OpenText in the machine era

At Enterprise World 2017, OpenText carved out new market space in the AI and machine learning future with launch of a new AI platform – Magellan.

“The Information Company” is moving with machines to a new digital era marked by software automation. OpenText, the Waterloo, Ontario based startup that has grown to a position of global leadership in enterprise content management, is now reinventing itself, transitioning from a provider of enterprise platforms for humans to a creator of data platforms for humans and machines. At OpenText Enterprise World 2017, held in Toronto last week, CEO Mark Barrenechea described the profound change he expects to see over the next five years that is driving this transformation: a Singularity University report estimates that 50 million robots will be produced by 2025; to drive IoT, 4.2 billion people will be connected on the same network by 2020 and 1 trillion machines will be connected by 2036; currently there are 500 digital currencies that are making the ‘Internet of Money’ a reality and turning the network into an API; AI is booming and mobile is “eating the world”; new business models are emerging in areas such as healthcare that will extend human life to age 150; and with machine assisted learning, soon everyone will have an IQ of a 1,000. At the event, Barrenechea also outlined how the company is working to address this change, offering several specific examples of OpenText product innovation is these areas. In IoT, for example, location based analytics are being integrated into the Adobe Prelude product to enhance media capabilities, in the digital currency field, OpenText is working on a project with Citibank, and in AI on risk management for banking, the company is building a new database system with straight through processing, and using Rosetta Stone as a natural language processor to ease tech consumption. But across each of these areas and more, OpenText has focused on AI as the true Deus ex machina – the real ‘god from the machine’ that is poised to solve problems across the range of OpenText and customer solutions.

CEO Mark Barrenechea in conversation with OpenText Enterprise World 2017 guest Wayne Gretzky

The name of OpenText’s new AI platform launched at Enterprise World 2017 is Magellan. Combining open source machine learning via the integration of Apache Spark’s distributed ML library, OpenText Analytics based on the company’s 2015 Actuate acquisition, and (Nstein) semantic navigation – or keyword search that retrieves related content based on a “semantic footprint” – the new platform boasts management of massive amounts of structured and unstructured data, and can be deployed across any EIM architecture to speed and ease implementation of data driven decision making and task automation. At the event, Mark Gamble, senior director, product marketing, OpenText Analytics, put Magellan through its paces in several different scenarios designed to demonstrate the potential of “drag and drop AI” algorithms. By applying embedded algorithms to additional data sets to fine tune marketing targets, he demoed Magellan’s ability to deliver “progressive insight” to staff who may not qualify as data scientist – “democratizing AI” through greater accessibility to analytics models. In a predictive maintenance example, he showed how internal staff – OT engineers, in this case – can program the system using a range of standard languages such as Python or R to create proprietary algorithms, using common code to replace expensive professional services that otherwise might be required to produce custom results. And returning to the robot theme, OpenText presented a final demo, in which Marcel Hoffmann, program manager, ecosystem solutions for industrial robotics company Kuka described the potential to use the Kuka cloud and OpenText Magellan and xECM (extended Content Management System) to discover and link case study documents on specific maintenance issues to support collaboration on the development of automated repair systems.

Kuka robot

Dubbing its new AI platform a ‘cognitive system’, OpenText is creating space for itself in the fast growing AI market, which by most analyst accounts is doubling on an annual basis. In this effort, the company is taking sharp aim at the grandfather of commercial cognitive systems, IBM’s Watson. In his keynote, Barrenechea reinforced the OpenText cost advantage, noting that Magellan may be delivered in a pre-certified x86, 6 or 7 U hardware configuration at a sixth the cost of IBM system ($100,000 vs 1 million), generating savings that are additional to the avoidance of professional services fees that Mark Gamble pointed to. But are Magellan and Watson apples to apples? On this score, there are a number of capabilities that OpenText has on board for future development. As EVP OpenText Engineering Muhi Majzoub explained, Magellan combines content analytics, semantic analysis, semantic navigation and autoclassification to support contextual analysis, and is now working on machine learning to better train machines to execute on specified rules. Search capabilities, however, are textual – based on keywords: “Video, image recognition and audio are things that we are working on, but not announcing this in phase one of Magellan capabilities,” he added. “These are projects that are work in progress.” Similarly, a natural language interface is a work in progress, which Majzoub advised will be part of next product iterations, along with advanced machine learning and processing of alternative media formats.

Tom Dong, VP product marketing, OpenText

According to Tom Dong, VP of product marketing at OpenText and former director of product marketing for IBM Watson Analytics, some key differences set the platforms apart: “There’s a lot of education of the market that is required, and it’s important not to confuse ‘deep learning’ with ‘machine learning’, as these effectively point to different techniques for doing classification. Typically, deep learning is neural networks. We actually have neural network technology within our portfolio in our cognitive capture capabilities, but with version 1 [of Magellan], we have taken the pieces that are easy to connect to Apache Spark and MLlib.” But in his view, there are also limitations on the practical utility of the deep learning approach: “When IBM is talking about machine learning, they’re talking about deep learning. Neural networks is a concept that was very popular in the 60, but it fell out of favour because if you didn’t have the right data, there is a huge trust issue that could be especially challenging in areas like control systems. Deep learning has come back – like disco – but what is the value of finding patterns in data? You still need to do you complex analytics on top of that in order to do more than simple monitoring.”

For Dong, control functionality is the key differentiator for any platform: “Prescriptive analytics should be the ultimate definition of machine learning, where you make the machine smart enough to mimic how the human would actually do something. The last frontier is technology that actually tells you what to do” or automates the activity. While this capability is not currently available in Magellan, Dong noted that the platform’s open architecture would allow the enterprise to independently build in the “do.” And through its integration with the OpenText EIM portfolio, he argued, Magellan is better positioned to support the creation of practical, useful business automation – based on the wealth of data that can be accessed in enterprise systems and on OpenText process expertise. “The advantage we have is that we own the data models. We own the data models for business networks, the models for customer engagement management. We know what the data for a customer represents and can connect the dots a little bit better than say, Tableau.” OpenText’s firm grasp of the models lies in its long experience helping customers manage their documents and processes. “We know what their invoices look like, and what are the transactions in their business networks,” Dong explained, and this knowledge of processes and information flow offers unique understanding of the right data to capture, the right questions to ask of the data, and the right reports to deliver to whom. Barrenechea called this focus on process automation and machine learning to solve common customer challenges “Applied Analytics” – machine learning applied to well defined use cases to solve business problems where the company has domain expertise (invoice management, financial transactions in the OpenText business network, HR, for example) and has demonstrated the ability to help customers automate – harnessing software automation to overcome the looming scarcity of data scientists needed to make AI an action item.