How to exploit the insights you just discovered with AI
By Samuel Irvine Casey | 5 minute read | January 12, 2018
– Max Kelsen is building a business by refining raw analysis into competitive advantage for customers; here are a few best practices, distilled, for harvesting the benefits of AI
– AI promises incredible things but to begin, use your AI data for productive projects that can deliver a fast ROI
– We use Watson Knowledge Studio to create tailored lexicons that reflect a specific company’s terminology and language. This sets us up to understand context when mining its unstructured data using Watson Natural Language Understanding
– AI knowledge bases are gifts that keep on giving, because they continue delivering new intelligence depending on who is looking at it, and what they are looking for
Waking up to the benefits of artificial intelligence
Every day, companies around the world are waking up to the benefits of artificial intelligence (AI). Every day, insight engines built with IBM Watson Discovery services find patterns that human operators miss. They are faucets that distill ‘dark data’ –that unstructured information that traditional analytics tools couldn’t unlock –into actionable intelligence.
How can you best put this fresh analysis to use in enhancing your customer experience? Watson’s algorithms are great at analyzing structured and unstructured data, but it takes human imagination and ingenuity to extract and apply analysis for the best results. Executives and knowledge workers must understand this analyzed data, spot the most useful revelations, and use them most effectively in their business decisions.
At Max Kelsen, we’re building a business on bridging this gap for customers by refining raw analysis into competitive advantage. Through numerous projects, we have distilled some best practices in harvesting the benefits of AI.
Here are some of the things we’ve learned about distilling post-analysis AI data into actionable intelligence using the Watson family of cognitive services:
Break down your silos
The data that you need is already in your organization, but it is often ‘dark data’, locked up in unstructured files that no one has thought were even possible to mine for value. Our customer engagements have taught us that you can often find value where you don’t expect it, such as in compliance data sets.
Be thorough. Before extracting and applying the insights from your cognitive computing project, be sure that you’ve drawn on all the data at your disposal. Find out where your organization’s data is, and who has access to it. Cross-departmental boundaries and create allies to unlock that data and bring it into the fold. That will enable you to produce a richer, more productive set of insights that will ultimately benefit the entire organization and its customers.
Pick low-hanging fruit
AI promises incredible things. The media carries stories of self-driving cars and AI-powered robotic delivery drones. They’re cool, but they are also long-duration, high-risk projects. Begin using your AI data for productive projects that can deliver a fast ROI.
These are the projects that automate highly repetitive, low-cognition tasks, such as extracting entities from web pages, classifying images, or assessing automobile claims and that human operators struggle to complete quickly and accurately, and that often takes them away from more interesting and high-value activities. With a service family like IBM Watson, now a machine can handle them. Automate these clerical tasks now and prove your business case before moving on to the cool stuff.
Apply the Pareto principle
Even these typical AI tasks won’t always be simple for a machine to do. In many cases, an automobile claim will take some extra attention. Problems with an image may mean that a computer can’t automatically identify damage to a car door. You might find 20% of the tasks that you’re targeting too difficult to automate.
This is a natural law. The Pareto principle (otherwise known as the 80-20rule) says that 80% of the effects come from 20% of the effort.Don’t spend a disproportionate amount to chase diminishing returns. Leave the 20% of your tasks to human operators, and look for the next project where you can gain big wins by applying what you’ve already learned from AI. That may involve improving your customer experience with a chatbot, or analyzing trends to find operational efficiencies.
Look for context in unstructured data
In the past, analytics programs had a tough time mining unstructured data such as social media posts and call center transcripts, especially at scale. Consequently, companies have ignored that data. Watson changes all that.
Take sentiment analysis as an example. Traditional analytics programs characterized human sentiment as either positive, negative, or neutral. Real people experience a range of emotions, and they relate to specific things. As an example, a customer may experience a full range of emotion in one interaction with a call center operator.
In the past, analytics programs were not smart enough to understand what these emotions were, or what they were about. They tried to work it out using rudimentary dictionaries and keyword maps, but delivered simplistic, inaccurate results that companies could not trust.
We use Watson Knowledge Studio to create tailored lexicons that reflect a specific company’s terminology and language. This sets us up to understand context when mining its unstructured data using Watson Natural Language Understanding.
The two combined enable us to identify not just the emotions in a customer interaction, but the context, too. That isn’t just useful –it’s crucial when creating actionable intelligence that our customers can rely on.
Look for nuances and context in your AI data to truly understand what your customers are frustrated, angry, sad, happy, or enthusiastic about –and why.
Take the insights to the users
By analyzing data from a complex array of sources using Watson Knowledge Studio and Natural Language Understanding, Max Kelsen delivers data-driven insights for our clients to drive deeper understanding and better decision-making. This knowledge base is the gift that keeps on giving, because it continues delivering new intelligence depending on who is looking at it, and what they are looking for.
To fully unlock the value of your data, Watson Discovery helps us to form ad hoc queries for our clients. We also surface unexpected insights for customers by distilling complex data from Watson’s AI engine using data visualization tools. We create custom dashboards that show us where to drill down for more detailed results and understanding.
By making it easier for non-technical users to manipulate these data sets, we can bring instant understanding to the executives in our client organizations. They can use self-service systems to access it when and where they need it most.
Watson’s AI lets enterprises interpret data that they previously could not. To fully exploit its capabilities, we must think about our data in new ways to harvest the richest results. It starts by fully understanding your data portfolio, and by articulating your goals. Get it right, and the benefits may never end.
Learn more about creating insight engines with Watson Discovery AI to solve business problems today.