OpenAI Head of Startups Mark Manara said at FutureLab that AI start-ups should build products that become stronger as foundation AI models improve, because cheaper inference and more capable models are changing how young companies compete. He said the edge for AI start-ups is moving away from sophisticated prompting and engineering workarounds toward speed, execution and distribution.

Manara said AI product development has changed over the past two years. Start-ups previously spent significant effort building scaffolding around models to compensate for limits, but models have become more reliable at reasoning, following instructions and using tools autonomously. Founders are now more often trusting models to make decisions and execute tasks, while focusing on giving systems the right tools and evaluation systems.

Manara warned that companies built around temporary model limitations may quickly lose their advantage. Echoing OpenAI CEO Sam Altman's advice to founders, he said, "If you're upset that the model got smarter, that's probably a less defensible and harder business to build." He argued that durable AI companies are those whose products improve when underlying models become more capable.

Manara said smaller, lower-cost models are opening applications that were previously too expensive to deploy at scale. Lower inference costs allow start-ups to support larger free tiers, run more agentic workflows (AI systems acting toward goals) and test use cases that would have been commercially difficult just months ago, without sharply raising compute costs.

Manara said future differentiation will depend on distribution, deep customer understanding, product quality and execution. Production-grade AI systems still need evaluation frameworks, guardrails, latency optimisation, memory systems and rigorous testing, beyond an impressive demo.

Manara said AI is changing start-up staffing, citing companies with fewer than 10 employees generating tens of millions of dollars in revenue by using AI across engineering, hiring, customer support and internal operations. He said engineers are taking on product duties, product managers are contributing code, and specialised AI engineering skills in model evaluation, orchestration and deployment are gaining value alongside Forward Deployment Engineers who implement AI systems for enterprises.

OpenAI Head of Startups Mark Manara said at FutureLab that AI start-ups should build products that become stronger as foundation AI models improve, because cheaper inference and more capable models are changing how young companies compete. He said the edge for AI start-ups is moving away from sophisticated prompting and engineering workarounds toward speed, execution and distribution.

Manara said AI product development has changed over the past two years. Start-ups previously spent significant effort building scaffolding around models to compensate for limits, but models have become more reliable at reasoning, following instructions and using tools autonomously. Founders are now more often trusting models to make decisions and execute tasks, while focusing on giving systems the right tools and evaluation systems.

Manara warned that companies built around temporary model limitations may quickly lose their advantage. Echoing OpenAI CEO Sam Altman's advice to founders, he said, "If you're upset that the model got smarter, that's probably a less defensible and harder business to build." He argued that durable AI companies are those whose products improve when underlying models become more capable.

Manara said smaller, lower-cost models are opening applications that were previously too expensive to deploy at scale. Lower inference costs allow start-ups to support larger free tiers, run more agentic workflows (AI systems acting toward goals) and test use cases that would have been commercially difficult just months ago, without sharply raising compute costs.

Manara said future differentiation will depend on distribution, deep customer understanding, product quality and execution. Production-grade AI systems still need evaluation frameworks, guardrails, latency optimisation, memory systems and rigorous testing, beyond an impressive demo.

Manara said AI is changing start-up staffing, citing companies with fewer than 10 employees generating tens of millions of dollars in revenue by using AI across engineering, hiring, customer support and internal operations. He said engineers are taking on product duties, product managers are contributing code, and specialised AI engineering skills in model evaluation, orchestration and deployment are gaining value alongside Forward Deployment Engineers who implement AI systems for enterprises.