The use of computers to automate human tasks and simulate the human thought process—the field now widely called cognitive computing—is attracting the attention and investment of some of the most well-known names in technology. More and more often, we are seeing automation and cognitive computing solutions that use statistical modelling, machine learning, natural language processing and other sophisticated capabilities to accomplish complex and learned operations. Google’s RankBrain artificial intelligence system processes search data to produce increasingly relevant search results based on what it has “learned” from past searches. Twitter has recently announced the acquisition of Magic Pony Technology that uses neural networks and machine learning for visual processing. Jack Dorsey, Twitter CEO and co-founder said, “Machine learning is increasingly at the core of everything we build at Twitter.”
For the past few years, advances in automation and cognitive computing also have been making their mark on the IT and business process services world, making certain repetitive, rules-based and mundane tasks faster and less labor-intensive. Because they can make such a dramatic impact in specific domains, these technologies are changing the way companies operate. Expectations for these solutions are high—and potential returns are great. For many enterprises, robotic automation has delivered 20-50 percent direct savings.
Though it’s easy to get caught up in imagining the potential of these solutions, their efficacy in the context of IT services can be validated by just a couple of simple proof points:
- Do they achieve higher efficiency? How much faster, cheaper and more consistently than humans do software bots complete repetitive tasks?
- Do they improve solutions for complex problems? How well can cognitive technologies solve (or assist a human in solving) higher-order problems that yield better results than is possible using human judgement and data?
While automation solutions can help achieve the first of these, cognitive solutions, powered by artificial intelligence, can help achieve the second. But various automation and cognitive solutions are at differing levels of maturity, making it difficult to know which will work best in any given environment. Enterprises need to be able to reap the efficiency gains in the short-term, while they experiment and solve higher-order problems in the longer term.
In today’s market, we are seeing four basic models for implementing automation and cognitive computing solutions.
- Embedded automation
In this operating model, the outsourcing service provider drives the automation initiative to achieve higher efficiencies for specific tasks in a traditional outsourced delivery model. The investments in automation and the benefits (typically cost savings) are shared with the client, depending on the nature of the commercial model. Since the service provider drives the automation, the client has less control over the nature and degree of automation. This model is ideally suited for less complex use cases.
- Automation as a Service
Here the enterprise client identifies an objective for implementing automation and engages an automation-savvy provider to automate these specific process areas. To support the initiative, some enterprises create an internal automation center of excellence that engages with the ecosystem and internal business units to evangelize and manage change. Because the enterprise drives the initiative, it realizes the benefits (and savings) and compensates the automation provider for the services. This model is best for clients that prefer more control over the automation initiative and for use cases that are low-to-medium complexity. Automation as a Service is the most mature operating model in today’s market.
- Provider-owned point solutions
Certain service providers have invested in developing best-in-class point solutions for specific domains. Most of these solutions contain elements of cognitive technology, such as social listening for marketing, image matching and natural language processing. This model is ideal for cases in which the business problem is complex and there are specialized solutions available in the market. In such cases, enterprises find it more beneficial to plug in such point solutions than trying to build their own, though of course this means the intellectual property is the provider’s or software vendor’s.
- Client-incubated innovation labs
This is ideally suited to address complex industry-specific problems for which solution components (but not whole solutions) exist. Given the iterative nature of such pursuits, this model works when a buy-side enterprise partners with a capable provider to set up a joint innovation lab to develop a custom solution in an agile approach. The ownership of the intellectual property and commercial model varies depending on the capabilities and investment of the two parties. Though the client-incubated innovation lab model is the least mature in the market today, it shows the greatest promise for taking advantage of cognitive software solutions that provide competitive advantages.
Automation, driven by A.I., and cognitive computing offer exciting opportunities to help organizations drive business results. Understanding which model is the best fit is the first step to enabling successful automation and cognitive implementation.
Sathish Seshadri is a Principal Consultant with Information Services Group. As an experienced consultant with 13 years of experience in management consulting and outsourcing advisory, Sathish helps clients achieve operational excellence and cost optimization through outsourcing and offshoring. He focuses on defining business-led strategy and taking advantage of emerging technologies.
Subscribe to Data Informed for the latest information and news on big data and analytics for the enterprise, plus get instant access to more than 20 eBooks.