How to Start Your AI Pilot and the Journey Beyond

How to Start Your AI Pilot and the Journey Beyond

Rigvinath Chevala

Rigvinath Chevala, Chief Technology Officer | Evalueserve

Artificial intelligence (AI) and machine learning (ML) continue to grow and are expected to contribute significantly to the global economy. According to a PwC report, global GDP could be up to 14% higher in 2030 due to AI  — an additional $15.7 trillion.

Businesses have taken note of this, and many are looking into the possibility of AI adoption. As stakeholders, you know this comes with risks. AI adoption is no longer a yes or no decision, however. It’s a matter of when and how to begin. Businesses must identify use cases that AI can solve, providing immediate ROI to justify the means. 

But getting started is important, so let’s dive into what that looks like. Begin with identifying the right use case. It is essential to start with the right resources, including data, subject matter expertise, and talent. Most importantly, the pilot should be designed in a way that it can scale up for production use. Try and make it repeatable and reusable in the enterprise. That way, it has continued ROI. 

Choose Your Business Case

Select a business case or automation opportunity that is small and simple. An impactful use case requires a sponsor to articulate a value statement and clearly define expected outcomes. With multiple variables at play,  it is advisable to start small and not risk too much at once. This way, if it fails, the impact on your organization is minimal. Your chosen case should have the potential to add value to your operations. Will your selection help reduce costs and enhance efficiency? Will it support innovation and improve sales? Successful AI pilots begin with a well-thought-out and well-defined business case.

Identify Data Sources

A successful AI project requires two types of data: technical data about the model and how it will perform and business-related data to measure how AI will affect the value of the business. Go for reputable data from a cross-section of sources. These sources could be industry websites, government and regulatory organizations’ websites, and journals. Assess the quality and create an Extract Transform Load (ETL) pipeline to help train and test your algorithm.

When planning an AI pilot, consider your internal and external data sources and account for quality and cleanliness variances. Will you use real data, or will a lack of sources require synthetic data? Does your data governance strategy account for finding, cleaning, and storing data? If your company has not yet collected data, one of your AI pilot project’s initial tasks will be setting up a data collection system.

Assemble Required Skill Sets

The key considerations when assembling your team are experience and balance. Build your team as you choose your project. This will ensure you base your decisions more on what you already have access to as opposed to projected possibilities.

A mix of data scientists and engineers working together under the leadership of a project manager well-versed in the requirements of both the business and the technical sides is best. The data scientists’ role — to train and oversee machine learning algorithms — is complementary to the data engineers’ role of building systems relevant to your organization. It is the project managers’ job to explain the business context to the data scientists and engineers.

This will help the technical team develop the most suitable solutions to your problems. The team must be dedicated to the project full-time and fully apprised of the end goal and expected outcomes.

Find Subject Matter Expertise Within Your Organization To Help Validate Output

Keep in mind that domain-specific AI implementation needs actual subject matter expertise. You must procure the data science talent necessary to create the algorithms and analyze the data. This should include domain experts who can ensure your model serves the enterprise business needs as dictated in the use case. If your organization needs to have the required talent on staff and your budget allows it, you should consider external partners. A hybrid team helps get the job done and provides cross-training to benefit your staff.

Create And Execute a Well-Defined Project Plan with Corresponding Success Criteria

Create a clear governance structure to show how the pilot is monitored and executed. This creates confidence within the organization to innovate and experiment further. The success of your company’s digital transformation strategy will benefit from top executive buy-in, so creating confidence and investment in the project is crucial. 

Next Steps With a Successful Pilot and Continued ROI

A pilot is generally considered a mechanism for rapid experimentation and failing fast. However, you need to be aware of what happens when a pilot succeeds. How quickly can you operationalize for full-scale use after you achieve your pilot’s goals? Once your AI pilot helps solve your business case, you can apply the process in-house to benefit other cases. An AI pilot facilitates change management, acting as a change agent who brings transformational change within an organization. This contributes to cost-cutting and validates your organization’s digital transformation strategy. Consider building your domain-specific AI as a product. Deploy algorithms so they are configurable and can be mapped for other use cases while remaining domain-sensitive for the entire organization.

Conclusion

It may seem counterintuitive to expend more energy upfront to make your pilot not only successful but reusable. It is the cautious approach to managing risk as you work toward digital transformation. Unless AI is purely an experiment within the organization, it will be just the tip of the spear in most cases. As you expand the use cases and scope beyond the initial pilots, you will reap the benefits of making your architecture and design modular and reusable.

About the Author

Rigvi Chevala, Evalueserve's Chief Technology Officer (CTO), has more than 16 years of experience leading high-performing product engineering teams in building enterprise-scale products and applications.

As the global head of all technology teams within Evalueserve, Chevala works with multiple lines of business to assess, strategize, and deliver software products and projects based on market and customer needs. Before joining Evalueserve, Chevala served in various leadership roles at Trimble, MRI Software, and Brandmuscle to lead the strategy, delivery, and implementation of enterprise-scale SaaS products. He started his career as a full-stack software engineer.

Chevala holds a master’s degree in computer and information sciences from Cleveland State University and a bachelor’s degree in computers and electronics from Jawaharlal Nehru Technology University in Hyderabad.

Related Stories

No stories found.
CDO Magazine
www.cdomagazine.tech