Implementation Guide

Implementation Guide

The AI-Augmented GTM Playbook • Version 1.0 • 9 October 2026

Implement one valuable intervention through the smallest reliable workflow. Add AI only where a bounded comparison can establish its additional value. Dates below are review windows, not delivery or profit guarantees. Outcome maturation, regulation, sample size and staff availability determine the actual pace. [RT §§48,76; OPS §§55–67]

Where to start

Outcome: A material problem with an executable intervention and a credible conventional alternative.

Prerequisites: A commercial owner who can act; relevant source access; Finance and method support proportionate to the decision.

Key actions: Use T1 to diagnose the constraint. Review recent actual decisions and avoidable losses. Compare direct product/service repair, native tools and no change before AI. Select one Play and specify the affected customer, treatment, economic horizon and decision it could improve. Set a maximum learning investment and a review date.

Owner: The existing accountable commercial, Product or Service leader; the sponsor resolves real resource conflict.

Measurement: Decision headroom, materiality, available treatment, relevant data, authority and capacity. These are entry facts, not a promise of ROI.

Decision gate — Relevance: Continue if a valuable action could change. Fix first if the question, economics or authority is unclear. Stop if no material action is available or a cheaper direct repair already settles the choice.

Monday morning: Reconcile last period’s facts, inspect one customer problem, write the alternative and no-change option, name the execution owner and book a decision date. Use existing systems.

First 90 days

Outcome: One operationally credible comparison or controlled intervention, with an honest continue/simplify/stop decision.

Prerequisites: Relevance gate passed; named owner; learning cap; actual staff time; customer and data permissions for the selected work.

Days 0 to 14 select and bound

Complete T1 and identify the minimum T2 economics. Define the customer entity, cost scope and outcome. Reject attractive use cases without intervention headroom, ownership or affordable measurement. Choose one workflow; zero agents are required. Record work that will be deferred or stopped to make capacity available.

Days 15 to 45 establish the comparison

Observe the current process A, repair it into B and establish accepted-work and commercial baselines. Repair only relevant tracking, linkage, permission or workflow gaps. The method owner assesses useful effect size, independent units, variance, interference and maturation. Choose a feasible experiment or explicitly limited alternative.

Complete T3’s protocol and T4’s authority contract. Test any current-state checks, denial paths, receipt/reconciliation and fallback required by the actual action. A read-only assistant does not need an enterprise data migration; a customer-impacting integration cannot skip its relevant controls.

If an unstructured step remains a material bottleneck, specify C as B plus one AI step. Keep human authority unchanged during the first comparison. Record all review, rework and operational costs. Dynamic tool selection needs a reason a fixed workflow is inadequate.

Days 46 to 90 run and decide

Deliver the selected intervention or workflow comparison within scope and limits. Monitor customer and operational guardrails; preserve the comparison. Retain negative evidence and deviations. At the scheduled review, reconcile costs and read out only outcomes that have matured. Decide whether to continue within the cap, fix a named dependency, simplify to B or stop.

Owner: Commercial owner accountable; Operations delivers; Finance owns economic definitions; method owner validates the readout; Product/Service and risk partners supply the required expertise.

Measurement: Accepted-work quality and cost, decision usefulness, actual intervention delivery, customer guardrails, evidence integrity and matured commercial outcomes. Capacity remains separate from cash.

Decision gates — Readiness, Evaluation and Safe Action: Continue only when the selected work has sufficient facts, credible evaluation and appropriate execution controls. Fix first where a specific repair can close the gap. Stop if the promised claim cannot be evaluated economically or exposure cannot be contained.

Day 90 Value review: Incomplete evidence can justify a further capped learning period, with a date and reason; it does not justify broad scale. A 12-month customer outcome cannot be declared proven after 90 days.

Months 3 to 6 prove and standardise

Outcome: A repeatable workflow that earns continued use, or a documented simplification/retirement.

Prerequisites: First comparison delivered correctly; outcomes mature enough for the intended decision or a funded extension has a defined information value.

Key actions: Compare B−A and C−B separately. Determine whether higher accepted quality, useful decisions or removed expense pays for the additional workflow cost. Verify Finance’s benefit realisation action. Document the operating procedure and failure cases. Train through actual work and test absence cover. Retire one replaced report, handoff, licence or process after verifying the replacement. Keep a fixed workflow if it is the best result.

Owner: Business owner and Finance make the value decision; Operations owns standardisation; the functional manager owns real capacity disposition.

Measurement: Mature outcome and uncertainty; accepted-work cost; reviewer load; realised versus forecast benefits; actual adoption; reliability and completed retirement.

Decision gate — Value: Continue if the option beats the best feasible alternative within the agreed risk and economic limits. Fix first for a specific correctable cost or quality problem. Stop the AI addition if B performs as well for less, even when the overall programme beats A. Preserve useful process improvements.

Months 6 to 12 replicate selectively

Outcome: Evidence that a useful workflow transfers to one further population or team without losing its economics or control.

Prerequisites: Value gate passed; explicit transfer assumptions; enough support/review capacity; a funded second need.

Key actions: Run P12. Compare customer, product, price, channel, data and permission differences before reuse. Price the marginal implementation and full shared operating cost. Start a bounded second release. Establish shared metric, evidence or action services only where repeated funded workflows need them. Test real budget mobility before introducing portfolio scenarios. Preserve craft ownership and customer-contact arbitration.

Owner: Original and receiving business owners accept transfer boundaries; Operations supports delivery; Finance prevents double counting; method owner confirms scope.

Measurement: Target-context outcomes, transfer errors, support/review cost, source/evidence reuse, action reliability and retirement of redundant local work.

Decision gate — Repeatability and Structure: Continue if the second context supports the case and ongoing demand is real. Fix first for a bounded transfer gap. Stop or adapt if the lesson does not travel or shared service cost exceeds reuse value. No automatic organisation redesign follows.

Months 12 to 24 formalise only what has earned it

Outcome: A sustainable operating arrangement for proven recurring work, with funded ownership and retirement discipline.

Prerequisites: Repeatability demonstrated; durable backlog; known coordination load; actual funding/delivery rights; positive net case against simpler organisation and technology options.

Key actions: Evaluate whether dedicated roles, a persistent outcome team or a shared platform reduces net operating burden. Use T5 to compare existing functions, temporary missions and persistent structures. Adjust incentives to customer quality, valid evidence and portfolio cooperation. Consider predictive value, uplift or aggregate portfolio models only where their data and measurement prerequisites hold. Price support through vendor/model changes and exit.

Owner: Existing executive sponsor and Finance approve structure and investment; named service owners accept ongoing operations; functional leaders manage roles and skills.

Measurement: Contribution, customer consequences, total ownership cost, real authority, adoption, service demand, review queues and reduced obsolete work. Distinguish structural benefit from the original intervention gain.

Decision gate — Repeatability and Structure, renewed: Continue only where a persistent arrangement beats the lighter alternative. Fix workload or interfaces before adding titles. Stop an uneconomic shared service and retain the useful local workflow.

Advanced state

Outcome: Reliable, economically useful decisions at the required scale, with selective specialist systems and explicit human authority.

Prerequisites: The relevant lower gates remain satisfied; adequate outcomes and treatment variety; funded engineering/method capacity; current controls; positive incremental case for each addition.

Key actions: Add validated prediction or treatment models where they improve decisions. Add durable action adapters when repeated writes require them. Consider bounded dynamic investigation where it beats fixed preparation. Graphs, specialised retrieval, real-time services and multi-agent patterns each need their own query, retrieval, latency or isolation evidence. Maintain evidence expiry, independent evaluation, fallback and retirement.

Owner: Commercial owner retains outcomes; named technical and method owners retain reliability and validity; the authorised risk/budget owners approve action boundaries.

Measurement: Same-action benefit from reduced delay or review, empirical error by risk slice, policy coverage, override reasons, incidents, recovery, full extra cost and net economic value.

Decision gate — Autonomy: Assist or recommend when independent action lacks a case. Retain human-approved execution if review is valuable. Permit only a specific bounded action when its additional value and controls are demonstrated. Keep human-only authority for strategy, material allocation, pricing, novel claims and ambiguous high-impact treatment. Demote or retire when assumptions fail. Advanced does not mean maximum autonomy.

One resource and cost plan

Use T5 with local estimates. These rows are responsibilities, not new hires or universal staffing ratios.

Add staff-days across concurrent commitments; do not allocate the same specialist twice. Convert additional purchases to cash cost and existing staff time to separately labelled opportunity cost. Include review queues, outcome lag, dual running, maintenance, incident handling and exit. If no capacity is available, reduce scope; do not leave a fictional fractional allocation in the plan.

Transformation scorecard

Choose a small subset appropriate to the Play. Every measure needs an owner and an action when it changes. Neither the scorecard nor the roadmap supplies a universal success threshold.

Context adaptations

Small teams combine roles and use current tools; buy scarce methodological advice only where the decision warrants it. Enterprises retain domain ownership and stronger access/operating boundaries without assuming every layer needs a platform. B2B and low-frequency firms allow longer horizons and explicit limited inference. Subscription firms examine retention and service treatment; transactional/marketplace firms include inventory, frequency, unit economics and supply capacity. Regulated or high-impact actions keep stronger qualified human gates, while low-risk drafting can remain lightweight.

Closure and handover

At every stop or scale decision, record what remains useful, what is being retired and who owns customer or system continuity. Preserve necessary evidence and source records, export portable configurations, remove obsolete access, close contracts where authorised and confirm the final benefit classification with Finance. A decision to keep a simple manual or native workflow is a completed implementation outcome.