If necessity is the mother of invention, commercialization is its handmaiden. Defined as the process whereby new products are brought to market, commercialization has been the subject of countless business and academic studies, and with good reason: though estimates vary, there is some consensus that close to 95 percent of inventions fail and only 5 percent actually generate monetary return; at the same time, 90 percent of an invention’s success has been attributed to “getting it out there.”
Typically, new product development is viewed as a multi-stage process that moves through research, development, engineering, production and marketing, with commercialization progressing in parallel through idea, PoC, product development, pre-commercial trials and sales to commercial sales and marketing. In the technology world, however, this linear model begs modification: platforms for continuous innovation, enabled by rapid scale, economical test/dev cloud environments, and disciplined through social and other forms of user feedback, have created more dynamic interplay and an acceleration of commercialization stages including PoC, prototyping, pre-product testing, alpha and beta testing, and pre-commercial product sale. Add to this, subtle variation in demand profiles requiring a variety of different deployment references which can inhibit mainstream technology adoption – the “chasm” between “early adopters” and “early majority adopters” à la Geoffrey Moore – the complexity of channel distribution for advanced technology solutions, and intense market competition, born out of ease of market entry for many digital products (or product extensions), and the challenge in high tech commercialization is clear. To increase odds of success for new products in this fast moving, hyper competitive marketscape, many organizations have transitioned from linear approach to a more fluid commercialization strategy. A good example of this can be found in Watson, IBM’s natural language cognitive platform.
Watson’s path to commercialization has been anything but conventional, entering public life as a winning contestant on the game show Jeopardy! back in 2011, and moving quickly to the opposite end of the gravity spectrum – as a research and decision aid in the medical field. Cancer care hospitals, including Memorial Sloan Kettering, University of Texas MD Anderson Cancer Center and the Mayo Clinic now boast Watson implementations as do the Cleveland Clinic and the New York Genome Center genomic research hospitals. Cloud powered ecommerce implementations followed next, with a more powerful (24x faster), stripped down (90% smaller) Watson designed for customer engagement applications – the “Engagement Advisor” – picked up by USAA, Genesys, DBS Bank (Singapore), Deakin University and ANZ Banking Group (Australia) and Metropolitan Health of South Africa. Other applications released by IBM include Watson Analytics, a freemium offering aimed at bringing analytics and natural language processing to the LoB user, and the Discovery Advisor, a system that provides visual insight into data patterns and connections to support discovery across a number of industries.
But to encourage Watson deployments in all their potential variegated glory, IBM has shifted from internal solution development to construction of a broader Watson ecosystem populated by developers and partners that can innovate on top of the platform, creating a continuous, fluid stream of new applications – in the sense of both use case and software – for the technology. Release of the Watson Developer Cloud in the late fall of 2013, and the establishment in early 2014 of a Watson cloud service at the IBM Research headquarters in Yorktown Heights, NY focused on collaboration with startups for R&D and commercialization of Watson cloud-based cognitive apps and services are two good examples of this approach. To manage commercialization initiatives, IBM also established a Watson Group, investing $1 billion in the new business unit (2,000 researchers and experts), including $100 million in venture investments to support entrepreneurs developing Watson-powered apps. IBM Senior Vice President Mike Rhodin, who heads up the Watson Group, explained at launch of the New York facility: “we recognized that while many of our early partners in healthcare and financial services had done incredible work with us, the applicability of this technology was widespread, and would extend to all the industries that we work in today.” At that time, Rhodin claimed over 100 active development partners, 3,000 partners who were working with IBM to get started on the platform, Watson customer experience centres in Singapore, Dublin, London, Melbourne and Sao Paulo, relationships with several US and Canadian universities for Watson training, and more than a dozen applications in the marketplace as evidence of Watson innovation diversity across geographies, (human) languages, use cases, roles and industries.
IBM’s reliance on partners for Watson innovation is one part function of the creative advantage of broad ecosystem development that is a familiar theme in IT (think Apple) – “What’s coming out of our partners, we couldn’t have imagined in our laboratories,” Rhodin noted – and one part function of the platform itself. Built on IBM’s DeepQA technology and POWER7 processors, Watson is a powerful, self-learning cognitive system that is capable of answering questions posed in natural language. IBM has equipped Watson with millions of documents, including dictionaries, encyclopedias, and other reference sources, and when connected to the Internet, Watson has access to vast content resources to expand its knowledge base. But to develop the context needed for accurate response to queries in specialized disciplines, Watson relies on the input of additional content materials and the creation of language, taxonomies and query structures associated with a particular industry, field of research, etc. For example, LifeLearn Sofie, a mobile app for veterinarians built on the blending of Guelph, Ontario-based LifeLearn IP and Watson technology, was created to leverage the intelligence of content providers, including the American Animal Hospital Association and Western Veterinary Conference, as well as industry leading practices such as The Animal Medical Center in New York City and the Ontario Veterinary Group and Associate Veterinary Clinics. To meet the specific decision support needs of veterinary practitioners, the platform was designed and trained by veterinarians.
Describing development of LifeLearn Sofie, which began in April, 2014, president and CEO of LifeLearn Inc., James Carroll explained that “the essence of the cognitive computing platform that is Watson is really content,” which uniquely positioned LifeLearn as a partner for IBM since the company has depth in information resources for the animal health industry. To prepare for the “ingestion” of content on emergency and critical care (the first subject focus), the LifeLearn development team adapted propriety software code to develop a layer – Sofie – that sits on top of the Watson technology. According to Mark Stephenson, chief veterinarian and chief corporate development officer at LifeLearn, part of the production process involved needs assessment to determine the user interface and user experience requirements to develop good app design, and another part involved development of back end software which connects to Watson via a series of APIs delivered by IBM. IBM also provided a dashboard – the Watson Experience Manager – to help a team of veterinarians and veterinary students led by Stephenson manage ingestion of content, and to perform “data training” to help Watson understand the language of veterinary medicine.
Interestingly, the bulk of the LifeLearn team’s efforts were expended at the beginning of development. The system was initially trained through a series of “episodes,” essentially Q&A sessions in which a question and a paragraph of text (approximately 150 words) containing the answer were submitted into the system. This process was repeated thousands of times to train Watson on questions and answers that typically would be asked by a veterinarian in practice: it’s critical that this process be carried out by subject matter experts, as, Stephenson explained, “the more relevant the question, the better.”
Educating Watson, a “gargantuan effort” requiring an initial commitment of six months to a year, will be an ongoing process. Ultimately, however, algorithms built into the Watson system will take over through self-learning processes. As Stephenson explained, “eventually, as Watson understands the linkages and connections [in veterinary medicine] and starts to understand context, it will come smarter on its own.” To ensure that answers are most relevant, the LifeLearn team will continue to add updated material to the Sofie corpus of information. LifeLearn will also perform ongoing QA testing, and has built analytics into the app to evaluate which questions are posed most frequently as a means to guide future development. Going forward, LifeLearn also intends to add new content categories, expanding beyond emergency and critical care, for example, to address subjects such as cardiology or dermatology, “repeating and enhancing the training process as we ingest that content,” Carroll noted.
Sofie has emerged as a web based app for mobile or desktop use that features LifeLearn design, functionality and content, which engages Watson technology and veterinary data sets residing in the IBM cloud. The veterinarian can ask the system about pre-set topics or ask a question in their own words, and have a set of ranked, relevant responses from multiple sources (assuming the practice has reasonable Internet service) returned in two to three seconds, an enormous time savings over traditional methods of information search that Stephenson argued often mean the vet has to leave the patient in order to retrieve a single answer. The result is better decision support for a group that is indeed in need: Stephenson explained “Our patients are not able to verbalize their symptoms. Veterinary practitioners are generalists and yet they have to deal with every body system as well as multiple species. And veterinarians also often take on the role of manager, financial advisor and marketing/HR resources manager. So we have tremendous challenges as compared to many medical professionals, and this is the reason that LifeLearn’s evolution of the industry is so important to us.”
To support the kind of innovation and investment in the Watson platform that LifeLearn has made, and to better address the budgetary realities of the startup and other communities that are likely to serve as first adopters, IBM has chosen an approach to market that is also less conventional. Instead of charging license fees for the technology, there are no upfront costs for access to the Watson ecosystem – IBM shares revenues from the commercial product with Watson development partners, recurring (rather than transactional) revenues in this model representing an approach that is better aligned with the services orientation that IBM is now focused on in businesses like cloud (another key component of the Watson solution). In addition to tools to facilitate technical integration and development, IBM also provides support aimed at enabling the partner’s business success: “from IBM’s perspective, it’s very important that we and other partners are highly successful. They have been incredibly helpful, and very much a partner with us in terms of development cycles and very active in terms of helping us develop commercialization plans,” Carroll explained, in advance of full commercial launch expected in early 2015 – a first for the Watson Group.
