Joseph Olassa, CEO, Nuivio Ventures & Ignitho Technologies, with expertise in Digital Engineering, Data Science and Cloud Infrastructure.
The need to effectively generate insights from your business data is undisputed. According to Gartner’s business composability study51% of the more than 2,000 CIOs surveyed will invest more in analytics by 2022.
Astute practitioners will harness the power of the cloud, create a solid governance model, connect applications to a data fabric, and often revisit the models to refine them. However, it is often a complex undertaking to drive business composition backed by meaningful business insights. In this post, I want to highlight two aspects that can make your AI initiatives even more successful.
The first is about purposeful design thinking that allows you to ask the right questions at the right level, and the second is about creating a closed loop to really amplify the power of your insights. This approach is powered by the data science framework at one of my companies, Ignitho Technologies, in partnership with Cambridge University’s “economical innovation“conception.
Asking the right questions at the right level
A technique called “design thinking” can help with this. However, let me start by outlining a common risk. Often organizations start a data analysis project by understanding the use case and the issues that the stakeholders are facing. They then define a solution, quantify the benefits and get to work on implementation. But by zooming in on a problem and then immediately solving it, you risk missing the bigger picture.
Suppose a healthcare company uses analytics to improve the use of a customer portal. The company may be so focused on understanding user challenges when using the portal that it may miss needing an entirely different, mobile-first approach.
In addition to asking the right questions by evaluating the customer’s point of view, consider the goals and objectives the company wants to achieve itself. Using natural language processing, you can discover usage patterns to extend the traditional approach to design thinking.
Extending the simplistic example of the past, improving customer portal usage may not seem like a win when customers demand a push-versus-pull approach. Taking it a step further, the company’s long-term goals could be to adopt an embedded commerce and experience strategy, perhaps with a greater focus on mobile and Internet of Things capabilities. So this particular AI initiative, while useful and perfectly valid in its own right, will not find a good match with the direction the company is expected to take.
Given this additional context, the development of new customer interaction capabilities should take precedence over a specific AI project. You may even now want to use analytics and AI to increase the effectiveness and adoption of this new emerging capability area.
Using a design thinking approach may seem obvious, but it’s easy to develop tunnel vision when you’re in the weeds. Asking the right questions at the level of the customer, not just the users, and aligning them with the strategic goals of the business should be incorporated into the governance model around AI initiatives. In addition to investing in the right issues, this process can also increase team morale and productivity.
Creating a closed loop to increase the power of insights
There’s nothing more exciting than seeing how analytical models yield insights that you can quickly implement to realize productivity or revenue gains. In fact, a significant number of AI project lifecycles find their natural end with prototyping, testing and the first successful implementation.
In my experience, two common problems plague most AI programs. First, the input data used for the first AI launch has been secured (collected and cleaned up) after a lot of hard work. As a result, keeping up with testing and refining models becomes a challenge and is often prioritized after the initial implementation. After all, the model works and delivers results.
Second, it is well known that the wider the range of inputs, the better the long-term effectiveness of an analytical model. Unstructured data, such as user-generated content, should also be included. For example, consider a customer churn prediction model for a media publication. The model can provide excellent insights by taking into account readership patterns and innovation patterns. However, these are likely to lag behind customer engagement indicators.
You can probably improve the model by building in early warning systems that take into account data about broader customer interests and other publications they read. Securing such data may require integration not only with additional resources within the enterprise, but also by ingesting data from digital capabilities that may already engage customers in various interactive experiences.
Supporting such a continuous closed-loop capability that continuously improves your model requires robust data processing and data pipeline infrastructure. In the excitement of building and testing an analytic model, this important consideration is often relegated to the background, with dedicated budgets to build it. However, the data processing infrastructure is usually better built incrementally as an integral part of several key AI initiatives. That makes it much more manageable and also more reflective of the main goals and priorities of the organization.
To analyze the maturity of your company in the field of analysis, you can use the analysis of my company short online review† By harnessing the power of design thinking at the right level and building responsive data operations along with your models as much as possible, you can truly unlock the power of data in your enterprise.