The strategic signal
As foundation models become easier to access, the cost of building an AI-enabled feature can fall. That expands the number of competitors able to enter a market and makes it harder to defend a product based only on a prompt interface or a thin layer over a third-party model.
But the conclusion is not that every AI wrapper is doomed. A product can create durable value through distribution, trusted data, deep integration, workflow expertise, reliability, compliance and customer relationships. The key question is whether the company solves a recurring problem better than alternatives and can retain a meaningful share of the value it creates.
Where defensibility can come from
- Workflow ownership: The product becomes part of a critical business process, not an optional chat window.
- Distribution: The company can reach customers efficiently and build trusted relationships.
- Proprietary context: Permissioned data, feedback loops and domain-specific knowledge improve results in ways competitors cannot easily reproduce.
- Reliability: Evaluation, monitoring, security and support reduce the operational risk of using the product.
- Integration: The product connects to systems of record and handles exceptions, approvals and auditability.
- Economics: The company can deliver measurable outcomes with sustainable gross margins after inference and support costs.
The proxy-layer problem
A product that merely forwards a prompt to a model provider can be vulnerable to price changes, new model features and imitation. Its position improves when it solves the harder surrounding work: data access, workflow design, permissions, evaluation, integrations, customer-specific context and change management.
That does not mean every company should train its own model. Owning model weights can be expensive and may not create differentiation if customers value the complete workflow more than the underlying model. Founders should invest in the layer where they can build a meaningful advantage.
Questions for founders and investors
- What customer problem remains valuable if the model provider changes?
- What evidence shows repeat usage, retention and willingness to pay?
- Can competitors reproduce the outcome with a few prompts, or would they need data, integration and operational capability?
- How do gross margins change as usage grows and models become more capable?
- What distribution advantage brings customers back and reduces acquisition costs?
FutureIsNow’s related analysis explores AI’s effect on SaaS economics, capital intensity in AI infrastructure and turning AI pilots into business value.
Bottom line
AI makes it easier to build a feature and more important to build a business. Durable startups will combine model capability with workflow ownership, customer access, reliable execution and economics that improve as they scale.
