Article Synopsis :
In 2013 the MD Anderson cancer institute tried a ‘moonshot’ to deploy AI for the detection and treatment of certain cancers. After costs topped $62m, they had nothing to show. They also used AI for less ambitious jobs, such as making hotel reservations for patient’s families, determining which patients needed help paying bills, and addressing staff IT problems. The results here were much more promising, driving higher levels of patient and staff satisfaction.
“Artificial Intelligence for the Real World,” from the Harvard Business Review, explores the current market in AI with a focus on achieving real business results.
There are presently three classes of AI:
Process automation: the automation of digital and physical tasks—typically back-office admin and finance—using robotic process automation (RPA). Tasks include:
- Transferring data from e-mail and call center systems into systems of record (e.g., address changes, service additions)
- Replacing lost credit or ATM cards, reaching into multiple systems to update records and handle client communications
- Reconciling failures to charge for services across billing systems by extracting information from multiple document types
- ‘Reading’ legal and contractual documents to extract provisions using natural language processing
RPA is the easiest and cheapest of the cognitive technologies to implement, it’s also the least smart, i.e., not programmed to learn and improve. Look for repetitive tasks. A good rule of thumb: If you can outsource a task, you can probably automate it.
Cognitive insight: using algorithms to detect patterns in vast columns of data and interpret their meaning. “Analytics on steroids” to:
- Predict what a particular customer is likely to buy
- Identify credit fraud in real time; detect insurance claims fraud
- Analyze warranty data to identify safety/quality problems in automobiles (& other mfg. products)
- Automate personalized targeting of digital ads
- Provide insurers with more accurate, detailed actuarial modeling
Cognitive insights via machine learning differ from traditional analytics in three ways:
- More data intensive and detailed.
- Models tuned on some part of the data set.
- Models get better—using new data to make predictions or categorize improves over time.
Deep learning—mimicking the activity in the human brain to recognize patterns, such as facial recognition and speech—though historically labor intensive in term of data curation, is increasingly able to identify probabilistic matches (data likely to be associated with the same person or company but that appears in slightly different formats) across databases. Use cases include data extraction from contracts, programmatic ad buys, and massive data crunches typically beyond the range of humans.
Cognitive engagement: engaging employees and customers with natural language processing chatbots, intelligent agents, and machine learning:
- Intelligent agents for 24/7 customer service for a growing array of use cases, from password requests to technical support (all in the customer’s natural language)
- Internal sites for answering employee questions on IT, benefits, and HR policy
- Product and service recommendations for retailers, increasing personalization, engagement and sales
- Health treatment recommendations to help providers create customized care plans taking into account health status and previous treatments
The article provides the following four-step framework for integrating AI to achieve business objectives:
- Understand the technologies: Rule-based expert systems and RPA, for example, are transparent, but neither is capable of learning/improving. Deep learning, on the other hand, is great at learning from large volumes of labelled data, but it’s almost impossible to understand how it creates the models it does, typically not good for regulated industries. Data scientists, big data engineers, and statisticians are essential to success.
- Create a portfolio of projects: Identify the opportunities and develop a prioritized portfolio of projects, typically via workshops/small consulting engagements. Dig into parts of the company where ‘knowledge’—insight derived from data analysis or a collection of texts—is at a premium but for some reason isn’t available. Look for bottlenecks in the flow of information and scaling challenges (knowledge exists but too time-consuming or expensive to scale).
Determine the use cases: Ask the following questions—How critical to overall strategy is addressing the targeted problem? How difficult to implement, technically and organizationally? Worth the effort? Prioritize on short- and long-term value, and how it integrates with other cognitive capabilities.
Select the technology: Take incremental steps with current technology, planning transformational change in the not-too-distant future. For example, you may want to turn customer interactions over to bots, but for now automate internal IT help desk as a step toward that ultimate goal.
- Launch pilots: The gap between current and desired AI capabilities isn’t always obvious; create pilot projects for cognitive apps before rolling out across the enterprise. Utilize PoCs for projects with high business value or to test different technologies at the same time. Cognitive Centers of Excellence (COEs) not a bad idea.
- Scale up: Pilots are relatively easy, not so rolling them out organization-wide. Need detailed plans for scaling up, requiring collaboration between tech experts and the business process being automated. Because cognitive tech supports tasks and not entire processes, scale-up requires integration with existing systems and processes—typically the greatest challenge on AI initiatives. For scale-up to achieve desired results, improved productivity is a must. Will you grow into productivity (adding customers and transactions without adding staff)? Or achieve headcount reduction (attrition and/or elimination of outsourcing)?
Link to Full Article:: click here
Digital Insurer's CommentsWith cognitive technologies, information-intensive domains such as marketing, health care, financial services (insurance), education, and professional services could become simultaneously more valuable and less expensive to society. Business drudgery in every industry and function—routine transactions, repeatedly answering the same questions, data extraction from endless documents—could become the province of machines, freeing humans to be more productive and creative.
Amid valid concerns about worker displacement, the upside of smart machines is their enablement of new levels of productivity, work satisfaction, and prosperity. Especially in the US insurance industry, where 400,000 workers are expected to retire by 2020, cognitive has a place in the enterprise, not tomorrow, but today.
Link to Source:: click here