IBM, whose Watson computer system last year famously beat two of the "Jeopardy!" game show's top contestants, is now looking to deliver the AI-powered computer framework through the cloud.

In a Twitter chat on Thursday afternoon, IBM officials discussed Watson's future.

"IBM will work with clients on deployment preferences but intends to deploy #IBMWatson on both private and hybrid cloud," the IBMWatson account posted.

Now, Watson clones are beginning to show up in the market. The company's first agreement to use Watson technology in a commercial setting involved health insurer WellPoint last year, which is an ongoing pilot that entails Watson being run on an IBM-hosted environment, not on the WellPoint premises.

This week IBM officials inked an "exploratory agreement" with Citigroup to explore how Watson could be used in the financial services sector. Watson computers could be used in a variety of other ways though, the company hinted. IBM may even use its SmartCloud offering to deploy Watson in the future.

"#IBMWatson may look at using cloud to handle data ingestion which process and analyse data for a particular use case," the IBM account stated. The Watson framework requires users to input data into the computer that can then be analysed, which could also be done through the cloud, the company suggested. "Another possible pattern is hosting core #IBMWatson QA engine in the cloud and keeping proprietary data on-prem."

The computing framework will only be available "as a service," the company said, and it even used the term WaaS, or Watson as a Service.

"They're definitely commercialising it," said Adrian Bowles, an independent analyst who tracks business analytics and the software market, and who has worked with IBM on the Watson project.

"If you look at what Watson is, people were thinking about it in terms of blocks and racks of computers competing on Jeopardy!" he said. "But conceptually, Watson is a hardware-software combination that can be tailored and workload optimised for a variety of applications."

For Watson to work effectively, he said, there needs to be a high volume of data that's inputted into the computer system. It requires a "fair amount of training with the customer," Bowles said.

Manoj Saxena, general manager of Watson Solutions for IBM, said Watson computers have the ability to engage in a written dialogue, ask users for additional information and display the logic and evidence the computer uses to create its recommendation.

For example, Saxena explained that in financial settings, Watson could use news reports, blogs, Twitter feeds and transcripts of earnings calls with analysts to "provide a 360-degree view of a customer and a complete contextual view of many decisions that a bank has to make."

In another instance, Watson could assist loan officers in researching if an applicant is legitimate, or recognise people who have a high likelihood of fraud, he wrote. It could help sales representatives assist in financial planning for clients based on the customer's finances, their past banking activity and what people of similar demographics have done.