Data-driven Applications (part 2)

Recently I posted an article on “data-driven applications” with a question on three example scenarios(refer:?link). The answer is, that the first two scenarios give a sense of intelligence from a user's perspective but do not require AI/ML. In scenario 3, the operator dashboard application behaves differently across multiple manufacturing site implementations. A data-driven application's logic depends on the transactional data within the application.

Businesses considering ERP / Customer service / Sales Force Automation / Manufacturing / Supply Chain solutions need to take a hard look at the “data-driven aspect” of the application, to remain competitive.

So, what are data-driven applications and what benefits do they bring? There are many benefits, but two key ones that come to my mind are:

Help businesses get more proactive:?Many organizations are struggling with the fact that most of their work is reactive rather than proactive. Businesses having a better ratio of proactive activities aligning with their objectives are more successful. A key aspect of data-driven applications is that they can foretell things before they happen, giving the necessary lead time for organizations to address them. Examples:

• Predictive asset maintenance: An unplanned downtime of business-critical operational assets can lead to huge losses. Data-driven maintenance applications analyzing data (sensor information) from assets can automatically schedule preventative maintenance to minimize or eliminate unplanned downtimes. This use case, often quoted in keynotes or executive presentations, is estimated to be a 25$Bi annualized addressable market by itself.
• Quality Assurance in Process Manufacturing: This is scenario 3 in my earlier post. A data-driven manufacturing software application can provide operators proactive guidance to meet quality requirements and thus avoid batch rejections.

Automate parts of the business:?Data-driven applications simplify business processes with automation. With its inherent nature of being intelligent, these applications can make timely decisions automatically on behalf of or even mimic real users. Examples:

• Quality Inspection in manufacturing: Quality inspections on large volume manufacturing (made-to-stock scenarios), are carried out randomly. Based on the manufacturing data gathered, a data-driven intelligent application could automatically recommend the number of “finished products” to be inspected for each production run, to make this process more accurate.
• Customer Service Management: A data-driven customer service management application can review textual descriptions in customer complaints using natural language processing. These applications can automatically recommend appropriate resolutions to the service technician attending to the customer's issue. These recommendations are based on learnings from the past solutions provided for similar complaints and telemetry data of the equipment usage at the customer.

In all examples above, the machine learning models (intelligence logic) must be trained on the customer's dataset, requiring data scientists’ involvement during implementation. Advanced data-driven applications can guide and propose appropriate models based on customer data, minimizing data science efforts.

In the customer service management example, the natural language processing part of the application is not dependent on customer data. Incorporating well-proven machine learning models for audio, visual, and language interpretation in enterprise applications can be the first step to creating intelligent data-driven applications, making them easier to implement.

The pitfalls of such applications are obvious! Imagine a situation where the application suggests adding an incorrect additive during manufacturing, which can result in the entire batch being rejected. Incorrectly configured and trained models can provide inaccurate results that can have adverse impacts.

So which parts of your business do you think will hugely benefit from data-driven applications? Thoughts?