Your AI Moat Is Probably Not the Model
A healthcare AI company reportedly spent about $20 million over three years building a proprietary system.
Then a newer general-purpose model arrived.
Suddenly, core elements of what the company had built could be replicated by competitors at a fraction of the cost. The technology had not stopped working. It had stopped being a durable moat. Building a Competitive Advantag…
That story captures one of the most important strategic questions emerging around AI:
If increasingly capable intelligence can be bought, rented, copied, or replaced, where does durable advantage move?
The answer is unlikely to be “the model” for most companies.
As AI capability becomes more widely available, competitive advantage may increasingly relocate toward what is harder to reproduce:
- Proprietary context
- Workflow knowledge
- Decision history
- Trust
- Relationships
- Institutional learning
- Value capture
That changes the AI strategy question. Instead of asking: Which model should we own? leadership may need to ask: What do we need to own that actually makes us different?
AI Capability Is Not the Same as Strategy
The HBR article Building a Competitive Advantage in the Healthcare AI Market makes this distinction directly. It argues that having the most advanced AI technology is not enough to win. The more important challenge is making a coherent set of choices about where AI contributes to long-term competitive advantage. Building a Competitive Advantag…
The article adapts Roger Martin’s strategic-choice framework into five interconnected questions:
- Strategic goals — Why are we investing in AI?
- Scope — Where will we compete?
- Moat — What advantage can persist?
- Capabilities — What should we own versus outsource?
- Management systems — What organization and governance fit those choices?
Importantly, the author argues that these choices should be revisited iteratively rather than sequentially. Changing one may require adjusting the others to preserve strategic coherence. Building a Competitive Advantag…
That matters. AI strategy should not be treated as a one-time technology plan. It is an operating hypothesis that gets tested, observed, and revised. Small changes in scope, authority, economics, workflow, or evidence can change what the right decision looks like.
Start With the Competitive Question
One of the strongest distinctions in the article is between an AI investment that improves efficiency and one that creates strategic advantage. Those are not the same thing.
An AI system may reduce cost, speed up work, or automate a process and still do nothing to improve the organization’s competitive position.
The article gives several examples:
- A hospital using AI to improve clinician retention may create strategic advantage because talent is scarce and workforce quality matters competitively.
- A biotech company using AI to shorten time to market may create strategic advantage because speed directly affects lifetime economics.
- Using AI to reduce the cost of a process that competitors can reduce just as easily may simply become table stakes. Building a Competitive Advantag…
That gives leadership a better first question: Is this AI investment merely improving performance—or changing our competitive position? Both may be worth doing, but they should not be confused.
Maximum Capability Is Not Always Maximum Value
The article also offers a useful example of deliberate constraint. Viz.ai could have pursued a stronger diagnostic claim for one of its products. Instead, it limited the system to identifying patients likely to have a particular type of stroke and alerting specialists. The narrower role still delivered the core value—faster treatment—while reducing regulatory and clinician resistance. Building a Competitive Advantag…
This suggests a broader AI design principle:
How much does the AI need to do to create the desired value at an acceptable level of risk?
That is different from: How much can the AI do?
The technically impressive solution is not necessarily the strategically superior solution. Sometimes greater autonomy adds little incremental value while increasing regulatory, operational, reputational, or decision risk. The best system may therefore be deliberately less autonomous. Capability should serve the business decision; the business decision should not exist to justify the capability.
The Moat May Be Somewhere Else
The most interesting section of the HBR article is its discussion of where defensibility may actually come from. The author identifies several potential moats:
- Proprietary data and network effects
- Regulatory approval and reimbursement
- Distribution and trust
- Workflow integration
- Customer co-development
- Proprietary technology
- People-led AI capability Building a Competitive Advantag…
Notice how few of these require owning the best foundation model.
The article suggests that OpenEvidence’s advantage may come less from its underlying model and more from its distribution among clinicians and the trust it has accumulated. Building a Competitive Advantag…
It also argues that workflow integration can become defensible when an AI system accumulates workflow-specific context and becomes deeply embedded in day-to-day activity. Building a Competitive Advantag…
And customer co-development creates another kind of advantage: a competitor may be able to copy the features, but it cannot instantly copy years of accumulated fit, trust, domain knowledge, and relationships. Building a Competitive Advantag…
This is where the strategic implications become larger than healthcare. As generalized intelligence becomes easier to access, scarcity can move toward things that remain organization-specific. The model may be shared; the organization’s accumulated context is not.
Own Your Moat. Outsource the Rest.
The article summarizes the capability decision simply:
Own your moat; outsource the rest. Building a Competitive Advantag…
That principle deserves more attention. Organizations can spend enormous amounts of capital rebuilding capability that becomes commoditized months later. The HBR article warns that companies often discover too late that what they developed internally has either been absorbed into general-purpose models or can now be replicated much more cheaply. Building a Competitive Advantag…
That suggests a more disciplined capital question: Are we spending proprietary capital on proprietary advantage?
For many organizations, that may mean spending less on rebuilding general intelligence and more on:
- Differentiated context
- Decision logic
- Workflow integration
- Trust
- Learning
- Value capture
Commodity intelligence can increasingly be rented. Organizational advantage has to be accumulated.
Value Creation Is Not Value Capture
This is another distinction AI strategy often skips. The usual sequence is:
Capability → Adoption
Can the system perform the task? Will people use it? Those questions matter, but they are incomplete.
A more useful strategic sequence is:
Capability → Value Proposition → Value Capture → Economics → Competitive Advantage
An AI system can create value for a customer, employee, or workflow without creating durable economic value for the organization that paid for it. The HBR article explicitly warns that even an excellent AI product can fail if the underlying business model does not support sustainable economics. Building a Competitive Advantag…
That leads to one of the simplest—and most neglected—AI strategy questions: If AI creates value, who captures it?
Does the value become:
- Margin?
- Revenue?
- Market share?
- Pricing power?
- Customer retention?
- Faster time to market?
- Better talent retention?
- Lower risk?
- Stronger switching costs?
Or does the gain simply get competed away?
AI activity is not the same as value creation. Value creation is not the same as value capture. And value capture is not automatically durable advantage.
Strategy Needs a Decision System
This is where the HBR strategy article connects directly with IBM’s 2026 CFO research. IBM argues that intelligence is becoming less scarce and that decision design is increasingly the practical constraint. Defining execution in the AI era
That matters because competitive strategy is not just a set of ideas. It requires repeated decisions:
- What should we fund?
- What should we build?
- What should we buy?
- What should we constrain?
- What deserves more authority?
- What should we scale?
- What should we stop?
IBM recommends beginning with a small set of high-value enterprise decisions and defining the data and ownership those decisions require. It describes the combination of high-value decisions, aligned data, and ownership as a decision spine. Defining execution in the AI era
An AI-Forward Competitive Operating Model
Taken together, the research suggests an operating sequence like this:
- Strategy: What are we trying to win at? Define the competitive outcome before choosing the technology.
- Decisions: Which choices determine whether we succeed? Identify the few decisions that matter most.
- Evidence: What would prove the advantage is real? Define what deserves confidence before scaling.
- Authority: How much autonomy does this use case actually require? Give the system enough authority to create value—but no more than justified.
- Capabilities: What must we own, and what can we rent? Invest proprietary capital where proprietary advantage actually exists.
- Operating Clarity: Do people understand how the strategy translates into work? The strategy must become executable.
- Action: Put the decision into operation.
- Outcome / Value Capture: Where did the value actually land? Measure the business consequence, not merely the activity.
- Learning / Reallocation: What should change next? Scale what works. Redesign what does not. Move capital when evidence changes. Stop what no longer creates value.
Competitive Advantage Has to Learn
IBM’s work adds an important final piece. It argues that capital allocation should become more dynamic in the AI era, with resources moving as evidence changes rather than waiting for annual planning cycles. Defining execution in the AI era
That means strategy itself becomes more adaptive. A company may make a good AI investment today and discover six months later that:
- The capability has commoditized
- The economics changed
- The workflow created more value than expected
- A competitor copied the feature
- Customer behavior shifted
- Governance became the bottleneck
- Another opportunity now deserves the capital
That is why iterative strategy matters. The objective is not to make the perfect AI bet once. It is to build an organization capable of updating its bets as the evidence changes.
The Moat May Be the Organization
As AI capability becomes broadly available, the durable advantage may increasingly come from the system surrounding the intelligence. The organization that knows:
- What matters
- Which decisions create advantage
- What evidence deserves confidence
- What capability should remain proprietary
- Where authority should be bounded
- Where the value actually lands
- What outcomes teach
- Where capital should move next
may outperform an organization with better technology but weaker decision discipline.
That is the deeper strategic shift. The AI model may be replaceable. The organization’s accumulated context, workflow knowledge, trust, decision history, relationships, and learning may not be.
So the most important AI strategy question may no longer be: Which model gives us the strongest capability?
It may be: What are we building around intelligence that competitors cannot easily copy—and does our decision system know how to keep improving it?