Strategic Value Realisation: Artificial Intelligence Costs Management and Value Attribution

Observations

Digital transformation means organisations must keep adapting, and AI offers many ways to create value and achieve goals. However, managing AI’s full strategic value is difficult due to complex financial and operational issues. Without a solid way to measure and prove this value over time, transformation efforts may be scattered and lack clear accountability.

This paper introduces SVR’s TAONexus (the TBM-AI-OKR Nexus)², a specialised framework designed to identify all costs and the real value of AI, solving the main financial governance problem in AI projects. TAONexus provides a clear financial structure for handling AI investments.

Understanding the AI Cost Conundrum: Opacity and Duality

AI costs can be primarily categorised into two distinct financial components. These components have distinct characteristics, and a nuanced understanding of them is essential to effectively manage AI expenditure.

• Training Costs (Capital Expenditure): High, one-off costs for research and development, such as buying specialised hardware (like GPUs), gathering data, and paying data scientists. These costs support the organisational objective of innovation.
• Inference Costs (Operational Expenditure): Ongoing costs for running AI models to produce results. These scale variably depending on how much you use the model (e. g., tokens or API calls) and support the organisational objective of operational efficiency.

AI introduces a new, multifaceted cost concept that defies the simple, linear pay-per-use models prevalent in traditional Cloud Software-as-a-Service (SaaS). AI costs are fragmented, driven by pricing metrics such as per-token, per-query, per-hour, per-GPU, and per-GB, making visibility a significant challenge.

Organisations also need to consider AI total cost of ownership (TCO), which includes often-overlooked shadow costs that sit outside direct Cloud bills. These indirect costs are important for working out the true return on investment (ROI), and include:

• Labour for data management.
• Cyber security.
• Legal review.
• Change management.
• Employee training.
• The cost of underutilised tooling or capacity.
• Database fees and related consumption costs.
• Ongoing monitoring of model creep and refinement of AI calls.
• Implementation and monitoring of AI guardrails and compliance.
• Data ingress and egress charges.
• Governance costs.
• Cloud infrastructure fees.
• Increasing consumption costs as agentic AI services grow through integration complexity³.

The Limitations of Standalone Frameworks

Neither FinOps nor TBM, when applied in isolation, provides a complete solution for managing this AI cost duality and complexity.

FinOps: This is a cultural practice that excels at providing real-time visibility, accountability, and continuous optimisation for variable Cloud spend. Its collaborative, ground-up approach is crucial for managing the highly volatile inference costs and operational efficiencies of AI. Key FinOps tactics, such as rightsizing GPU instances and leveraging committed-use discounts for interruptible training jobs, are essential for driving cost efficiency in AI pipelines.

However, traditional FinOps is inherently OpEx-centric. It struggles to account for the substantial CapEx required for model training (e. g., on-premises hardware) and fails to provide a comprehensive TCO analysis, omitting critical indirect costs like change management and security. This blind spot renders a purely FinOps approach insufficient for strategic investment planning, where the total financial commitment must be amortised against long-term value.

Technology Business Management (TBM): This is a strategic framework that provides a standardised taxonomy for categorising and reporting on all technology costs across Cloud, on-premise, labour, and software to provide a holistic TCO perspective. TBM offers the essential top-down context for mapping technology expenditure to business services, products, and units. TBM explicitly enhances AI modelling, including an AI models sub-tower in the technology resource towers and support for tracking GPU-based AI compute, with AI solution offerings represented in the solutions layer⁴.

Nonetheless, TBM is traditionally designed around a slower, periodic reporting cadence (monthly or quarterly), which is insufficient for managing the dynamic, real-time consumption patterns of Cloud-based AI. Furthermore, TBM faces challenges in allocating sunk costs associated with the R&D nature of AI, including failed experiments and discarded models that do not yield a deterministic final product.

Introducing the TAONexus Framework

The key is to combine the best parts of FinOps and TBM together with OKRs into one clear approach: TAONexus (the TBM-AI-OKR Nexus). TAONexus is designed to help organisations get the most value from AI by making all costs visible and giving better control over spending that furthers the organisation’s goals. This model builds on the main ideas of the FinOps foundation and fills in the AI cost visibility gaps, so you can see and manage every part of AI spending for an impact that matters.

The components of TAONexus are:

1. TBM as the Strategic Backbone: This establishes the standardised language and TCO model, ensuring all direct and indirect costs are captured, including labour and specialised AI infrastructure. TBM provides the context for prioritising FinOps optimisations based on business value.
2. FinOps as the Operational Engine: This manages the ground-up, granular, real-time consumption of variable Cloud resources (inference costs). It provides the necessary operational agility for day-to-day cost control and optimisation.
3. OKRs for Strategic Alignment: These provide the mechanism for aligning both FinOps and TBM activities directly to aspirational, measurable business goals. OKRs translate high-level business objectives (organisational value drivers like cost, risk, innovation, experience) into quantifiable, easily measurable outcomes.

Value Attribution through Rigorous Measurement

The framework’s power lies in moving beyond simple cost allocation to sophisticated value attribution, thereby linking AI spend directly to measurable Key Results (KRs).

TAONexus helps organisations see all their AI costs and gives them better control over spending. It uses TBM as the foundation to capture all costs and FinOps for real-time management and optimisation. OKRs (objectives and key results) link cost management directly to business goals, enabling teams to work towards clear, measurable outcomes.

TBM helps create useful financial measures (such as cost per AI interaction) by aggregating all AI solution costs and linking them to the value they deliver. TBM facilitates the creation of relevant AI unit economics (e. g., cost per AI-assisted interaction) by aggregating the fully burdened TCO of an AI solution and mapping it to the value stream⁵ it supports.

These measures support outcome-focused key results, such as reducing the time to solve problems while staying within budget.

The TAONexus hybrid approach brings together the fast, flexible actions of FinOps with the thorough, big-picture planning of TBM. It is designed to work towards clear, measurable business goals (OKRs).

Next Steps

Success hinges on instilling a culture of shared financial accountability across engineering, finance, and business domains, ensuring that every dollar spent on AI is financially traceable and contributes verifiable, measurable value to defined strategic outcomes. Specifically, leaders must mandate the shift from tracking aggregated spend to quantifying value through AI-specific unit economics, such as cost per prediction or cost per query, providing the critical basis for ROI calculation and performance-based contracting⁶.

• Integrate with Ideation and Innovation: Ensure the TAONexus framework aligns with existing ideation and innovation processes and culture.
• Mandate TBM by Design and Scenario Modelling: Shift governance to embed TCO, capacity forecasting, and ROI modelling directly into AI architecture and funding decisions before proofs of concept (PoCs) are approved and investment commitments are made. Leverage TBM Taxonomy 5.0’s capabilities to run scenario planning, evaluating financial trade-offs (e. g., Cloud vs. on-premise inferencing, or build vs. buy decisions).
• Formalise TAONexus: Mandate the use of value streams as the required bridge between AI solutions and oganisational objectives (OKRs). Ensure all AI initiatives define clear, measurable OKRs focused on business outcomes rather than just technical achievements.
• Establish Granular Accountability via Tagging: Implement a robust, unified tagging and cost allocation strategy for all AI-related Cloud resources, aligning metadata with the TBM taxonomy (e. g., project name, business owner, workload type). This is critical for visibility and accurate cost attribution.
• Prioritise Unit Economics: Define measurable key results derived from AI TCO, focusing on cost-for-performance KPI’s such as cost per prediction or cost per model run, rather than solely tracking total spend.
• Implement Performance-Based Contracting: Review and update sourcing and supply plans to adopt performance-based contracting frameworks for AI services, linking a substantial portion of supplier fees directly to the achievement of relevant, measurable buyer outcomes.
• Embed Governance: Create a TAONexus operating rhythm with a regular value council reviewing FinOps telemetry, TBM allocation variances, and OKR progress, with clear ownership across TBM Office⁷, FinOps, machine learning operations (MLOps), and product finance.
• Integrate FinOps Practices: Establish a continuous feedback loop and a regular cadence for reviews in which FinOps, TBM, and AI teams discuss spending and progress toward OKRs. Empower engineering teams with the necessary tools and visibility to take ownership of their AI usage and optimise resources continuously (e. g., rightsizing GPU instances).
• Invest in Objective Attribution Capability: Begin planning to integrate machine learning-driven attribution tools to objectively measure each AI solution’s unique impact on business outcomes. Enforce collaboration between the TBM Office and MLOps teams to tag consumption data at the source for accurate value allocation.

Footnotes

1. TAONexus combines real-time cost control (FinOps) with Technology Business Management’s strategic planning (TBM), all tied to your objectives and key results (OKRs) to keep AI spending efficient day-to-day and aligned to your top goals, so that results are clear and measurable.
2. FinOps is © The FinOps Foundation, 2025; Technology Business Management and TBM is © The TBM Council, 2025; The OKR framework was created by Andrew Grove at Intel, 1983 and later formalised and popularised by John Doerr, 2018; Strategic Value Realisation, SVR, TAONexus, and TBM-AI-OKR Nexus are © Barta Global Services, 2014–2025.
3. As agentic AI connects to more apps and Clouds, integration complexity drives up usage-based costs. Put simply, more connections mean more compute, API calls, and data transfers across multi-Cloud and app-to-app links.
4. ‘TBM for AI’, TBM Council, 2025, ‘Technology Business Management (TBM) Taxonomy’, TBM Council, 2025.
5. The Value Stream framework and its associated metrics provide the operational context for achieving goals set by OKRs. These value stream metrics, functioning as key performance indicators (KPIs), ensure that performance is measured directly against the end-to-end flow of value to the customer.
6. See also: ‘AI Product Operating Cost Drivers: Demystifying the Path to Strategic Value Realisation – Special Report’, IBRS, 2025.
7. The TBM Office is a dedicated team that implements the TBM framework to align technology investments with business goals.