What are some of the tasks your enterprise must address to get started with using Watson? While it’s going to be somewhat different for everyone, Dayhuff Group’s methodology provides some general insights.

Dayhuff Group is a consulting company and longtime IBM Business Partner headquartered in Ohio that specializes in consulting for enterprises seeking to improve their business intelligence (BI) and enterprise content management (ECM) systems. Dayhuff characterizes their function as “helping companies move from enterprise search to cognitive exploration” and has a track record of more than 200 successful installations.

To reach that goal, one of Dayhuff’s services is to provide strategy sessions at which their consultants analyze a client’s business and provide an “IBM Watson value assessment.” The assessment “leverages the broad portfolio of Watson technologies as well as architectural principles to define the client’s cognitive strategies…. This includes helping define how Watson is used, how it is architected, and how it is sized.” The consultation defines specific use cases for Watson, proposes an end-to-end technical solution, develops a client roadmap, defines implementation of a project to adopt Watson, and carries out a preliminary content assessment.

“Watson enablement,” as Dayhuff calls it, includes assessment of client objectives and goals, design of a Watson architecture for on-premise or cloud use, configuration of Watson technologies, integration of Watson capabilities with existing systems, and employee training in using the new combined system. “Our business intelligence practice helps companies to reduce the complexity of organizing and distributing information and leads to competitive advantages, overall better decisions, and an improved bottom line.”

The Particulars of a Watson Project

In outline, this methodology is fairly standard for most any consulting project. Diving a bit deeper gives some insight into some of the unique challenges of a Watson implementation project.

“We perform one of two engagements,” reports Stephanie Coates, Dayhuff’s senior director of sales and marketing. “We provide innovation collaboration for the client that is exploring the value and applicability of a cognitive journey. We also provide an operational walkthrough for the client that has a clearly defined path for transformation where we need to document the current and future states to frame the solution approach. If the client is in exploration mode, we mentor and collaborate. If the client already ‘gets it’ about Watson’s value and have a path defined, we help them execute. Out of these exercises, we define the best path.” What these exercises entail is at least partly spelled out by Dayhuff’s “BI-Assessment Services” document.

The procedure involves meeting with client users to identify business needs, collecting samples of reports and other business docs that show what needs are currently being unmet, offering IT developer sessions to explore platform options and review existing architecture, reviewing existing data warehouse tools and report designs, and providing some best-practice examples. These sessions are available as one-day workshops or multiweek engagements, depending on the client’s commitment level.

Assessment deliverables include a BI solution roadmap, an implementation statement of proposed work by Dayhuff, an estimate of the financial investment needed to implement the solution, a project timeline, a business-case description of how users expect to use the solution, an inventory of existing systems and apps, an inventory of end users and their skills, and a description of the technical components of the solution. Dayhuff also offers data cleansing and curating procedures “that we offer as the first order of business for our BI clients,” Coates emphasizes.

This methodology isn’t affected by the type of industry in which a client company may be involved. “It’s the solution approach that varies depending on the identified ‘time to value’ and sense of urgency as to whether we design and build or mentor,” Coates reveals. (“Time to value” describes the time between a request for some kind of value and that value’s delivery.)

The procedure also isn’t much affected by whether the client’s emphasis is on BI or ECM. “There is significant overlap, as teaching these systems involves data mining, pattern recognition, and use of information assets from all facets of the business,” Coates points out.

Cognitive Computing and Watson Pitfalls

Watson is all about cognitive computing. Coates defines this term as “the process of using computing power to simulate and mimic the way the human brain works.” Getting clients to define the questions they want cognitive computing to answer “usually comes from identifying the problem they are trying to solve or the advantage they wish to gain,” Coates explains.

In keeping with the concept of mimicking the way the human brain works, part of the Watson cognitive journey is training the system itself. Just like a new employee, a client’s implementation of Watson must be taught how to sift through and evaluate the information to which it has access. “IBM Watson has been trained using massive quantities of data and content so that it provides Answers, not just more information,” Coates expounds. “We advise companies that if you want Answers, use Watson.”

This training process has its own protocols, according to Dwight Bowman, senior solutions architect at Dayhuff.

“First, it’s most important to know that training Watson is an iterative process. Watson learns by example,” Bowman explains. “We’ve found that it’s best to anticipate several iterations of training and reviewing results to get the best answers. Different Watson tools require different input for training. In each case, giving Watson examples to learn from, reviewing the results, and refining the examples enables Watson to learn and improve its accuracy.”

“The tasks start with gathering good examples for training,” Bowman continues. “The samples should match known results that you want to get out of Watson. You should feed them to Watson so it can learn. Then review the results and look for where Watson’s analysis differs from what you expected and use the results to refine a new training set. After a few iterations, Watson’s understanding of what you want to analyze, and your insights into what can be analyzed, will improve to the point where you can count on consistent, quality results from Watson.”

Coates does point out that getting everyone in an enterprise to be on board with a shift to using Watson can sometimes be problematic. “A first task is to demystify cognitive concepts for them,” Coates elaborates. “We educate to help them understand that cognitive computing, and Watson in particular, is not a magic box but rather a collection of computing facilities that have been and are continually being trained to mimic the way the human brain senses, reasons, and responds.  We explore as to how they think about a particular challenge or opportunity and then help them imagine deriving answers faster and more efficiently. Just like a child, it is training, training, training. This is why we call it a cognitive ‘journey’—because it is not only revolutionary but also evolutionary.”

Another stumbling block to adopting Watson can be a basic misunderstanding of how Watson functions. “From time to time, business and IT stakeholders are not like-minded in their understanding of cognitive computing in general,” Coates observes. “In one instance, the business was under the impression that Watson was born smart and that all they had to do was to buy it, plug it in, turn it on, and fire away, asking questions, searching for insights and answers. The time and dedication to training Watson has to be understood and committed to up front in order to maximize the value to the business and to manage expectations. In short, all of the stakeholders need to be engaged collectively on the journey.”

Watson and Power Systems Users

Dayhuff’s clients include users of IBM Power Systems servers. “We have one client that just purchased one to run WEX,” Coates begins. (WEX is Watson Explorer, an engine that can search and analyze nearly all kinds of public and external content.) “They’re developing turnkey systems using Watson and Power Systems horsepower. Dayhuff is also using Watson and Power Systems bundled together for our Pharma Solutions, and we have two other solutions under development.”

Dayhuff Pharma tackles the problems of adverse drug interactions. Using Watson Explorer and custom software called Dayhuff Annotators, Watson sorts through reports of adverse reaction events, manufacturing data, quality control reports, clinical trials, ingredients, and other sources to identify potential causes quickly. This can shorten investigation time from weeks to minutes and enables better identification of drug interaction issues that may have been clouded by inconsistent reporting.

“We are currently developing a proof of concept for a client using the Watson Visual Recognition API bundled with IBM Case Manager for quality control in food production,” Coates adds.

When asked whether Dayhuff sees a future convergence for Watson and Power Systems or the systems operating in parallel for some time, Coates admits, “We have no current insights.”

Coates does report that content management is generally of more interest to clients than straight BI. “ECM is the most prevalent in our experience, but we also see Hadoop, data warehouse, and data entry systems as the most common business platforms. We even have a client who explores large collections of email systems and some social media,” Coates says.

When asked about ways clients can evaluate the dollar value of adopting Watson-based cognitive computing at their enterprises, Coates simply replied, “Reducing the costs associated with regulatory compliance is a common theme.”

While adopting Watson as a means of gaining insight into business problems isn’t a simple task, neither is it endlessly complicated. Companies like Dayhuff Group are available to help both those who want to learn more about it and to sort through the options and come up with a coordinated, comprehensive plan for implementation.

John Ghrist has been a journalist, programmer, and systems manager in the computer industry since 1982. He has covered the market for IBM i servers and their predecessor platforms for more than a quarter century and has attended more than 25 COMMON conferences. A former editor-in-chief with Defense Computing and a senior editor with SystemiNEWS, John has written and edited hundreds of articles and blogs for more than a dozen print and electronic publications. You can reach him at This email address is being protected from spambots. You need JavaScript enabled to view it..

Also Read

Upskilling and reskilling for talent transformation in the era of AI

Dispelling Some Fog About Edge Computing

Laying the Corporate Cultural Groundwork for an AI Project