AI-enabled business transformation: Closing the gaps

• Understand the required change for development and consumption of AI: Assessing process readiness and people readiness in creating AI capabilities and adopting data-driven insights will help in planning the right organizational training, construction of analytics translator roles, and creation of the proper incentives to drive adoption. From an AI development point of view, leaders in early adopting companies should focus on explainable models that help build trust.

•Define a program/plan to scale from a few use cases to core operations: Communicate the roadmap, as well as long- and short-term goals, and plan the required business process changes accordingly.

While the foundational capabilities, such as digitization, organizational readiness, and adoption, are intertwined with the progress of AI capabilities, realizing that you are collectively embarking on a journey towards perfection will help manage the expectations of the organization.

Execution
This is the process of realizing the goals which were set in planning. Key success factors here are, again, recognizing the right skill sets and taking a holistic approach. Companies frequently execute AI projects as technical software development projects and unintentionally lose sight of the broader strategy. This is an unfortunate but not an uncommon landing zone, as they are fighting overwhelming data management challenges, struggling to fill skill gaps with emerging AI, and attempting to make sustainable long-term technology decisions and investments. Regrettably, as a result, some organizations can struggle to justify the data and technology, investment made and lose their appetite for further AI implementation.

• Treat it as a business problem:The role of humans in the loop at every step of execution cannot be underestimated. For example, machine learning models (supervised) are “trained” (as opposed to “programmed” in a software development world) requiring extensive human involvement and collaboration in obtaining, understanding, and contextualizing/ labeling data as well as validating it for any potential bias. Business data and process owners are an essential part of the development process and product management teams.

• Recognize the difference between “science” and “software development”: Do not push traditional scrum methodologies in all circumstances. Time-boxed approaches create unnecessary inefficiencies and overheads when the goals and next steps are constantly altered by findings/insights gained during the exploration phases. Unlike software development, clear requirements do not guarantee product development success in a typical data science setting.

•Address the “last mile” challenge: The value of any sophisticated model is limited to the extent to which it is adopted and has a positive impact on the relevant outcome. To ensure adoption, embed AI in business processes and complementary products. Secondly, address behavioral problems that destroy the value of AI by ensuring the right incentives are in place for business users, and encourage data-driven performance management.

• Instrument processes and products for “Active” or “Reinforcement” learning: (Business applications can be instrumented with for monitoring usage, measuring outcome metrics and collecting feedback of insights.) Organizations experimenting with AI tend to evaluate AI solutions to decide on further investments and future strategy. While it seems trivial, instrumenting the process/product for essential feedback on a product recommendation and or sales lead closure not only helps to prove success, but also allows you to improve the solution as necessary.

Though this is not an exhaustive list of success factors, if any of those listed are missed out, they can morph into missed expectations and frustration among executive leaders and stakeholders. Non-digital native companies, especially, struggle with the human elements of AI transformation. It is a relatively easier task to choose the right technology platform, hire/grow technical skills, and optimize machine learning algorithms. On the other hand, poor judgement around the non-technical aspects of AI transformation can have long-term impacts, cause strategic failure, and cost businesses their precious competitive edge. Each organization and their business problem is unique, and the cultural shift underlying these success factors will have to be planned and executed in an informed manner.