Artificial intelligence: fulfilling the failed promise of big data

Register now

The topic of artificial intelligence is dominating discussions of data management this year. But while a growing number of organizations are interested in AI, many don’t fully understand what the technology can do to help boost their customer engagement or the bottom line.

Forrester Research analyst Brandon Purcell has recently authored two reports on the current strong interest in artificial intelligence and what can be expected from it. In part one of Information Management’s interviews with Purcell yesterday, we discussed “The Top Emerging Technologies in Artificial Intelligence.” In part two today we discuss the report “Artificial Intelligence Technologies and Solutions, Q1 2017.”

Information Management: Artificial intelligence seems to have replaced big data as the big data theme for 2017. What is your sense of exactly how many organizations are working with artificial intelligence and where are they in the process?

Brandon Purcell: I’d agree that AI has replaced “big data” as the buzzword du jour, but in my mind it actually has the ability to fulfill big data’s failed promise. Big data really focused on capturing massive amounts of data from multiple sources. Companies got really good at that, but they’ve struggled to turn that data into insights and insight into action. The promise of AI is to complete that process - from data to insight to action - in a virtuous cycle that optimizes continuously.

According to Forrester’s Business Technographics survey of over 3,000 global technology and business decision makers from last year, 41percent of global firms are already investing in AI and another 20 percent are planning to invest in the next year.

Most large enterprises’ first foray into AI is with chatbots for customer service, what we call “conversational service solutions.” These run the gamut from hard coded rules-based chatbots which aren’t artificially intelligent to very sophisticated engines using a combination of NLP, NLG, and deep learning. From a customer insights perspective, many companies are starting to uses some of the “sensory” components of AI such as image and video analytics and speech analytics to unlock insights from unstructured data.

IM: What the top reasons that organizations are adopting artificial intelligence and what gains to they hope to realize?

Purcell: Organizations are adopting AI to optimize the customer journey from discovery through conversion, all the way to the end of the customer lifecycle. AI promises to automate the process of understanding customers and anticipating their needs, then delivering the right experience to them at the right time. Organizations are hoping to impact the top line by acquiring new customers and increasing the value and lifetime of existing ones, and they’re hoping to impact the bottom line as well by reducing costs through automation.

IM: What are some of the top obstacles or challenges to achieving success with an artificial intelligence effort?

Purcell: The primary challenge is and will always be the data. Data is the lifeblood of AI. An AI system needs to learn from data in order to be able to fulfill its function. Unfortunately, organizations struggle to integrate data from multiple sources to create a single source of truth on their customers. AI will not solve these data issues - it will only make them more pronounced.

After data, traditional people and process challenges come into play. Who owns the AI initiative? Typically the group in the organization with the technical skills to implement AI is not the same group that will actually own its execution. We see companies fumble this handoff all the time. And how will you measure success to prove the ROI of the effort? Rigorous measurement processes still remain elusive for most companies.

IM: What are your thoughts on artificial intelligence best practices that organizations should use to best achieve success?

Purcell: Start with a narrow use case and make sure you have data for it. Then bring together internal stakeholders and agree upon how you’ll measure success. For example, a subscription-based business may want to decrease customer churn.

They probably have historical data on past customers who have churned that they can use to train a model. They may also have data on retention incentives that have worked in the past. Assemble the marketers who will oversee the retention campaign as well as the data engineers and scientists responsible for building the model. And agree upon a measurement methodology.

Traditional text and control works quite well. Treat one set of customers and see how much higher their retention rate is than a holdout sample after a specified period of time. Assuming the success of the project, you can build the business case for further investment.

For reprint and licensing requests for this article, click here.