When a company goes bankrupt, its physical assets are typically the first things auctioned off to pay creditors. When budget carrier Spirit Airlines shut down operations in May under the weight of billions in debt, the expected items went on the block. Planes, airport gates, and real estate were all liquidated. But an August 14 court filing revealed a completely different kind of asset up for sale. The airline auctioned off its internal enterprise data and software code.
Google won the bid for $10 million.
The tech company purchased decades of finance and accounting records, operations data, pricing and revenue management models, internal emails, Teams chats, and HR records. The deal explicitly excluded customer passenger profiles and frequent flyer records, and a third party was contracted to scrub the remaining files of any personally identifiable information. Google stated it would use this operational data to help improve its products and AI models.
Ten million dollars is pocket lint to a company like Google. The purchase signals a strong hypothesis that years of logistical and operational data contain patterns useful for training models in specific domains.
The lesson for membership organizations is direct. You sit on a treasure trove of specialized information. Your association data monetization potential is likely far higher than a defunct airline's operational logs. The question is how you choose to leverage it.
To understand why Google bought Spirit's data, you have to look at how artificial intelligence is currently advancing. The frontier models from companies like Google, Anthropic, and OpenAI are already incredibly capable. They have been trained on essentially everything available on the public internet.
To make these models smarter, AI labs have two primary paths. The first is algorithmic improvement. A recent example of this was when OpenAI released the model internally codenamed Strawberry, which launched as o1. This was the first true reasoning model. Previous generations of AI would essentially give you the very first thought that came to mind when answering a prompt. A reasoning model has the ability to edit its thoughts and think about a problem for a period of time before returning an answer. These models can take different compute budget sizes, categorized generally as low, medium, and high, allowing them to reason for longer periods if needed. Giving an already intelligent model more time to think dramatically increases its power.
The second path to smarter AI is data. Because the public internet has largely been scraped, specialty data is the next frontier.
Frontier labs want to make their models better at everything, but there is a parallel movement to build models that excel in very narrow domains. Developers are taking highly capable enterprise models and tuning them to work exceptionally well in specific fields like industrial engineering, nuclear science, medicine, or law.
This is where your proprietary data value becomes a tangible asset. You hold the structured and unstructured knowledge of your entire industry. You have clinical registries, decades of published journals, and specialized frameworks that do not exist on the public web. You have always known this data was valuable, but until recently, there was no practical mechanism to unlock that value at scale.
Once you recognize the value of your data, the immediate question is how to use it. The default path many organizations take is licensing.
Medical societies, for example, are frequently partnering with platforms like OpenEvidence, an AI tool built specifically for the medical community. In some cases, the society receives a substantial check to license its content. In other cases, the society previously licensed its content to a traditional publisher, and that publisher strikes the deal with the AI platform directly. The aggregator pulls the society's data into its environment to train its models and answer user queries.
This approach generates revenue, but it comes at a steep cost. You get a check, but you lose control of the asset. The third-party platform becomes the primary interface for your members, and your association becomes a silent data provider in the background.
A more sustainable AI data strategy requires thinking like an entrepreneur. Instead of selling off your data, you can build a layer of AI on top of it that only your organization controls. You retain the asset under lock and key while selling access to the intelligence it generates.
Building an owned AI product starts with value creation. This requires setting aside concerns about technology or data security for a moment and focusing entirely on the people you serve.
You need to talk to your members and the professionals you want to become members. Do not ask them what they need from the association. If you ask that question, they will likely give you tactical feedback, like asking you to make the website easier to navigate or reducing the number of clicks required to register for a webinar. That feedback is helpful for reducing friction, but it will not lead to a breakthrough product.
Instead, ask them about their daily lives. How do they practice their profession? Where do they experience the most pain?
In many professions, the routine work is highly efficient. A professional might handle the standard cases they see every day with almost instant recall. But the edge cases, the problems that fall outside the norm, consume a massively disproportionate amount of their time. They spend hours doing research, vetting solutions, and consulting colleagues to figure out the right approach.
If you find a pattern of pain across your industry, you have found an opportunity. Your association holds the collective knowledge of the sector. You can activate that knowledge to build a tool that helps people in their day-to-day work. Instead of being an episodic partner that a member interacts with once a year at an annual conference, you become an indispensable daily resource. You provide an AI assistant that sits on their computer or in their pocket, trained exclusively on your vetted, specialized data, capable of crushing that specific professional pain point.
A common hesitation when building these tools is the fear of giving away the association's intellectual property. If an AI can answer any question based on your archives, does that devalue the original content?
Consider how Disney managed its intellectual property during the transition to home video. Historically, Disney kept its classic films in a metaphorical vault. Every seven years, they would re-release a movie in theaters, generating a tremendous amount of high-margin revenue. When VHS became the standard for home entertainment, there was significant internal debate about whether releasing a film on video would destroy this business model.
They tested the concept with Pinocchio, a film that had already been re-released in theaters several times. The VHS release did not cannibalize their intellectual property. Instead, it built more durable demand for everything else Disney offered.
There are ways to slice and leverage intellectual property that open up new revenue streams without diluting the core asset. The technology simply enables you to package the knowledge differently. You are not giving the data away. You are providing a highly efficient interface to query it.
Creating immense value for your industry is only half the equation. The second half is value capture.
Many organizations and businesses create massive value in the world but fail to capture any of it financially. If you build an AI tool that dramatically reduces the time it takes for a professional to solve complex problems, you must monetize it appropriately.
This type of specialized, daily-use tool should generally not be bundled into basic membership. Members do not currently expect this level of software service as part of their standard dues. Imagine you are serving Certified Public Accountants. You might discover that CPAs spend 75 percent of their time solving the top 10 percent of most difficult cases. The standard work flows through easily, but the complex edge cases demand heavy research. If your association can provide an AI tool trained on your proprietary data that saves them 20 percent of that time, you can directly calculate the financial return on investment for that member or their firm.
If the association captures zero percent of that financial upside, the business model is flawed.
You can certainly offer a basic tier of the service as a member benefit, but the full capability should be a premium product. This is the core of non-dues revenue AI. People are willing to pay for tools that create measurable value in their daily operations.
This requires an entrepreneurial mindset, which sometimes conflicts with traditional association culture. But associations are businesses that exist to solve problems in their markets. The difference is simply where the money goes. For an association, profitability is not about enriching shareholders. It is about sustainability. The revenue generated from a successful AI product feeds directly back into the organization, allowing you to create even more value for your community.