The horror… the horror…
Ok, maybe that’s a little dramatic. But if you lead a content or data company, there is one question I suspect is underneath a lot of the discussion and thinking about AI licensing.
Could a deal that creates valuable new revenue today eventually weaken the business we already have?
That is the conundrum.
There is clearly a large new market forming around AI, and content and data companies have good reasons to participate. But if an AI company can use your content to give its customers much of the value they currently come to you for, what happens to your subscriptions, advertising, licensing revenue, traffic, or customer relationships?
So the challenge is not simply whether to engage with AI. It is how to engage without giving away so much value that you undermine the business you spent years building.
How do you move forward without betting the farm?
What is your farm?
The answer is different for every company.
For some businesses, the primary model is B2B data licensing. The farm might be a proprietary database and the rights, technology and relationships that make customers come directly to the company to access it.
For a publisher, the farm might include a subscription offering to valuable content, advertising revenue, intellectual property, the brand customers trust, and the position the company has established as the place people go when they need reliable information.
There are many types of farms, but the ones I’m talking about have valuable content and data that bring people to them, that they can sell, and that they control the customer relationship.
Although we can’t predict the future, it’s a bit easier if you’ve seen a similar movie before
Like many of you, I’ve lived through several technology transitions. These seem to happen faster and faster lately.
When I worked at Liquid Audio and MusicNet during the early days of digital music, it was all about creating a new way of distributing music – downloading and streaming songs. So, similar to now, companies that controlled music had to decide: should we work with these new distribution platforms, or should we fight against them (Sue or Woo, as I’ve heard people explain it).
I realize this is an obvious, simple example, but I bring it up because now we know where it all landed (we all know the future is much easier to understand after it has already happened).
When the world you’ve been living in suddenly changes and you don’t know the future (of course), what do you do? Act? Wait?
I say both.
Protecting an old distribution model forever can be dangerous if customer behavior is moving somewhere else. Moving too quickly can be equally damaging if a company gives up durable value before it understands what will replace it.
We can make educated guesses about where it is going with AI, but we are living inside the transition. That should create some humility about predicting exactly how content will be distributed and monetized five or ten years from now.
If customers eventually prefer to consume information through an AI interface, a content company may decide that allowing an older business model to decline is perfectly rational, provided the new model creates a larger and sustainable business.
The content or data company must understand the trade.
It’s about digging into the nitty-gritty
Part of the difficulty is that what we mean by the phrase “AI licensing” is evolving so quickly. A few years ago, the AI licensing deals getting the most attention were largely ‘training deals.’
“AI licensing” isn’t one thing anymore. The same content can be licensed in very different ways, with very different consequences for your existing business. The more precisely you understand what you’re licensing, how it can be used, and what the AI company’s customer ultimately receives, the more intelligently you can protect what matters.
So you should know exactly what you have and what you want before making a big commitment
With so many options, and so many unknowns, I think companies should treat their early engagement with this market as a carefully designed learning process.
Start by understanding what you have (see above). How does your content or data fit the AI market? Who might actually need it? What makes it different or useful? Are the rights clear? Is it structured and technically ready to be delivered? Which licensing models make sense for your business, and where are the major risks?
Then determine your commercial guardrails before you enter a serious negotiation.
What content are you comfortable making available? For which uses? What happens if the AI company wants to use it in other products or provide access to other companies? How much control do you need over user-facing outputs? What survives after the agreement expires? What reporting do you need to understand how the content is being used? How much of your existing business could conceivably be displaced, and what economics would justify taking that risk?
Some of these questions eventually become contract terms that need to be handled carefully with qualified legal counsel. But the underlying decisions are business decisions, and leadership should understand them before the lawyers start turning them into language.
There also needs to be internal alignment. Business development may see a large new revenue opportunity while marketing worries about the brand, product worries about the customer experience and finance worries about cannibalizing an existing revenue stream. Those conversations should happen before outreach, not after somebody puts an attractive offer on the table.
And when uncertainty remains high, I am increasingly convinced of the value of pilots.
A pilot can use a subset of the content, a limited scope and clear reporting to answer questions with actual market behavior. Is there real demand? How is the content being used? What does it contribute to the AI product? What economics emerge? Are customers using the AI product in ways that could substitute for the existing business? What have we learned that should change the next agreement?
There are already examples of content owners experimenting with this kind of controlled approach. HarperCollins, for example, sought author permission for selected nonfiction books in an AI training arrangement. The reported arrangement also included output restrictions designed to limit how much of a book could be reproduced. (See: The Authors Guild)
That does not mean those particular terms are right for everybody. It shows that “yes” and “no” are not the only available answers. Scope, use, content selection, outputs, economics and reporting can all become part of the discussion.
You don’t need to know exactly where this ends
I feel the underlying anxiety for those running content companies today, and I certainly don’t know exactly where AI is taking the industry.
Yes, obviously, you can’t do “nothing.”
There is too much investment, product development, and growing use of AI for content and data companies to simply wait and see what happens before acting Moving early can create advantages too: better relationships, faster learning, and potentially stronger negotiating positions as the market takes shape.
The answer, at least as I see it today, is preparation followed by deliberate engagement. Know what you’ve got and have a professional way of presenting it to AI companies, and know the type of deal you want to do. You may not get that exact deal, but you should know before getting led in the wrong direction.
Understand what makes your company valuable. Learn how AI companies might use it. Know what you are willing to trade and what you are trying to protect. Do the math on what could be gained and what could be lost, including several years into the future. Start with controlled experiments when you can. Capture what you learn and let that evidence guide the next move.
The more a company understands its assets, the market, the economics, and its own guardrails, the less this has to feel like one enormous wager on an unknowable future.
You can learn your way into the AI market.
And hopefully, five or ten years from now, you will still have the farm.
Maybe even a bigger one.
