By Ravi Mehta, Partner; Sushant Kumaraswamy, Director; Vedant Agarwal, Associate Director and Akshay Kumar, Senior Consultant, Deloitte India
One of the biggest paradox in implementing Artificial Intelligence (AI) is that while most (84% of C-Suite Leaders1) believe that scaling AI is a key strategic imperative for their organization, yet very few (16% of C-Suite Leaders1) have managed to go beyond the pilot phase in their AI implementation journey. Clearly, the strategies that made companies successful in the ‘pilot’ phase do not work in the ‘scale up’ phase and hence leaders need to devise a different strategy to successfully scale up AI in their organizations. AI is a complex ‘portfolio’ of multiple types of technologies (e.g., NLP – Natural Language Processing, ML – Machine Learning) and each of these types of technologies have their own ‘capabilities sweet spot’ as well as their unique implementation rythyms, challenges and requirements. In addition, businesses and processes tend to be typically hyper-fragmented these days and hence implementing AI at scale across multiple business units and functions becomes a very intricate and complex exercise. Companies can take a three pronged approach to help resolve the complexities associated with scaling up AI in their organizations – 1) Adopting a portfolio-based implementation approach 2) Building a robust data strategy 3) Streamlining AI governance and policies.
Adopting a portfolio-based implementation approach: Typically, companies either adopt a ‘low hanging fruits’ strategy or a ‘big bang’ strategy. As AI is not a ‘monolithic’ technology but a ‘portfolio’ of technologies, companies may benefit from adopting a ‘portfolio’ based approach to AI implementation. In this approach, companies can construct a ‘project portfolio’ (similar to a ‘stock portfolio’) wherein few processes can be ‘quick wins’ while others can be ‘big wins’. Adopting this approach helps companies to get the best of both worlds – realize some early demonstratable benefits as well as learn some important lessons on the applicability of these AI technologies in the unique context of their organization. Another important consideration is to create a healthy usecase pipeline by marrying technology capabilities with process requirements across multiple business units, regions and functions (for example, once an organization has created an effective invoice reading solution in ‘pilot’ phase, then that solution needs to be scaled up across business units rapidly to maximize benefits).
Building a robust data strategy: Data is the food of a successful AI program. Without right type, quality and volume of data, any AI program is bound to be under-nourished and likely to ultimately fail. As per Deloitte’s latest ‘State of AI in the Enterprise’ survey2, 33% of executives surveyed identified data-related challenges among the top three concerns hampering their company’s AI initiatives. Specifically, there are 2 critical aspects of ‘Data Strategy’ (‘Data Generation’ and ‘Data Governance’) that are critical for AI programs. For many companies, ‘Data Generation’ is still not a focused initiative and hence many AI programs do not receive the right data at the right time. Additionally, ‘Data Governance’ helps in bringing more clarity in data ownership and data usage in a multi-disciplinary and multi-project team environment. For example, a leading telecom major has implemented comprehensive data strategy that helped them to accelerate AI implementation and realize better outcomes.
Streamlining AI governance and policies: AI, unlike most other technologies, typically learns over time and hence has a much more calibrated response to the definition of ‘success’ at various stages along the implementation journey. While in most technology implementation programs, companies typically get into a ‘hyper care and stabilization’ phase post deployment in production, the companies need to get into an ‘enhanced learning and refinement’ phase in case of AI programs. Additionally, companies also need to ensure compliance with their own organizational policies as well as other national and international regulations (e.g., GDPR for data privacy and security). Hence, companies need to create a robust ‘fit-for-purpose’ governance model and risk control framework (e.g., what roles will AI and humans play in a customer facing process?) for implementing AI programs at scale. A key part of this governance model is to clearly define the roles and assign the right leaders for the key roles (e.g., companies can think of creating a role of ‘AI Adoption Leader’ to give defined accountability for increasing AI adoption in the organization).
AI promises to change the way humans live and work. However, this big change is dependent on many important things (as outlined above) happening successfully at the right time. Companies will need to dig deep and take few bold decisions to prepare the ground for AI programs to scale up and succeed in their organizations. The stakes are big, the challenges are bigger, but the human spirit has the knack to soar above the challenges and find ways to succeed in most challenging situations.
📣 The Indian Express is now on Telegram. Click here to join our channel (@indianexpress) and stay updated with the latest headlines