and causation distinct.
- Buy evidence when it can change a material decision at proportionate cost.
- Choose the least complex reliable mechanism.
- Retain human authority for strategy, major investment, pricing, novel claims and ambiguous customer consequences.
- Enforce exact action limits outside the model and fund meaningful review.
- Record saved time as capacity until expenditure or useful output changes.
- Reuse scoped evidence and retire stale findings and obsolete work.
- Make complexity, specialist roles and autonomy earn their full ownership cost.
These principles consolidate STR §46, AGT §5, OPS §3 and RT §83. They govern the Plays; they are not estimates of financial return.
A decision-led cadence
Cadences are configurable. A small firm can combine reviews. Every new recurring review replaces existing work or has an explicit capacity cost. Train through the selected workflow: practitioners verify evidence, managers practise real decisions, operators rehearse failures and Finance checks actual benefits.
Daily where needed
Exceptions, broken delivery, current safeguards and containment
Avoid — Reopening strategy on ordinary noise
Weekly
One-page outcome, diagnosis, alternatives, action, owner and readout
Avoid — Slide production without a decision
Monthly or at a mature readout
Finance reconciliation, benefit realisation and feasible tranche choices
Avoid — Moving budgets during valid experiments without a method decision
Quarterly or on material change
Strategic assumptions, evidence reuse, ownership cost and retirement
Avoid — Adding new forums without closing old ones
In this release
What the case studies show
Two examples, with clear limits on what they can tell you.
Booking.com linked machine-learning applications to business decisions and tested their effects. Its practitioner paper describes roughly 150 successful applications. It does not show the success rate of all attempts, a pooled profit return, or what a smaller business should expect. The useful lesson is to evaluate what a model changes in a real decision.
Unilever reports lower production time and cost from a digital-twin and AI workflow for product images. That is a company-reported production example, not independent proof of an AI effect, incremental sales or a return you can apply elsewhere. Compare the cost of assets your team can actually use, including review and rework.
The original sources and their limits are recorded in the evidence section. Neither case establishes a general financial return from AI.
Make marginal thinking practical
Those customer counts and values are hypothetical. A real decision needs effect uncertainty, aligned horizons, cohort quality, cash, delivery capacity and alternatives. If incremental acquisitions plausibly include zero, present the customer-effect and profit range rather than a precise CAC ratio. “Hold the budget” is a valid decision. sourcesRT §19AI Augmented GTM Red Team Verdict — Red Team; claim audit, investment challenge and simplification.Evidence for this chapter · 6 claim records
GRW §6 illustrates a CU20,000 tranche that yields 43 incremental customers valued at CU600 contribution before media: CU25,800 less CU20,000 gives CU5,800 before any further costs. The next CU30,000 tranche produces 48 customers: CU28,800 less CU30,000 gives a CU1,200 loss. Historical average CAC can look acceptable while the next tranche destroys value.
Fill T2 for one cohort and reconcile it with Finance.
compare the next feasible tranche with product, service and lifecycle alternatives.
use validated forecasts or response models.