The comfortable additive assumption
Organisations often build AI investment cases by addition. A better model adds capability. More data adds accuracy. Training adds adoption. Governance adds control. Each improvement is expected to contribute its own share of value, and weakness in one area appears to be a contained problem rather than a threat to the whole investment.
That way of thinking is attractive because it matches the way technology is usually bought. The product, implementation, training and controls can be listed as separate workstreams. Their projected benefits can then be added together in a business case.
But organisational value does not necessarily arrive in separate parcels. An AI system may produce accurate output and still change nothing because the output enters the wrong workflow, serves no agreed priority, is not trusted by the people expected to act on it, or cannot be used under the organisation’s regulatory, contractual or market conditions.
The technology can work, and the investment can still fail.
The non-linear penalty of misalignment
A 2026 paper by Panagiotis Theodorou and Petros Theodorou applies the Strategic Alignment Framework to AI-enabled procurement in energy companies. Strategic Alignment is the organising theory: capability must fit the organisation’s objectives, structure and readiness, while remaining responsive to the external environment, if it is to create value. 1
The authors use a multiplicative model to formalise what happens when those conditions do not align. The multiplication sign captures a non-linear penalty: a low factor does not simply subtract a fixed amount; it suppresses the combined result. The strongest tool in the market cannot independently overcome a strategically irrelevant objective, a missing decision process or an organisation that cannot act on its output.
Theoretical foundation and authors’ contribution: The authors trace their version of the Strategic Alignment Framework to research undertaken by Petros Theodorou at the University of Macedonia in the 1990s, initially in the context of restructuring industrial production. The 2026 paper applies that line of work to AI-enabled energy procurement through the formal model, six application groups and a five-stage roadmap. Andrew’s Intel adds: the additive-versus-multiplicative explanation, the “fund the limiting condition” interpretation, the illustrative vignette, the cross-sector examples and the practical decision check. Evidence boundary: the procurement model is conceptual and model-driven. The general Strategic Alignment logic predates this application, but the wider examples and practical adaptation in this article should not be treated as universal results established by the paper.
Per = α × AI × S × O × E Within the Strategic Alignment Framework, procurement performance equals translation efficiency multiplied by AI capability, strategic alignment, organisational readiness and environmental conditions.
Why the scale of the loss is easy to miss
Suppose AI capability, strategic alignment and external conditions are all strong, but organisational readiness is weak. In the paper’s conceptual scoring system, lifting an already strong tool from 0.9 to 1.0 changes the combined result only slightly. Raising organisational readiness from 0.3 to 0.7 more than doubles it.
On a narrow screen, each row is shown as a separate card.
| Change made | Illustrative configuration | Combined result* |
|---|---|---|
| Improve the tool only | 1.0 × 0.9 × 0.3 × 0.8 | 0.216 |
| Repair readiness | 0.9 × 0.9 × 0.7 × 0.8 | 0.454 |
*Illustrative arithmetic only; α omitted. These values are not measurements from the paper or a validated assessment instrument.
The numbers are not the point. The direction of attention is. Once the tool is capable enough, another model upgrade may produce less value than clearer ownership, cleaner data, a revised workflow, better training or a decision rule that turns output into action.
Find and fund the limiting condition. The easiest component to buy is not necessarily the component constraining the result.
The four conditions that decide how much value survives
The Strategic Alignment Framework is the overarching lens; the factor labelled S in the equation is the narrower measure of whether the particular AI initiative supports the organisation’s agreed objectives. The model can then be understood through four practical questions. A separate scaling factor, α, represents how efficiently useful output becomes a changed decision or behaviour.
1. AI capability
Can the system reliably perform the specific task with the data and operating conditions available? Capability includes more than an impressive demonstration; it includes repeatable performance in use.
2. Strategic alignment
Which declared priority, decision, or risk does the initiative support? In a commercial organisation this might include cost, resilience or return on invested capital; elsewhere it may be service quality, compliance or another priority set in the organisation’s plan. A technically advanced use case can still be strategically weak if it is detached from the outcomes that matter.
3. Organisational readiness
Can data, roles, processes, governance, skills and decision rights turn output into action? A licence does not become an organisational capability merely because it has been purchased.
4. External conditions
Do regulations, contracts, market conditions and stakeholder expectations permit or constrain the intended use? These conditions may either amplify the benefit or prevent it from being realised.
A technically successful pilot can still be organisationally inert
Consider a large infrastructure operator introducing AI contract analytics. The system can identify renewal dates, unusual clauses and possible compliance gaps, and reassess contract value as market conditions change. Technically, the pilot works.
But the contract repository is inconsistent, legal and procurement teams use different taxonomies, no one owns the alert queue, and high-risk findings arrive by email rather than entering a review process. The organisation has purchased capability without creating a decision system.
When definitions are agreed, exception categories are prioritised, ownership is assigned, and alerts enter the contract-review workflow, the same technology begins to create value. The model has not become dramatically more intelligent; the organisation has become more aligned.
The vignette is fictional, but the mechanism follows the paper’s logic. It explains why an AI pilot may be technically successful and economically disappointing at the same time.
Why energy procurement makes the lesson visible
Energy procurement is a demanding setting in which to examine alignment. Buyers operate across volatile commodity markets, capital-intensive assets, long supply chains, sustainability obligations and strict regulatory regimes. A procurement decision may affect cost, resilience, compliance and the physical reliability of essential infrastructure.
The paper groups AI use into six applications: automated sourcing and RFx, robotic process automation, supply-chain optimisation, contract-management analytics, supplier-risk management, and price optimisation and negotiation support. The list matters because the applications do not require the same organisational arrangements.
Automating invoices and purchase orders is largely an operational workflow problem. Supplier-risk management is cross-functional: risk, finance, operations and procurement must be able to coordinate around the signal. Contract analytics becomes strategic only when contractual priorities, specialist oversight and decision authority are connected to the output.
“AI-enabled”, therefore, describes a capability. It does not describe an outcome. The same capability can be useful, inert or disruptive depending on where it is placed and how the surrounding organisation responds.
Sequence matters more than enthusiasm
The authors turn the framework into a five-stage implementation roadmap. Its value lies less in the novelty of the individual stages than in their order: assessment and alignment precede scaling; skills, roles and governance are part of implementation rather than repairs added after installation.
Figure 2. The five-stage implementation sequence derived by Theodorou and Theodorou. Andrew’s Intel visualisation and condensed wording based on the source paper, CC BY 4.0.
The strategy-and-roadmap stage should respond to the external environment and to the gaps identified during assessment. A bounded pilot should then test more than whether the model can produce an answer: it should test whether the workflow, data, ownership and decision path work together. Scaling should follow evidence that the whole system can create a repeatable outcome.
Governance is not the last administrative hurdle. It is the continuing process through which models, roles, risks and strategic priorities remain aligned as circumstances change.
The practical judgement
The paper’s most useful contribution is not the promise that leaders can reduce AI strategy to an equation. It is the Strategic Alignment perspective behind the equation: value depends on the fit between capability, objectives, organisational readiness and environment. The multiplication makes the penalty for misalignment harder to ignore.
That question is useful before an investment, when a pilot disappoints, and when a technically successful system fails to change results. In each case, buying a better tool may be premature. The limiting condition may be strategic ambiguity, data, process design, governance, skills, decision authority or the external environment.
As a management lens, “AI value is multiplicative” is memorable because it challenges a comfortable assumption. Poor implementation does not merely reduce one line of benefit. Misalignment can suppress the return on the entire investment.
The value comes from the system, not the tool. The next unit of investment should go to the condition that limits the system.
PRACTITIONER LAYER / USE AFTER READING
Find the condition suppressing value
This check translates the paper’s model into a leadership discussion tool. It is an Andrew’s Intel artefact, not a validated diagnostic. Its purpose is to stop a team from averaging away the factor most likely to limit the result.
On a narrow screen, each row is shown as a separate card.
| Dimension | Question to answer | Evidence to look for |
|---|---|---|
| AI capability Score: ___ / 5 | Can the system reliably perform the specific task with the available data? | Test results, failure modes, data quality and deployment conditions. |
| Strategic alignment Score: ___ / 5 | Which formally agreed priority, decision or risk does the initiative support? | Named outcome, owner, strategic-plan or business-plan link, trade-offs and success measure. |
| Organisational readiness Score: ___ / 5 | Can people, processes and governance turn output into action? | Workflow integration, roles, skills, decision rights and exception handling. |
| External conditions Score: ___ / 5 | Do regulation, contracts, market conditions and stakeholder expectations permit or constrain use? | Legal and regulatory review, data rights, volatility and procurement conditions. |
| Translation efficiency (α) Score: ___ / 5 | How efficiently does useful output become a changed decision or behaviour? | Time to action, adoption, escalation path, accountability and feedback loop. |
1. Score each dimension independently. Do not average first. 2. Begin with the lowest score. A low factor is the most likely place to direct the next unit of attention or investment. 3. Match action to the constraint. Scores of 1–2 suggest foundation work; a score of 3 suggests a bounded pilot and repair; scores of 4–5 support cautious scaling subject to evidence and continuing governance.
Questions to ask before approving or scaling an AI investment
What decision, risk or operational outcome is this initiative intended to change?
Which condition is currently the weakest, and what evidence supports that judgement?
Where will the AI output enter an existing workflow, and who is authorised to act on it?
What must be true about data quality, process consistency and user behaviour for the result to be repeatable?
What evidence distinguishes a working organisational capability from a persuasive technical demonstration?
Which regulatory, contractual, market or stakeholder conditions could suppress the expected value?
What would make us stop, redesign or defer the initiative rather than scaling it automatically?
Reference panel: the six applications in the authors’ framework
The paper distinguishes operational and strategic applications. The categories are retained here as a reference layer rather than as part of the main reading path.
On a narrow screen, each row is shown as a separate card.
| Primarily operational | Primarily strategic |
|---|---|
| Automated sourcing / RFx Drafting and running sourcing events, standardising evaluation and improving speed or transparency. | Contract-management analytics Monitoring obligations, performance, compliance and potential leakage. |
| Robotic process automation Routine invoice and purchase-order work, fewer transaction errors and release of capacity. | Supplier-risk management Financial, operational and sustainability risk, continuity and disruption signals. |
| Supply-chain optimisation Forecasting, inventory, logistics and service-level decisions in volatile conditions. | Price optimisation and negotiation support Market intelligence, price analysis and preparation for negotiation decisions. |
The descriptions are condensed into decision-maker language. The application groups and operational/strategic distinction are drawn from Theodorou and Theodorou (2026).
Evidence and limits
The paper is theory-building and model-driven. Its procurement specification and scenarios are illustrative, and the authors call for empirical validation using firm-level procurement data, longitudinal case studies and cross-company comparison. The variables and application weights are not yet calibrated for reliable investment forecasting. 1
The model should therefore be used as a structured lens, not as a settled universal law. The general Strategic Alignment logic has wider roots, but the additive-versus-multiplicative explanation, cross-sector examples and decision tool in this article are editorial adaptations rather than universal results established by the procurement paper. The defensible claim is that the framework provides a coherent way to examine how capability, strategy, organisational readiness and external conditions interact — and to ask better questions before scaling an AI investment.
Sources and evidence notes
Evidence posture: This is a theory-building, model-driven paper. The equation and scenarios are useful as a structured decision lens, not as a calibrated forecasting instrument or a universal empirical law. The source, DOI metadata and licence were rechecked on 29 October 2026.
- Panagiotis Theodorou and Petros Theodorou, ‘Artificial intelligence for strategic procurement in energy firms: a modeling framework’, Academia Green Energy, volume 3, issue 1, published 16 March 2026. The paper presents a conceptual and mathematical framework, six application groups and a phased roadmap; it calls for empirical validation and firm-level calibration.
- Creative Commons Attribution 4.0 International. The source permits sharing and adaptation with appropriate credit, a licence link and an indication of changes. Andrew’s Intel has condensed, visualised and editorially interpreted the framework; no author or publisher endorsement is implied.
Author-review note: Panagiotis and Petros Theodorou reviewed an earlier draft and commented on theoretical framing, provenance and attribution. Responsibility for the final editorial interpretation remains with Andrew’s Intel.
This article provides general decision support. It is not legal, regulatory, investment, procurement or technical advice.