Interview It was inevitable that AI would be deployed to help enterprises navigate the labyrinth that is modern software licensing, given the myriad options available from the tech giants.

Chris Brown, co-founder of Onyx, a business which says it is aiming to demystify the process, told us it all kicked off when Microsoft began transitioning to a model where it would sell directly to larger customers.

money
Best pricing model for AI? Work in progress, says Salesforce
READ MORE

“The old model really kind of revolved around the single player, which was the LSP [Licensing Solution Provider], as that kind of gatekeeper,” he tell The Register.

A move to a one where Microsoft sold directly to customers was on the cards for a while. “The way we see it is there are far more options for customers… but with that has come another layer of complexity,” says Brown. “It’s just become really difficult for a lot of customers to get a great answer.”

He compares selecting the correct licensing model to “having to spin the wheel,” where Microsoft might say one thing, the partner ecosystem might say another, and LLMs such as Copilot, Gemini, or ChatGPT might come up with something else. That’s hardly ideal for an organization faced with spending several million dollars or more per year on licensing.

It may be tempting to say that any administrator relying on the output of ChatGPT for licensing advice richly deserves whatever befalls them.

Brown’s approach is to use an LLM, but with a twist. First, the training, is “heavily curated” and based on decades of licensing experience, he tells us.

Then there are inaccuracies in product usage rights that his team needs to catch and correct. Teams being unbundled from Microsoft 365, which took some time to be updated, is an example of a moving target that required full-time staff to monitor the knowledge source and keep answers accurate.

The Onyx tool is intended to provide a licensing helpdesk, “which is based upon any licensing question you can think of.”

Want to know the difference between DSP and MCA licensing? Not a problem. What about the finer nuances? Again, the bot, with its custom training, can come up with answers.

Brown reckons models can perform a deep dive into a user’s questions, however, it won’t come up with an individual SKU. “We provide the licensing information, and then we do provide benchmarking advice,” he says, “but that’s really only a benchmark on the target price that they’re going after.”

As a result, the model is more like an advisor. If it can’t answer a question, it won’t hallucinate an answer, he says. “It will actually push itself to the QA team to go find an answer,” he adds, rather than coming up with something that looks like it might be right. “And that is what customers really like.”

Onyx currently only supports Microsoft licensing. Oracle, IBM, and SAP are on the roadmap, but development slowed slightly after the company realized the depth to which customers wanted to go – more than a simple recommendation engine. Just getting Microsoft’s options into the models has taken just over a year.

Has licensing become so complex that an AI is needed to parse its finer points? Brown isn’t so sure. “Licensing rules have always evolved,” he says. “And I think it’s really that evolution.”

What has actually happened, Brown posits, is that LSPs led customer engagement in the past. The world has changed, and customers are being channeled one way or another. “There’s still advice, but that advice might be biased based on what else is being sold,” Brown tells us.

“So we’re that kind of missing layer, between myself, the partner, the customer, because we don’t sell anything, we don’t sell the services, we’re not trying to push it. Here’s that clarity.”

The Onyx model is an interesting one, and perhaps an appropriate use for AI technology. It recognizes that AI is only a part of the toolset and will also defer to humans rather than spouting convincing nonsense.

That approach, as well as providing trustworthy licensing advice, is one that other AI vendors could maybe learn from. Whether it gets the backing of the storage asset management community – the eco system of licensing professionals that emerged as licensing complexities multiplied – is another point. ®