BYOAI (build your own AI): Enterprise leaders must evaluate low-code AI platforms to alleviate the talent crunch

The most recent HFS survey (December 2019) on the adoption of AI, analytics, and automation revealed that almost 60% of enterprise leaders consider talent scarcity as the biggest impediment in the way of progress with AI and automation.

Exhibit 1: 57% enterprise leaders highlight shortage of talent as their biggest obstacle in achieving their AI & automation goals

What is the biggest obstacle preventing your automation program goals?

Source: HFS Research, “State of Integrated Automation” 2019

Sample: Global 2000 Enterprise Leaders = 317

Enterprise leaders need to find innovative ways to beat this talent crunch. One way is to hone abilities to leverage the increasingly popular low-code/no-code platforms like Google AutoML and Microsoft Azure ML to implement AI and automation solutions.

57% of enterprise leaders say talent is the biggest problem—where will they find enough data scientists? Business!

In the context of AI, analytics, and automation, the answer to the talent shortage lies within the business itself. Business process teams and domain experts can equip themselves with basic training on low-code AI and automation dev platforms. They can become the data science assistants instead of waiting for a team of data science expertise to come to their rescue. Features like canvas-based visual programming and development environments with drag-and-drop modules are commonly available in advanced low-code/no-code platforms. These features can help business process owners translate their domain language and process knowledge into AI-powered modules that can help them augment their productivity, process efficiency, and consequently CSAT and NPS scores.

How a BFSI business process team transformed into an AI and automation team

The business process team responsible for loan default analysis and predictions in a global BFSI organization proactively took up this initiative of improving the accuracy and timeliness of their predictions of loan defaults, using AI and ML. They were a team of about 20 young professionals with no technical knowledge and with experience ranging from five to twelve years.

The team leader used the process knowledge, loan data, and default data from the past 10 years to train the ML modules to predict the probability of defaults. The team based the training on certain features and dimensions that they could easily extract from a digitized version of the loan applications (most of which were already digitized via online form submissions).

The team leader evaluated new low-code ML development tools and used them to experiment with dummy datasets and found the tools to be easy to learn and use, with loads of similar example code and process models available in online training programs and samples. Within a month, using the low-code ML modules from Microsoft Azure ML, the team built a simple but minimum viable solution prototype of their targeted loan default prediction module.

If the business teams have the right will and attitude to learn, low-code platforms offer the opportunity and show the way!

Five key benefits of using low-code AI and automation development platforms in organizations are:

1. Business users are becoming self-reliant instead of waiting eternally for enough data scientists to appear on the horizon when they are currently nowhere in sight—especially not enough to meet demand.
2. Business users can design and develop solutions to best fit their usage scenarios. Hence, their usage preferences, personas, and work behavioral nuances are embedded in the solution, and the problems of adoption, scale, and organizational change management are implicitly taken care of.
3. Business users get a unique opportunity both to build and use a new skill and to multi-leverage and add value to organizational goals in multiple ways, such as creating knowledge assets using their domain and functional knowledge.
4. Highly skilled AI-ML coders are not needed (but supervision is advisable).
5. Using low-code solutions enhances the reusability of knowledge assets. Diverse business process know-how and nuances from multiple actors enrich the business value of the AI usecases.

Exhibit 2: Example action items for business leaders next Monday morning

The Bottom Line: Enable business teams to use low-code AI development platforms. Also ensure the endeavors are overseen by a competent AI leader.

Enterprise AI and business process leaders must enable their teams with training on low-code AI development platforms so that they can experiment within the business taking the first steps toward becoming self-reliant by leveraging these transformative technologies to improve their process quality, delivery effectiveness, productivity, and efficiency.  

But be warned; low-code AI is not a magic wand; it will derive patterns from the data that feeds its algorithms including the inherent bias within. A scattergun approach is not advisable. A comprehensive understanding of the background math at play and pros and cons of available algorithms is essential. Ensure testing is comprehensive and the accuracy levels suffice for business purposes. Develop a means to process and close the loop with further training on the array of false positives and false negatives that emerge. Ensure you can explain any decision AI is making, if necessary. While these platforms will reduce coding effort and enable many shortcuts, the overall context of the tools and problems being solved is essential for success.