Gartner Predicts Open-Weight AI Models Will Run 90% of Enterprise AI Workflows by 2030

Open-Weight AI Models Set to Dominate Enterprise Workflows

Gartner has issued a striking forecast for enterprise AI: open-weight AI models will execute 90% of enterprise operational agent workflow steps by 20301

1

2

2

. This projection, outlined in a document titled "Open-Weight Models Will Run 90% of Enterprise Agentic Workflow Steps" by Gartner senior director Mark Petty and group vice president Raymond Paquet, signals a fundamental shift in how organizations will deploy AI infrastructure. The analysis reveals that nine in 10 agent workflow steps are routine tasks that open-weight AI models can handle at performance parity with closed models, but at dramatically lower costs1

1

2

2

.

The Economics Behind the Shift

The cost of frontier models has become a critical consideration for enterprises deploying agentic work at scale. According to Petty and Paquet, an open-weight AI model costs approximately one-quarter as much as the cheapest closed model achieving the same performance score1

1

2

2

. When compared against using a frontier LLM for every workflow step, the cost advantage expands to as low as one-twentieth1

1

. These figures represent best-case scenarios. Current open models tend to generate more verbose outputs and require more retries, which reduces the practical saving per task to approximately five or six times compared to frontier models1

1

2

2

. Still, AI model economics heavily favor open-weight alternatives for repetitive enterprise operations.

Routine Tasks Drive the Case for Open Models

Typical enterprise operational agent workflow steps involve hundreds of routine executions that don't require advanced reasoning capabilities. These include data classification, tool calling, passing information between systems, data validation, and reconciling data across platforms1

1

2

2

. Open-weight AI models excel at handling hundreds of these routine steps for every single step that demands frontier reasoning1

1

. This capability matters because enterprises running agentic work at scale execute these operations millions of times monthly, making efficiency and cost control paramount.

Workflow Control Trumps Cost Savings

While the economic advantages are substantial, Gartner positions workflow control as the primary driver for adopting open-weight AI models. "Years of workflow history are a large part of what an acquirer pays for in a SaaS company and every routine step sent to a frontier API moves a slice of it outside your perimeter," Petty and Paquet wrote1

1

2

2

. Open-weight models preserve organizational control over proprietary data and operational intelligence. The analysts describe cost reduction as "a bonus" rather than the main objective1

1

2

2

. This perspective reflects growing enterprise concerns about data sovereignty and competitive intelligence leakage through external AI APIs.

Domain-Specific Models and the Role of Harnesses

Gartner predicts enterprises will primarily adopt domain-specific open-weight AI models, particularly in scenarios where a small number of highly stable steps run millions of times each month1

1

2

2

. Most models are already commodities, according to the analysts. The real value now lies in harnesses—the tools that run agents and manage agent fleets1

1

. Effective harnesses route work to the most appropriate model for each task without users noticing any performance impact, save money, impose strong guardrails, and manage complex agent deployments1

1

. Less capable harnesses leave enterprises with significantly more manual oversight.

The Future Division of Labor

Petty and Paquet outline a clear division of labor emerging in enterprise AI by 2030. Frontier models will continue serving as the engine for research and development, functioning essentially as capital expense investments1

1

2

2

. Open-weight AI models become operational expenses after frontier models complete the heavy lifting of initial development and complex reasoning tasks1

1

2

2

. This bifurcation suggests that for-profit model makers spending billions on AI infrastructure will find themselves competing for revenue from just 10% of steps involved in agentic workflows1

1

. Watch for enterprises to begin piloting domain-specific open-weight deployments for high-volume routine tasks while reserving frontier LLM access for strategic innovation work.