AI and Strategic Positioning: Goodwill vs Expense
AI Value Assessment — Lesson 7 of 10
A company spends $15 million developing an AI-powered underwriting engine that reduces credit losses by 30% and processes applications 10 times faster than the manual alternative. The engine has a useful life of at least five years. Its replacement cost is estimated at $20 million. When the company is acquired two years later, the acquirer's purchase price allocation values the AI engine at $25 million as an identifiable intangible asset.
Yet on the company's own balance sheet, the AI engine may appear as nothing — or, at best, as a partially capitalised development cost. The $15 million investment was expensed through the income statement over two years, reducing reported profits and suppressing the balance sheet. The company's book value understates its true economic value by the full amount of the AI asset.
This is the accounting paradox at the heart of AI investment. The standards that govern financial reporting — primarily IAS 38 for individual entities and IFRS 3 for acquisitions — create systematically different treatments for the same asset depending on whether it was built internally or acquired. Understanding these rules is essential for CFOs making capitalisation decisions, investors evaluating AI-intensive businesses, and executives positioning their companies for M&A.
AI investments exist in an accounting grey zone. Most are expensed as incurred, which depresses reported profits and understates balance sheet value. Some qualify for capitalisation under IAS 38's development criteria. And in M&A, the same assets that were invisible on the seller's balance sheet are separately identified and valued on the acquirer's. CFOs who understand these rules can make strategic decisions about capitalisation, disclosure, and investor communication that materially affect how the market perceives their AI investment.
The IAS 38 Framework for AI
IAS 38 distinguishes between the research phase and the development phase of an internally generated intangible asset. Only development costs can be capitalised, and only when six conditions are simultaneously met.
The Six Capitalisation Criteria
| Criterion | IAS 38 Requirement | AI Application |
|---|---|---|
| Technical feasibility | Demonstrate the asset can be completed for use or sale | Model achieves target accuracy on validation data |
| Intention to complete | Entity intends to complete, use, or sell the asset | Board-approved project with committed resources |
| Ability to use or sell | Entity can use the asset or sell it | Clear deployment plan or licensing strategy |
| Probable future economic benefits | Asset will generate future revenue or reduce costs | Business case with quantified benefits |
| Adequate resources | Technical, financial, and other resources available to complete | Team staffed, infrastructure provisioned, budget allocated |
| Reliable cost measurement | Expenditure during development can be reliably measured | Time tracking, cost allocation methodology in place |
Research vs Development in AI Projects
The critical judgement for AI projects is where the research phase ends and the development phase begins. IAS 38 defines research as "original and planned investigation undertaken with the prospect of gaining new scientific or technical knowledge." Development is "the application of research findings to a plan or design for the production of new or substantially improved materials, devices, products, processes, systems, or services."
In AI terms, the research phase typically includes: data exploration, feature discovery, algorithm selection, initial model architecture experiments, and proof-of-concept work. The development phase typically begins when: the model architecture is finalised, the training pipeline is established, the model achieves target performance metrics on validation data, and the integration design is specified.
A fintech company's AI underwriting project had four phases: (1) data exploration and feature engineering (4 months, $1.2M), (2) model architecture selection and experimentation (3 months, $0.8M), (3) model training, validation, and optimisation (5 months, $1.5M), (4) production integration and deployment (4 months, $1.0M). Phases 1-2 were classified as research and expensed. Phases 3-4 met the IAS 38 capitalisation criteria and were capitalised as a $2.5 million intangible asset, amortised over 5 years.
The boundary between research and development in AI is genuinely ambiguous, and auditors increasingly scrutinise AI capitalisation decisions. Organisations that capitalise aggressively — classifying early-stage experimentation as development — risk restatement and audit qualifications. The conservative approach is to capitalise only from the point where a clear, measurable development plan exists with demonstrated technical feasibility.
What AI Assets Look Like in M&A
When a company with significant AI capabilities is acquired, IFRS 3 requires the acquirer to identify and separately value all identifiable intangible assets at their fair value. This is where AI assets that were invisible on the seller's balance sheet suddenly materialise — and the treatment has direct implications for the acquirer's reported earnings.
Purchase Price Allocation for AI
In a typical purchase price allocation (PPA) for an AI-intensive company, the identifiable AI assets might include:
Identifiable AI Assets
- Proprietary algorithms and trained models
- Curated training datasets
- Software and MLOps infrastructure
- Patents and trade secrets
- Customer relationships enhanced by AI
Residual Goodwill
- Assembled AI workforce
- Organisational AI capability
- Expected synergies
- Going-concern value
- Unidentifiable competitive advantages
The allocation between identifiable intangible assets and goodwill has significant financial reporting consequences. Identifiable intangible assets are amortised over their useful lives, which reduces reported earnings for the duration of the amortisation period. Goodwill is not amortised under IFRS (it is tested annually for impairment). An acquirer that allocates more value to goodwill and less to identifiable AI assets will report higher earnings in the years following the acquisition — but the asset remains on the balance sheet indefinitely and is subject to potentially large impairment charges.
Valuation Methods for AI Assets in PPA
The three standard methods for valuing intangible assets in a PPA each have specific applications to AI.
| Method | AI Application | When to Use |
|---|---|---|
| Relief from Royalty | What royalty rate would a licensee pay for the AI technology? | When comparable AI licensing transactions exist |
| Multi-Period Excess Earnings | What excess earnings does the AI asset generate above contributory asset charges? | For AI assets that are the primary value driver |
| Replacement Cost | What would it cost to recreate the AI asset from scratch? | For data assets and infrastructure where cost data is available |
Strategic Implications for CFOs
The accounting treatment of AI creates several strategic decisions that directly affect how the market values the company.
The Capitalisation Decision
Capitalising AI development costs increases reported assets and reduces reported expenses (improving EBITDA). Expensing them reduces reported profits but provides an immediate tax deduction and avoids the complexity of impairment testing. The decision is not purely accounting — it signals to investors how the company views its AI investment: as a long-term asset or a period cost.
The Investor Communication Challenge
A company that expenses all AI development will report lower profits but higher cash flow quality (no capitalised costs to amortise). A company that capitalises aggressively will report higher profits but may face investor scepticism about asset quality. The solution is transparent supplementary disclosure: explain the AI investment, the assets it creates, and the expected returns — regardless of how the costs are treated on the income statement. The Opagio Valuator provides the framework for this supplementary disclosure.
Pre-Acquisition Positioning
Companies anticipating a sale or fundraise should document their AI assets comprehensively. A well-prepared intangible asset register — documenting trained models, curated datasets, algorithmic IP, and technical infrastructure — makes the due diligence process smoother and increases the likelihood that AI assets are separately valued (rather than subsumed into goodwill) during the PPA.
The difference between a well-documented AI asset portfolio and an undocumented one can be worth millions in acquisition valuation. Acquirers assign higher values to assets they can identify, understand, and independently verify. An AI model with documented training data, performance metrics, and maintenance history is worth more in a PPA than an equivalent model with no documentation — even if the technical capability is identical.
The FRS 102 Consideration for UK SMEs
UK companies reporting under FRS 102 (rather than full IFRS) have a simpler framework but similar principles. FRS 102 Section 18 allows capitalisation of development costs when criteria similar to IAS 38 are met, though the detail and guidance is less extensive. The key practical difference is that FRS 102 offers less specific guidance on AI-related capitalisation, giving CFOs more judgement — and auditors more latitude for challenge.
For UK SMEs investing in AI, the practical recommendation is to follow IAS 38 principles as best practice even when reporting under FRS 102. The IAS 38 framework provides a more robust, defensible basis for capitalisation decisions and prepares the company for a potential transition to full IFRS if it grows or is acquired by a listed group.
What Comes Next
In Lesson 8: Building Your AI ROI Dashboard, we move from theory to practice, designing the KPIs, metrics, and reporting structures that bring the 4-Layer Framework to life as an operational management tool. The dashboard integrates cost reduction metrics, revenue attribution, competitive advantage indicators, and strategic option valuations into a single view that boards and investors can use.
Ivan Gowan is CEO of Opagio, the growth platform that helps businesses and investors measure, manage, and grow intangible assets. Before founding Opagio, Ivan held senior technology and leadership roles across financial services and digital platforms for 25 years. Meet the team.