Microsoft is a very distant second, with about 7%, and Google Cloud accounts for the rest. But when it comes to artificial intelligence products and services, Morningstar is betting big on Microsoft’s OpenAI. 

“We didn’t double down on our majority cloud provider,” says Rhodes. “Microsoft is well positioned with out-of-the-box services available today, so that you can get up and running quickly.” 

As the major cloud providers are also in the business of offering AI products to enterprises like Morningstar, a key decision that’s emerged is whether companies should stick to their cloud providers when creating their AI ecosystems, or perhaps work with another partner—or even multiple partners. 

“I do think that at the high end of the market, they’re going to be multi-cloud, that’s just their reality,” says Spencer Kimball, cofounder and CEO of database startup Cockroach Labs. “And why would they change that for AI? I think they’ll embrace it even more.”

In deciding on OpenAI, Morningstar evaluated the terms of use from the offerings available in the market, because the company was concerned about how the data it would send over to a provider would be used. It felt Microsoft was further than the rest in clarity as to what it would do with the data and how it may be fed back into the models. 

Microsoft had been clear with Morningstar that the company’s data would only be captured for debugging purposes and purged after 30 days. That gave Rhodes peace of mind. 

“I do think what happened was Microsoft was kind of really ahead with the enterprise use case,” says Rhodes. “Honestly, I kind of felt like Amazon and Google were asleep at the wheel.”

Last year, when Morningstar launched the company’s AI chatbot, called Mo, it paired its investment research library with Microsoft Azure’s OpenAI service. 

Cloud computing and enterprise software provider Appian took the exact opposite approach to Morningstar. AWS is the company’s primary cloud provider and the pair are collaborating closely on AI too. “We already had a great relationship, so we are diving in,” says founder and CEO Matt Calkins. 

In April, Appian announced it signed an agreement with AWS to support the company’s new private AI tool. Appian has been an advocate for private AI, which it says represents three core values: models must be trained exclusively on a company’s own data, that data remains firmly under an organization’s control, and the AI models are unique and never shared. 

The tie-up with Amazon Bedrock will give Appian the ability to host large language models within certain compliance boundaries that customers set, and then privately customize those models. The cloud-based machine-learning platform Amazon SageMaker will allow Appian’s clients to create, train, and fine tune those AI models using only their own data.

“Staying within one ecosystem is easier,” says Calkins, who acknowledges that the prior relationship with AWS gave it a leg up on AI. “Provided they didn’t mess it up, we were probably going to pick them. We have such a good history with them.”

Kimball says the cloud providers are extremely different from what they are offering with regards to AI, at least currently. “It is vastly different between the clouds,” he says. “The convergence isn’t there.” 

Larger firms will likely juggle a few different AI offerings, which would mirror their approach to cloud. But for smaller companies, going all in on one cloud and adding AI products and services could result in better affordability and the ease of working within one ecosystem. “At the smaller end, I think the argument around simplicity and discount can hold sway,” says Kimball.

Brian Raymond, the CEO and cofounder of Unstructured, recently raised $40 million from Nvidia, IBM, and Menlo Ventures, among others. The startup that sits at the start of the AI data journey, it ingests and pre-processes unstructured data including emails, documents, images, and video into formats that can be ready for use with foundation models. 

Unlike cloud, when, in the early years, switching providers could be costly, moving among AI companies is simple and inexpensive. “It is like fast fashion,” says Raymond. There are “extraordinarily low switching costs across models, and so almost no vendors lock in.”

Raymond sees two different AI strategies emerging. There’s vertical AI, which focuses on applications that could serve a specific industry, like healthcare or travel. And horizontal AI, which can enhance processes like cybersecurity or customer support for any sector. 

“You’re seeing these cloud service providers pursuing both vertical strategies as well as horizontal strategies simultaneously, in order to hit different market segments,” says Raymond.

ACI Worldwide operates with a hybrid cloud strategy, due to the nature of the payment system company’s business. It has generative AI production products running in Microsoft Azure and AWS, with Google still in the early days. 

Chief Technology Officer Abe Kuruvilla says the company recently rolled out Microsoft’s AI-powered chatbot, partly because it would be easy to deploy, but also due to the fact that it was part of ACI’s existing license. “It’s fairly cost-effective,” says Kuruvilla. “You can see each of these providers doing add-on services for incremental costs.”

But he’s wary of overspending. Kuruvilla says businesses need to keenly monitor some of their AI spend and ensure the return on investment and productivity gains are clear. ACI is currently testing out Copilot for Microsoft 365 and expects to make a decision on it by June. But the per-user cost will be a factor in where it may be deployed.

“We are piloting that; we have a couple hundred people using it and continuing to receive feedback,” says Kuruvilla. “More importantly, we are monitoring utilization.”