Many association leaders feel a quiet pressure to pick a side. In the early days of any technological shift, we look for the winner—the one platform or provider that will define the era. We ask whether we are an OpenAI house or an Anthropic house. We wonder if we should wait for Google to integrate everything into our existing workspace or if we should bet on a specialized startup. This instinct is natural. It is how we managed the transition to cloud computing, social media, and even association management systems. We look for a partner we can trust, sign a long-term contract, and build our workflows around their specific architecture.
But artificial intelligence is not following the traditional software-as-a-service trajectory. We are witnessing a rapid shift where the core product—intelligence—is becoming a commodity. For association leadership, the most dangerous move right now is to marry your organization to a single AI model or vendor. When the underlying technology is changing every few weeks and costs are plummeting, loyalty to a specific model is not a strategy. It is a constraint. To thrive, associations must adopt a model-agnostic approach that treats AI as a utility rather than a proprietary feature.
For a long time, the prevailing belief was that a few massive U.S. companies would hold a monopoly on high-level intelligence. The assumption was that the sheer cost of computing power and the scarcity of data would keep the "frontier" models—the smartest ones—behind a high wall. However, the recent surge in open-weight models is proving this wrong. When we look at the global landscape, labs in China and elsewhere are releasing models that rival the best in the world at a fraction of the cost.
Consider the release of Kimi K3 from Moonshot AI. This model arrived with capabilities that match or exceed some of the most respected public models from just a few months ago, yet it costs about a third of the price to run. This is not an isolated event. Alibaba recently announced a massive model, Qwen 3.8 Max, claiming it rivals the top-tier performance of the most advanced Western models. Whether these specific models become your primary tools is less important than what they represent: the democratization of high-level reasoning.
When intelligence can be packaged into "open weights" that anyone can download and run on their own hardware, the competitive advantage of the model developer starts to evaporate. If a lab in Beijing or a startup in Paris can produce a model that performs as well as the one you are currently paying a premium for, the value of that intelligence is no longer in its scarcity. It is in its availability. For an association, this means the "engine" inside your AI tools is increasingly interchangeable. If you build your entire member engagement strategy around the specific quirks of one model, you lose the ability to swap it out for a faster, cheaper, or more specialized version later this year.
To understand AI commoditization, think about how you buy gasoline. You likely have a preferred gas station near your home, perhaps because of its location or a loyalty program. But if you were driving and saw two stations across the street from each other—one charging four dollars a gallon and the other charging one dollar for the exact same grade of fuel—you would switch without a second thought. You have no deep emotional loyalty to the brand of the fuel because the product is a commodity. It performs the same function in your engine regardless of the logo on the pump.
Artificial intelligence is moving toward this same reality. Microsoft, one of the biggest investors in OpenAI, is already demonstrating this pragmatic approach. Reports indicate that Microsoft believes it can save hundreds of millions of dollars a year by switching some of its internal workloads to different models, including those from Chinese labs. They are looking for the best tool for the specific job at the lowest possible price. They are not letting their investment in one company dictate their technical choices for every single task.
Association leadership should adopt this same mindset. Your AI strategy should focus on the "workloads"—the specific problems you are trying to solve for your members—rather than the specific model. Are you trying to automate member support? Are you trying to summarize decades of technical journals? Are you trying to personalize event recommendations? Each of these tasks requires a different level of intelligence. By remaining model-agnostic, you can route the simple tasks to cheap, fast models and save the expensive, high-reasoning models for the truly complex problems. This is how you manage a utility.
One of the most common mistakes in AI adoption is using "too much" intelligence for a task. We often default to the most famous, most powerful model available because we want the best results. But this is like hiring Albert Einstein to answer basic member renewal questions. Einstein is certainly capable of the task, but his time is expensive, and his genius is wasted on routine data entry.
Most association tasks do not require frontier-level intelligence. A member services professional needs empathy, clear communication, and access to your database. They do not need to be able to invent new physics. We are reaching a point where models that are "smart enough" for 90% of association work are becoming incredibly cheap or even free to run. These models, often referred to as being at the level of the previous year's flagship, are more than capable of handling agentic workloads—where the AI runs in a loop to make decisions, check data, and execute tasks.
When you stop viewing AI as a single, all-powerful brain and start seeing it as a spectrum of available intelligence, your strategic options expand. You can afford to run AI-powered processes at a scale that was previously unthinkable. If the cost of processing a member's entire history to provide a custom career path drops from five dollars to five cents, the project moves from a "maybe someday" to a "must-do now." This shift only happens when you stop overpaying for intelligence you don't actually need.
Perhaps the most significant shift for association leadership is the need to plan for a future where high-level intelligence is effectively free. This sounds like hyperbole, but the trend lines are clear. Every few months, models get smarter, faster, and cheaper. If you are evaluating a new member benefit today and the "inference cost" (the cost of the AI thinking) makes the project too expensive, do not kill the idea. Instead, ask: "What would we do if this intelligence cost nothing?"
Many organizations are currently bounded by the constraints of today's pricing. They build small, cautious pilots because they are afraid of a massive API bill. But if you build your infrastructure to be swappable, you can design ambitious projects today that will become financially viable by the time they are ready to launch. If you build a system that is married to a high-cost provider, you are stuck with their price increases or their technical limitations. If you build a model-agnostic system, you can ride the wave of falling costs.
This approach allows for what we call "ambitious member value." Imagine a world where every member has a dedicated AI assistant that has read every article your association has ever published, attended every webinar, and knows the member's specific professional goals. This assistant could provide real-time coaching, suggest connections, and solve technical problems. Today, the cost of running that for 50,000 members might be prohibitive. Next year, it might be a line item in your budget that is smaller than your coffee service. The associations that win will be the ones that have already built the data structures and member interfaces to take advantage of that cheap intelligence the moment it arrives.
The goal of an AI strategy is not to have the best AI. The goal is to serve your members better than anyone else. The AI is just the engine. By treating intelligence as a commodity, you shift your focus from the technology to the outcome. You stop worrying about which lab is winning the race this week and start focusing on how you can use the resulting abundance of intelligence to solve the problems your problems face.
Build for swappability. Invest in your data, because your data is the only thing the AI labs don't have. Ensure your technical team is using frameworks that allow them to change models with a few lines of code. Most importantly, encourage your board and your staff to think bigger. The constraints of the past—the high cost of human-like reasoning and data processing—are disappearing. When intelligence is a commodity, the only limit is your organization's imagination and its willingness to experiment.