Artificial Intelligence has created a big impact across various industries and the use cases have varied from analyzing radiology images in healthcare to robotic advisors in financial industry. There is a consensus across the board about the usage and benefits of AI.
In a recent survey done by Kearney in South East Asia, we found that while 50 percent of the companies have piloted AI in some form or another, only 17 percent are devoting more than 0.5 percent of their revenues to embed AI solutions into their operations.
Despite the widespread acknowledgment about adopting AI, the real challenge many organizations face is increasing the adoption of AI across the organization and integrating it with the mainstream technology and business processes.
Early adopters of AI still recount the excitement around launching AI pilots, recruiting a team of data scientist, showing the pilot AI results. But since then, many have struggled to streamline and scale AI in a way that AI becomes mainstream just like IT.
The challenge to integrate AI with your core IT persists and is likely to widen which may result in organizations abandoning their AI efforts after few initial pilots. While the initial experiment of AI increases, the abandonment of AI projects is also likely to increase rapidly in next 2-3 years. We believe in order to scale up AI pilots, an organization needs to work on three key dimensions:
People and Organization Structures
While organizations focus a lot on tools and technologies, they underestimate the formidable cultural and organizational barriers which AI initiatives face while scaling up. Without organizational structure and leadership, the projects fail either because they do not have enough AI skillset or many times because of feuds on data and analytics ownership. Organizations which have successfully scaled up AI initiatives have adopted some of the following features:
- Enabled an organization with a platform mindset: To remove business vs IT silos which is the biggest challenge to scale AI, organizations have altered their organization structure from a current BU or geographic focused to a product centric. In this structure, business teams identify the use case and the IT team works together with them to scale up the use case. A typical product team will have roles from product manager, business experts, data stewards, IT analysts and developers, UI experts, AI/ML engineers and data scientists.
- Created an AI team in a ‘federated’ manner – Organizations in recent past have experimented to house their AI skills in different manner across the organization. It can be centralized, decentralized or a ‘federated’ model.
Organizations which have successfully scaled up their AI initiatives, start with a centralized model and gradually scale up to a ‘federated’ model. Once scaled up, the lean central team is still responsible for driving the AI mandate across the organization, standardizing tools and processes and more importantly play the part of incubators for those business unit who are starting their AI journey.
- Continuously educated employees on AI: Having established the need of AI at the leadership level and recruited a team of data scientists, most organizations miss out on training and communicating the importance of AI to their employees. Lack of knowledge on what is AI and how important it is for the organization, can be a key factor of AI not being adopted across the organization. Training can be imparted through various methods such as bootcamps, online courses, workshops etc.
Recently, a pharma company offered AI training across the enterprise, to help executives understand the broad range of possibilities with AI and separate the facts from the fiction. They conducted boot camps for more than 1000 employees to explain the basics of AI and how it will help them and the organization in future.
Though people, process and organization structure may constitute as much as 70% of the effort to scale AI, it is equally important that your AI tools and technologies are standardized and operationalized across the organization. In a future scenario, an organization should be able to seamlessly leverage cloud to choose an existing AI model, customize if needed and associate the data available from the massive data catalogue. In order to achieve the above scenario, many organizations are establishing the following:
- Adopting Cloud for all AI needs: Usually, many organizations start AI with a POC and may not realize how quickly the AI infrastructure needs would grow even if you attempt to scale it across a business unit. It is essential, that even for a POC the entire AI effort is planned over a cloud. This will not only include AI/ML models but also massive data which will be reused across the organization.
- Establishing Data Strategy & Pipelines: If you have the best ML model but you are unable to access the data, you cannot operationalize AI and scalability will still be a distant dream. Hence, it is essential to create a data management strategy which will ensure that all data is standardized and searchable in a data catalogue. Setting up an enterprise data lake over cloud and pooling organizational data in that is usually the first step in this direction.
- Enabling AI as a microservice: Business process is a combination of tasks and some of these tasks can be performed by AI. In order to reuse and scale AI, it is essential to enable AI output as a microservices which will help to streamline the process of developing, testing and reusing AI programs and features across their organizations.
As an example, in our Fintech organization we have created AI APIs for traditional and often used actions such as KYC, reading financial statements and insights generation which are now being used by our lending platform for various acquisition and lending journeys.
Even with the best tools, technologies, people and structure, scaling up AI across the organization will fail if the end-user does not adopt the new processes which AI entails.
- Redesigned processes and workflows to incorporate AI: Organizations have reworked on their legacy processes to enable AI output. For example, one of the most popular cases of AI is predictive maintenance where you can forecast when machines are going to break, and which enables you to fix them in time. Implementing AI means completely changing the process of how equipment’s are maintained, how maintenance staff is rostered, how spare parts are ordered, and how the contracts are formulated with vendors. If you just implement AI technology but you continue to use legacy processes, then you will never be able to realize its benefits.
- Create an AI culture from top: Clear articulation and communication of AI vision across the organization is key to adopting and scaling AI. In organizations which have successfully scaled AI, leaders encourage collaboration and more importantly drive the organization from taking an experience led to data driven decision making. For example, while earlier lending decisioning at various banks and NBFC’s used to take days and were manual, but now, as a process banks have started trusting and relying on AI enabled insights and data driven business rules through which in principal approval for loans are provided in a matter of minutes.
At the end of the day, the organizations need to have a culture where they trust AI and consciously remove human intervention in order to successfully scale up AI
The authors are Sandeep Gupta, Director and Head of Digital Center of Excellence at Kearney and Yogesh Sharma, Head of Digital platforms at Decimal Technologies