Part 2 in the 4-part series “Predictive Analytics and Machine Learning“
Forrester recently released a study titled “Powering The Intelligent Enterprise With AI, Machine Learning, And Predictive Analytics.” According to this study:
“The application of PAML capabilities is important to nearly all companies; 93% said PAML is integral to the ongoing success of their business, and 88% agreed that the next generation of enterprise applications will be infused with machine learning and other AI technologies.”
The problem: how do companies keep pace with internal demand? Today, almost every line of business can make a case for how PAML solutions can add value. Predictive maintenance (decrease cost, increase customer uptime); invoice matching (streamline accounts payable); fraud detection (minimize risk, prevent leakage) – these are just a few examples.
The growth in demand seems quite widespread. As the study shows, 79% of companies see a growing demand for machine learning models. This is driving demand for data scientists – who are in short supply, overwhelmed, and having trouble meeting demand.
One way to help address this problem is by automating tasks for data scientists and providing a platform that streamlines development and speeds deployment. You can read more on this in upcoming blogs in this series by my colleagues Mary Carol Madigan and Tina Tang. Here I want to take a different tack.
Prepackaged for business
If demands are too intense for data scientists, why not alleviate these demands at their source: the business? Why not empower everyday businesspeople to employ PAML to address business problems themselves in a controlled fashion?
Today, PAML technology is advanced enough for standardization. We can now model common business problems at high enough levels of abstraction that they can be used as starting points for a wide variety of common challenges.
With intuitive tools and the right data, business analysts can use these models (categorized according to the type of problem they address) to start realizing value quickly – all in a guided, easy-to-use manner, without having to know how to use R or Python. This alleviates demands on data science teams – while empowering the business to achieve results.
Another problem: Even while data science teams do a good job pumping out new solutions, there remains the challenge in getting them to scale. One interesting finding in the study is that “data-science professionals more commonly felt like their company had a broad application of PAML across various use cases compared to business end users who indicated [that] the number of PAML use cases was still relatively small. This illustrates that end users are not seeing the volume of PAML applications that they’d like, which validates the need for greater automation-focused solutions to address volume challenges.”
For PAML technologies to power the intelligent enterprise, they have to be put to use. Models that are not fully or properly deployed are not useful for anyone. By democratizing PAML via simple, easy-to-use, intuitive tools, business users and analysts can leverage predictive technology – scaling it out to answer business problems and drive business value.
We are talking about solutions that provide access to predictive capabilities via categorized algorithms where the complexity is masked and model training is automated. Ideally, such solutions should have at least three core capabilities (the last representing an emerging capability that is nevertheless vital for ease of use and adoption). Acting almost as microservices that can be reused in different contexts, these capabilities should include:
- Automated insight: There is a lot of data, more than a single human can manage, and a lot of trends and patterns to be identified. With machine learning, those correlations and relationships can be exposed and presented to business users to discover and understand what they mean. Maybe you want to weigh the value of a sale opportunity or compare one group of customers to another. Your PAML solution should have the ability to highlight key influencers and drivers of a particular outcome and present them to you using dynamic visualizations that aid understanding.
- Predictive analysis: After trends are detected and explained, business analysts need help executing. Where do you focus, who do you target, and how do you optimize your business for the best possible outcome? A PAML solution should help present prescriptive actions that allow business analysts to continuously take the steps needed to improve business processes and deliver better customer experiences.
- Natural language search: Common PAML scenarios require the ability to search. But where the data resides should not concern the business user. Nor should specialized query languages be allowed to exacerbate the data skills gap by bottlenecking inquiries with the data team. PAML solutions should support natural language questioning so that everyday users can get the most out of it.
It’s often said that data is the “new oil.” As oil was to the 20th-century economy, data is foundational to the new digital economy. As with oil, companies need to drill deep into this data to extract value. This is what PAML technologies are designed to do – and companies that can put the power of PAML in the hands of everyday businesspeople are the companies that have a better chance of success in the digital economy.
For an in-depth look into the intelligent possibilities for your business, review the August 2018 Forrester Consulting study, “Powering The Intelligent Enterprise With AI, Machine Learning, And Predictive Analytics,” commissioned by SAP.