Chapter Seven

A Practical Playbook for Becoming AI-Ready

The path from cloud-first to AI-ready is not a single transformation program. It is a set of reinforcing management disciplines across infrastructure, data, cost, risk and talent.

The following playbook is designed for senior executives who need a board-level view and practical next steps:

Pillar 1: Workload Placement by Design

Create an AI workload placement framework that covers public cloud, private cloud, edge, SaaS, and on-premises options
Evaluate latency, data gravity, sovereignty, egress cost, resilience, energy constraints, model proximity, and reversibility
Revisit placement regularly because inference patterns, model architectures, and commercial terms are evolving quickly

Power availability and grid connection should be treated as first-order placement constraints, not secondary sustainability considerations.

The IEA reports that data-center electricity consumption is set to double by 2030, while power use from AI-focused data centers is poised to triple; its earlier Energy and AI analysis also projected that AI-optimized data-center electricity demand could more than quadruple by 2030.

That means workload placement decisions must account for regional power availability, cooling, grid connection lead times, energy contracting and resilience, alongside latency, sovereignty and cost. [9][10]

Pillar 2: Data Products as First-Class Assets

Prioritize high-value data products with clear owners, lineage, quality rules, sensitivity classification, and AI-use permissions
Invest in semantic layers, metadata and knowledge structures that make enterprise context discoverable and governable
Treat unstructured data as a major AI risk surface, not an afterthought

Pillar 3: FinOps for AI

Extend cost management from cloud spend to AI spend across SaaS, APIs, models, data, infrastructure, and agent platforms
Define unit economics for priority use cases before scaling
Implement tagging and allocation for agents, workflows, and environments.

Pillar 4: Agent Governance and Observability

Maintain an agent registry with identity, owner, purpose, model, tools, data access, permissions, and budget
Require evaluation pipelines for accuracy, safety, compliance, and drift
Log decisions, actions, exceptions, and human interventions end-to-end

_{^{With the introduction of tokenization costing, driving up the cost of AI, it's imperative to broaden the control of agent development / AI usage costs, but without stifling innovation and creativity. Architectural layers will need to be implemented to maximize prompt, agent, sub-agent and skill reuse and assist with prompting minification to reduce cost.}}

Steve Crombie, CIO Bulgaria, ING Bank

_{^{Data quality has become a board-level issue because AI faithfully scales whatever information it is given. Well-governed data accelerates value. Poorly governed data accelerates mistakes.}}

Sumita Dutta, Global IT and Digital Head & CIO, Shell Downstream, Customer Operations, Shell

Pillar 5: Talent Redeployment and Learning

Map roles to tasks to understand what AI will automate, augment, or create
Reskill operations staff toward engineering, product ownership, data quality, risk analysis, and agent supervision
Create career paths for agent managers, AI product owners, model risk stewards, and platform economists

_{^{In regulated industries, the justification of a particular decision must be proven and demonstrated that it is compliant and corrective action taken when a process is deficient or inaccurate. Ownership of decisions will ultimately have an accountable individual with its agent coworker(s) contributing to, but not owning it.}}

Steve Crombie, CIO Bulgaria, ING Bank

_{^{Your role is not at risk if you embrace AI…it’s more at risk if you don’t! AI removes the slow traditional manual data collation, compilation, and analysis; however, there is still a human in the loop required, and that skill set has moved from data gatherers to real-time decision makers. The future Internship will be centered around learning the business sector and the consequences of those decisions.}}

Steve Crombie, CIO Bulgaria, ING Bank

_{^{The next generation of operational dashboards will track digital workers alongside human workers, measuring utilization, quality, costs, risk exposure, and business value. The most valuable employee in the AI era may not be the person who knows the technology best. It may be the person who knows where human judgement still matters most.}}

Sumita Dutta, Global IT and Digital Head & CIO, Shell Downstream, Customer Operations, Shell

How the Pillars Move the Enterprise Up the Maturity Curve

Playbook Pillar	Stage it Unlocks
Pillar 1: Workload placement, power and sustainability	Stage 2 → 3: Makes the estate cloud-effective and AI-enabled, with placement, economics and capacity under control
Pillar 2: Data products as first-class assets	Stage 3: The data foundation without which AI experimentation cannot reach production safely
Pillar 3: FinOps for AI	Stage 3 → 4: Makes consumption visible and allocable so autonomous workloads can scale without runaway cost
Pillar 4: Agent governance and observability	Stage 4: The identities, budgets, telemetry and kill switches that make agent-ready operation auditable
Pillar 5: Talent redeployment and learning	Stage 4 → 5: Builds the human capacity to orchestrate humans, agents and platforms as one adaptive portfolio

The five pillars are not a parallel list to the maturity curve; they are the mechanism that advances the enterprise along with it. Each pillar does the work that unlocks the next stage.