Your Company Has Been Producing a By-Product for 15 Years. Nobody Put It on the Balance Sheet.

1. The economic asset that accounting may not recognise

Imagine a manufacturer established fifteen years ago.

Every day, its equipment generates operating logs. Technicians record faults and repairs. Managers approve exceptions. Customers report failures, request replacements and describe conditions that formal engineering specifications never anticipated.

These records were created to operate the business, not to build an AI dataset.

Over fifteen years, however, the company may have accumulated thousands of examples of equipment behaviour, human judgement, failed interventions and successful resolutions.

The original operational purpose and the potential secondary economic purpose are different.

An engineer's report written in 2015 helped resolve a specific breakdown. In 2026, a sufficiently documented collection of similar reports might help evaluate whether an AI assistant recognises comparable faults and recommends appropriate escalation.

That possibility illustrates how historical information can acquire additional usefulness when technology, demand or organisational capabilities change.

But the chapter's title requires an accounting qualification.

Economic value is not balance-sheet recognition

The OECD has examined data as an intangible economic input and identified substantial measurement challenges. Its research distinguishes the productive value of data from the limited ability of existing economic and financial statistics to capture every form of data investment ¹.

Under IFRS, IAS 38 governs the recognition of intangible assets within its scope. Recognition requires relevant criteria to be satisfied, and internally generated intangible assets face particular difficulties involving identifiability, expected future benefits and reliable measurement of cost ².

Consequently, a company's possession of historical information does not automatically establish a separately recognisable intangible asset.

Nor does a hypothetical licensing valuation, by itself, justify increasing the carrying amount of an asset or the company's book value. Under IAS 38, initial recognition and subsequent measurement follow prescribed criteria; a revaluation of a recognised intangible asset generally depends on a reliably measurable fair value by reference to an active market, a condition that is rarely met for distinctive enterprise datasets ².

Some identifiable data-related development expenditure may qualify for accounting recognition under applicable standards and circumstances. Other expenditure is recognised as an expense. Acquired intangible assets may also be treated differently from internally generated ones.

The distinction matters because management should not confuse three separate questions:

Accounting: What, if anything, may properly be recognised in the financial statements?

Economics: What incremental benefit could the information generate for the business?

Commercialisation: Could a third party obtain enough value from a permitted use to justify paying for access?

An archive may have meaningful internal economic usefulness without appearing as a separately identified balance-sheet asset.

It may also have no credible licensing market despite significant historical expenditure on collecting and storing it.

The finance analogy is valuable precisely because it imposes discipline: future benefits must be assessed against costs, uncertainty, obligations and alternatives.

2. A possible market is not an investment thesis

Earlier chapters established why certain operational records can support training, retrieval or evaluation. At the board's final decision gate, that technical possibility is an input, not a forecast of demand. OpenAI's dated Data Partnerships announcement documents one company's stated sourcing approach ³; it does not establish an active buyer for this manufacturer's archive, a licensing price or a probability of conversion.

Management should distinguish what it knows from what it proposes to test. Internal records may have observable operational usefulness without a credible external counterparty. Conversely, a prospective buyer's interest does not prove that a provider can lawfully supply the requested package or recover its preparation costs. Rarity, volume and historical depth matter only through a specified task and competing sources. The strategic question is which uncertainty must be resolved before the next pound is committed.

Exhibit 1: From data possibility to capital-allocation evidence

This is a board decision framework, not a prediction of transaction value. A 'yes' in one row cannot compensate for a legal prohibition in another.

The distinction between evidence and speculation is essential to the accounting question. Commercial value may exist without qualifying for recognition under IAS 38, but management should not use an imagined licence price as either an accounting entry or an investment appraisal.

3. Conducting a portfolio inventory rather than launching a sales campaign

The board should treat enterprise information as a portfolio of competing uses, not launch a sales campaign around the largest dataset. A lightweight inventory records its business owner, permitted applications, minimum viable quality evidence, incremental preparation costs and strategic sensitivity. Detailed field-level reconstruction belongs to the dataset assessment chapters; this portfolio decision concerns where to fund the next controlled test.

Consider our hypothetical manufacturer, which identifies three possible datasets.

Exhibit 2: Initial portfolio assessment

This is an illustrative strategic assessment, not a quantitative ranking or assertion of buyer demand.

The three datasets are related, but treating them as one enormous database would obscure important differences.

Machine logs offer breadth and frequency. They may describe changes in equipment performance with great precision, but a fault code does not necessarily explain what ultimately caused a problem.

Technician work orders potentially contain richer operational interpretation. They may connect symptoms with investigations, actions, parts replaced and return-to-service outcomes.

Customer messages offer another perspective, including language used to describe problems. But they may be difficult to link reliably and may include material subject to significant restrictions.

A mature portfolio process evaluates both each dataset independently and the incremental benefit of linking permitted fields across sources.

Why the three assets should not be treated as one investment

Machine telemetry may help internal fault monitoring even where no reliable outcome labels exist. Technician records may provide deeper evidence for a specialist external evaluation but require costly expert review. Customer correspondence may improve an internal retrieval workflow while being unsuitable for external disclosure. Linking assets can expand possible uses, yet also introduces technical costs and fresh permission and confidentiality questions. A portfolio appraisal should show the marginal benefit and marginal cost of each proposed combination before authorising integration. It should not assume that more linked information is automatically more valuable.

A disciplined selection method

Rather than assign one supposedly precise score to every archive, management can use a two-stage process.

First apply non-negotiable eligibility conditions: a plausible task, sufficient access, a defensible route to lawful use and an accountable business owner.

Then compare eligible datasets across task relevance, outcome quality, distinctiveness, historical coverage, technical preparation effort and economic potential.

A numerical scoring system can help organise an extensive inventory, but scores must not conceal disqualifying risks. A dataset that performs strongly on technical usefulness cannot compensate for a lack of lawful permission through additional points elsewhere.

The purpose of selection is to allocate limited investigation resources.

Not every archive deserves a full valuation. Many deserve only enough initial examination to establish that further expenditure is unwarranted.

4. Valuing an opportunity under uncertainty

A company's operational data may have several potential sources of economic benefit.

Internal applications might improve service quality, reduce processing time, lower error rates or support better decisions.

External licensing might generate fees for access to approved datasets, evaluation services or recurring information products.

Strategic benefits could also include improved relationships with technology partners or capabilities developed through a pilot.

These benefits must be evaluated separately to avoid double counting.

An external buyer might find a dataset technically interesting but decline to pay for it. A company might negotiate an attractive licence fee but incur substantial preparation and governance costs.

A technically successful project might also reduce the firm's ability to exploit proprietary information exclusively.

The appropriate financial analysis therefore considers expected incremental benefits, required investment, operating costs, probability of completion, residual risks and the alternatives sacrificed.

A worked business case

Return to the manufacturer.

After preliminary review, the company identifies technician work orders as the most promising candidate for a restricted AI evaluation partnership.

It considers funding a pilot before committing to full preparation.

Assume the following entirely hypothetical conditions:

• Initial inventory, sampling and rights screening cost £10,000, incurred whether or not a licence is signed.
• If the opportunity progresses, full dataset preparation costs an additional £30,000.
• First-year secure delivery, support and governance cost £12,000.
• A prospective first-year licensing fee, if negotiated and completed, is assumed at £70,000.
• Following the initial pilot, management estimates a 40% probability that an acceptable, lawful commercial arrangement will actually proceed.

The assumed fee is not a market benchmark, buyer quotation or valuation of the manufacturer's data. The 40% probability is an illustrative management judgement, not a statistical industry conversion rate.

If the commercial arrangement proceeds, the first-year contribution after the additional preparation and support costs would be:

£70,000 − £30,000 − £12,000 = £28,000.

But management must account for the £10,000 it would spend investigating the opportunity regardless of the outcome.

Exhibit 3: Probability-weighted pilot decision

Illustrative one-year decision model. It assumes the additional £42,000 of preparation and support costs are incurred only when the licence proceeds, and that unsuccessful projects incur only the £10,000 initial pilot cost. It excludes tax, discounting, opportunity costs and unquantified risks.

The example breaks even when the illustrative completion probability exceeds £10,000 ÷ £28,000 = 35.7%, before discounting, opportunity costs, tax and residual risk. At the central 40% assumption, the indicated expected contribution is just £1,200.

That result is not an estimate of the dataset's economic value. It demonstrates that seemingly attractive headline licensing revenue can coexist with a marginal investment case.

A lower fee, higher preparation cost, delayed transaction or more demanding delivery agreement could turn the result negative.

The model deliberately assumes that all later £42,000 of preparation and support cost is incurred only after a completed contract. Viewed before committing funds, if an additional unavoidable £5,000 of preparation must be spent regardless of whether the arrangement closes, expected contribution at the 40% scenario falls to −£3,800 and the break-even probability rises to £15,000 ÷ £28,000 = 53.6%. These are ex-ante funding tests, not observed market conversion rates. Once any such pre-contract expenditure has actually been incurred and cannot be recovered, it is a sunk cost for the next go/no-go decision, which should evaluate only the remaining incremental costs and benefits.

A board paper should set out the cash timing, deposits, milestone payments, probability justification, abandoned-deal costs, recurring support, tax, discount rate, contract liabilities and downside scenarios separately. An expected contribution close to zero should rarely be presented as a precise asset valuation. The stage-gate approach is valuable because it caps how much can be lost before commercial evidence improves.

The internal alternative

The company should also assess whether the same historical records could generate better value internally.

A retrieval assistant might help technicians locate authorised procedures faster. Improved maintenance analytics might identify recurring component failures. More reliable documentation could reduce repeat inspections or training time.

These possibilities need measured operating baselines.

For example, the company could test a tool with a defined group of technicians and compare repair time, first-time fix rates, repeat visits and safety-related errors against an appropriate baseline.

Any expected savings should be reduced for implementation expenditure, ongoing model costs, human supervision and the risk of incorrect recommendations.

The decision should compare competing uses on consistent assumptions rather than assume either internal deployment or external licensing will succeed.

There may also be an option to do both, provided lawful permissions, technical constraints and contractual exclusivity terms allow it.

A good data strategy protects the company's ability to choose among future uses instead of committing its most valuable information to the first interested counterparty.

5. Creating a durable asset rather than a one-time export

A one-time extract may justify a bounded project, whereas an ongoing licence implies an operating capability with staffing, monitoring, correction and security costs. The board should fund that capability only where the buyer's use requires it and the contract allocates responsibility and payment. Retaining additional data without reliable definitions or a proportionate retention policy creates cost and risk rather than automatic strategic value.

Data carries an economic burden

Historical information consumes resources.

Storage and backup systems cost money. Security and access controls require maintenance. Retention and deletion obligations create administrative work. Data may become less interpretable as the systems and employees that created it disappear.

When a company attempts to commercialise the records, further expenditure may be required for extraction, reconciliation, expert review, documentation, rights remediation and secure delivery.

These costs vary substantially by source and use. There is no meaningful universal price-per-record that captures the full economic burden.

Governance also matters to valuation.

The ICO's data-sharing checklist, for example, requires organisations considering personal-data sharing to examine purpose, necessity, lawful basis, responsibilities, security, retention and documentation ⁵.

A commercial opportunity that cannot satisfy applicable requirements is not made viable simply by an attractive fee.

Preserving strategic advantage

Some operational archives contain confidential methods or accumulated know-how that contribute directly to a company's competitiveness.

WIPO explains that commercially valuable confidential information may qualify for trade-secret protection where the relevant conditions are satisfied, including reasonable protective measures ⁶.

Licensing such material can create additional revenue, but sharing also introduces disclosure and control risks.

Management must therefore consider the potential loss of exclusivity, restrictions on future use and whether permitted derivatives could reduce the original competitive advantage.

Sometimes keeping the information internal is the economically stronger decision.

Durability also depends on the ability to reproduce the information consistently.

A buyer receiving a monthly dataset needs to know whether new records remain comparable with earlier deliveries. Changes in source systems, field definitions or business processes should be monitored and documented.

This requires an identifiable owner, a governed extraction process, quality controls and defined responsibilities for updates, corrections and incidents.

NIST's AI Risk Management Framework provides a voluntary foundation for documenting intended use, data-related risks, governance responsibilities and evaluation practices ⁷.

These controls create value even if no external partnership materialises. A well-governed archive can support internal analytics, operational continuity and future technology projects.

The objective should not be to transform every company into a data vendor. It should be to prevent potentially useful information from remaining unmanaged, uninterpretable or unnecessarily exposed.

6. The boardroom decision: what should management do next?

The concluding decision is how much capital and management attention the company is prepared to put at risk, for which specific hypothesis and under whose authority. The board or delegated investment committee should appoint a sponsor, assign legal/privacy and operational risk owners, set an initial expenditure ceiling and approve explicit evidence gates. Funding the next phase should depend on a documented change in what management knows, not merely the passage of time.

For a business with limited technical capacity, the following ninety-day programme is an illustrative planning structure, not a standard timetable or a mandate to commercialise every archive.

Exhibit 4: Illustrative 90-day management programme

The timetable is an illustrative management framework, not an industry delivery benchmark. Complex, regulated or fragmented datasets may require longer investigations.

Each decision gate should have a named approver, an evidence owner, a maximum permitted expenditure and a stop condition. The board should distinguish permission to investigate, permission to share a controlled sample, permission to negotiate and permission to sign; none of those approvals should be inferred from the previous one.

The resulting decision should be proportionate to the evidence available.

A company might discover that its historical records are technically rich but unsuitable for external licensing. It may decide to develop an internal assistant instead.

Another company might identify a narrow, permission-cleared archive whose rare cases support a valuable evaluation task.

A third may conclude that neither internal development nor licensing justifies the required expenditure.

All three outcomes can represent successful asset management if they are supported by a disciplined assessment.

The executive closing checklist

Before authorising a material investment or approaching an external data buyer, management should establish:

• Purpose: A named workflow, intended AI capability and measurable use case.
• Evidence: A defined record unit, representative sample, meaningful context and reliable outcomes.
• Rights: Documented provenance, permissions, confidentiality and applicable privacy requirements.
• Economics: A realistic assessment of preparation expenditure, operating costs, buyer uncertainty and alternatives.
• Delivery: Reproducible extraction, appropriate security, quality checks and clear responsibilities.
• Strategy: An understanding of internal value, competitive implications, exclusivity and future optionality.
• Governance: An accountable owner, review process and explicit conditions for stopping.

These questions create a useful boundary between discovering a possible opportunity and claiming that a commercial asset already exists.

Conclusion: knowing what the company has produced

Fifteen years of operations may leave a company with a valuable record of decisions, failures and outcomes. Yet technical usefulness, contractual rights, an identifiable buyer and accounting recognition remain separate propositions. The correct conclusion to the handbook is not that every enterprise owns a hidden fortune.

It is that management should treat its information resources with the same discipline applied to other investments: establish what the company controls, test the use against an alternative, fund evidence incrementally, protect strategic choices and stop when the economics or permissions fail. Internal deployment, a restricted external pilot, a carefully scoped licence and no investment at all can each be sensible outcomes.

The archive may never appear as a separately recognised balance-sheet asset. The decision whether to commit resources to it still belongs in the boardroom.