AI and Competitive Advantage: Optionality and Risk
AI Value Assessment — Lesson 6 of 10
Cost reduction saves money. Revenue growth makes money. Competitive advantage protects money — and, over time, is often worth more than the first two layers combined.
When Warren Buffett describes a business with a "wide moat," he is describing intangible assets that protect the business from competition: brand loyalty, network effects, proprietary technology, regulatory barriers, switching costs. AI creates a new generation of these moats — ones that are often invisible to traditional competitive analysis but increasingly decisive in determining which companies dominate their markets.
This lesson examines the four mechanisms through which AI builds competitive advantage, introduces frameworks for assessing moat strength, and addresses the risks that can erode AI-driven advantages.
AI competitive advantages are compounding and self-reinforcing — more data improves models, which attracts more users, which generates more data. But they are not permanent. Organisations must actively maintain and extend their AI moats, because the same compounding dynamics that build advantages can also build advantages for competitors. The key question is not "do we have an AI advantage?" but "is our advantage widening or narrowing?"
The Four AI Moat Mechanisms
AI builds competitive advantage through four mechanisms, each with different durability and defensibility characteristics.
Mechanism 1: Proprietary Data Moats
Data is the raw material of AI. An organisation that has accumulated years of proprietary, high-quality, domain-specific data has an advantage that competitors cannot easily replicate — regardless of how much they spend on AI talent or compute infrastructure.
The strength of a data moat depends on three factors:
| Factor | Strong Moat | Weak Moat |
|---|---|---|
| Uniqueness | Data cannot be obtained elsewhere (proprietary sensors, exclusive partnerships, user-generated) | Data can be purchased from third-party providers |
| Volume | Years of accumulated data with millions of data points | Months of data that a new entrant could match quickly |
| Network effects | More users generate more data, which improves the product, which attracts more users | Data volume does not improve the product proportionally |
A B2B SaaS company in the logistics sector had been collecting shipment data — routes, timings, fuel consumption, delays, weather impacts — from 3,400 customers for seven years. When it deployed an AI-driven route optimisation feature, the model trained on this proprietary dataset outperformed commercial alternatives by 23% on fuel efficiency and 31% on delivery time accuracy. A competitor launching a similar feature would need to accumulate years of equivalent data before matching performance — a structural advantage worth significantly more than the development cost of the AI itself.
Mechanism 2: Model Quality Advantages
A model trained on more data, better data, and refined through more real-world feedback loops will outperform a competitor's model trained on less. This quality gap translates directly into customer outcomes — better recommendations, more accurate predictions, fewer errors — which reinforces market position.
Model quality advantages compound through a learning loop: better models attract more users, more users generate more training data, more data produces better models. The speed of this loop determines how quickly the quality gap widens.
Mechanism 3: Network Effects
AI-powered platforms can create network effects that are qualitatively different from traditional network effects. In a traditional marketplace, the network effect is direct: more sellers attract more buyers. In an AI-powered platform, the network effect operates through data: more users generate more behavioural data, which improves the AI, which improves the experience for everyone.
This data-mediated network effect is particularly powerful because it is often invisible to users. They simply notice that the product keeps getting better — the recommendations become more relevant, the search results more accurate, the predictions more reliable. What they do not see is that their own usage is part of the feedback loop that drives the improvement.
Mechanism 4: Switching Costs
As customers integrate AI-powered features into their workflows, switching costs accumulate. These costs are both technical (data migration, integration rebuilding, workflow reconfiguration) and cognitive (learning new interfaces, retraining staff, losing historical insights).
AI-specific switching costs include:
- Trained model dependency: The AI has learned the customer's patterns, preferences, and anomalies. Starting fresh with a competitor means losing that learned context.
- Historical insights: Three years of AI-generated analytics, predictions, and recommendations have become part of the customer's decision-making infrastructure.
- Integration depth: The AI connects to the customer's CRM, ERP, data warehouse, and other systems. Replicating these integrations is expensive and risky.
The Compounding Effect
These four mechanisms do not operate in isolation. They compound. Proprietary data creates model quality advantages, which improve the product, which drives user growth, which generates more data and increases switching costs. The strongest AI moats exhibit all four mechanisms simultaneously. When assessing an AI competitive advantage, evaluate each mechanism independently and then assess the compounding effect of their interaction.
Assessing Moat Strength
Quantifying competitive advantage is inherently more difficult than measuring cost savings or revenue growth. Three frameworks provide structured approaches.
The With-and-Without Method
The With-and-Without method asks: what would the business be worth with this AI capability versus without it? The difference is the value of the competitive advantage. This valuation approach is widely used in intangible asset valuation and translates naturally to AI moat assessment.
In practice, the "without" scenario is estimated by modelling the business with industry-average capabilities instead of the AI-enhanced capabilities. The revenue, margin, and growth rate differences between the two scenarios, discounted over the expected moat duration, provide a Net Present Value of the competitive advantage.
The Replication Cost Method
How much would it cost a well-funded competitor to replicate the AI capability from scratch? This includes the cost of acquiring equivalent data (if possible), hiring equivalent talent, building equivalent infrastructure, and waiting for equivalent learning cycles. The replication cost provides a floor valuation for the moat — the advantage is worth at least what it would cost to reproduce.
The Market Share Sustainability Analysis
How does the AI advantage translate into sustainable market share? Analyse customer retention rates for AI-enhanced versus non-AI-enhanced product features. Track competitive win rates in deals where the AI capability is a differentiating factor. Monitor the rate at which competitors are closing the AI gap.
Moat assessment is inherently forward-looking and uncertain. The goal is not to calculate a precise valuation but to establish a credible range that informs strategic decisions. A competitive advantage valued at $20-40 million justifies continued investment even if the exact figure is debatable. See Lesson 9 for case studies applying these frameworks in practice.
The Risks to AI Competitive Advantage
AI moats are powerful but not permanent. Three categories of risk can erode them.
AI Moat Risk Assessment
| Risk Category | Description | Mitigation |
|---|---|---|
| Technology disruption | A breakthrough (e.g., foundation models) makes proprietary models obsolete | Invest in capability to adopt new architectures, not just current models |
| Data commoditisation | Previously proprietary data becomes publicly available or purchasable | Continuously generate new proprietary data through customer interactions |
| Talent attrition | Key AI talent leaves, taking institutional knowledge and capability | Document models and processes; build team depth beyond key individuals |
| Regulatory change | Data protection or AI regulation limits the use of proprietary data | Design data practices for the most restrictive plausible regulatory regime |
The most dangerous risk is complacency. An organisation with a strong AI moat today may assume the advantage is self-sustaining. It is not. Every AI moat requires active maintenance: continued data investment, model retraining, infrastructure modernisation, and talent development. The moment investment in moat maintenance slows, competitors begin to close the gap.
From Competitive Advantage to Enterprise Value
For investors and boards, the critical question is how AI competitive advantage translates into enterprise value. The connection operates through three channels:
- Revenue protection: AI moats reduce customer churn and protect recurring revenue streams, increasing the reliability of cash flow projections.
- Growth premium: Organisations with AI moats command higher revenue growth multiples because the compounding mechanisms create self-reinforcing growth.
- Acquisition premium: In M&A transactions, AI capabilities and the associated data assets command significant premiums in purchase price allocation.
The Opagio Valuator provides tools for assessing the enterprise value implications of AI competitive advantages, mapping them to the intangible asset categories used in formal valuations.
What Comes Next
In Lesson 7: AI and Strategic Positioning, we examine the accounting treatment of AI investments — when to capitalise under IAS 38, when to expense, and how AI assets appear (or fail to appear) in M&A transactions where goodwill allocations determine reported earnings for years.
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.