Implement through evidence gates
What happens in 90 days and what earns the next tranche?
The roadmap is a sequence of decisions, not a technology purchase schedule. The Implementation Guide contains the full phase contracts and resource template. Each gate returns Continue, Fix first or Stop. sourcesOPS §§55–67AI Augmented GTM Operating Model and Transformation Plan — Operating model; derived design, not independent organisational evidence.RT §§48,76,81AI Augmented GTM Red Team Verdict — Red Team; claim audit, investment challenge and simplification.
| Named gate | Evidence required before continuing |
|---|---|
| Relevance | Material addressable problem, owner, available intervention and conventional alternative |
| Readiness | Relevant economics, data, permission, capacity and known limits |
| Evaluation | Feasible comparison, outcome horizon, cost baseline and predeclared decision rule |
| Safe Action | Correct authority, current checks, records, stop and recovery for the actual action |
| Value | Mature or appropriately qualified outcome, full cost and advantage over the best feasible alternative |
| Repeatability and Structure | Transferability, recurring demand, funded ownership and retirement of replaced work |
| Autonomy | Same-action benefit from removing review, reliable controls and acceptable residual exposure |
Text version
Each gate returns Continue, Fix first or Stop.
- Relevance: Material addressable problem, owner, available intervention and conventional alternative.
- Readiness: Relevant economics, data, permission, capacity and known limits.
- Evaluation: Feasible comparison, outcome horizon, cost baseline and predeclared decision rule.
- Safe Action: Correct authority, current checks, records, stop and recovery for the actual action.
- Value: Mature or appropriately qualified outcome, full cost and advantage over the best feasible alternative.
- Repeatability and Structure: Transferability, recurring demand, funded ownership and retirement of replaced work.
- Autonomy: Same-action benefit from removing review, reliable controls and acceptable residual exposure.
One selective starting plan
ModelM09ModelDays 0–14: choose one problem and reject weak alternatives. Days 15–45: establish economics and a baseline, repair relevant foundations and agree the measurement and action contract. Days 46–90: run one feasible comparison or controlled release. Optional AI adds one bounded comparison, not five parallel pilots.
At day 90, continue within the learning cap, simplify or stop. An outcome that takes a year cannot be proved in 90 days. Months 3–6 may be needed to mature evidence and standardise a useful workflow. Months 6–12 can test transfer to another population and earn reusable services. Months 12–24 can justify structural or specialist investment where workload and economics support it. None of those dates automatically releases funding or action authority.
Track commercial contribution, customer consequences, accepted-work cost, evidence quality, reviewer load, reliability, adoption and retired work. Licences, prompts, training attendance and agents launched are activity measures, not transformation outcomes.
Dates are review windows, not guaranteed results. Use this model with the conditions and limits described above.
Ten things to stop doing
- Equating platform ROAS with enterprise profit.
- Treating attribution or a fluent AI explanation as causal evidence.
- Scaling cheap acquisition without downstream cohort quality.
- Booking unused saved hours as cash.
- Funding AI against a deliberately unrepaired baseline.
- Using agents for exact rules and arithmetic.
- Starting more pilots than reviewers and operators can support.
- Building graph, streaming or Control Plane infrastructure without a decision-specific case.
- Launching experiments without useful sample, a horizon and a decision rule.
- Keeping duplicate reports, integrations and meetings after their replacement works.
Ten things to establish first
- One material customer or commercial constraint.
- An available intervention and a named decision owner.
- One reconciled economic metric and horizon.
- A competent conventional alternative.
- Relevant outcome linkage and customer safeguards.
- A feasible measurement plan with uncertainty visible.
- Exact authority and a stop procedure.
- A staff and review-capacity plan.
- A short decision and evidence record.
- A dated value review with a retirement option.
These are requirements for a selected workflow, not ten enterprise workstreams. sourcesRT §§65–76AI Augmented GTM Red Team Verdict — Red Team; claim audit, investment challenge and simplification.
Related