3 Principles to move from AI Pilots to AI @ Scale
In my last blog post, I covered how you can deliver an AI pilot in just eight weeks and at the same time design your program in a way to scale the AI across your enterprise. Culture, architecture and technology is fundamental to move from AI pilot to AI @ Scale. I also discussed how IBM is helping one of biggest ANZ-region banks to do AI @ Scale to deliver personalized offers using a bot to automate risk modeling and crediting the money in their customers’ accounts. It is all digital, without the friction of traditional banks processes.
There are three critical principles you need to build AI @ Scale at any enterprise company.
1. Culture of AI @ Scale
People are what drive any business. Almost everybody thrives in a culture which fosters a growth mindset and open collaboration. With AI @ Scale, a culture of cultivating talent, encouraging change agents and fostering seamless collaboration to seed new data initiatives is extremely important.
You can create a culture of experimentation, being agile and permission to fail fast. Business leaders, decision makers and data scientists all can collaborate together with trust and security. The leadership team needs to trust in data science teams and the models they have built.
A platform like IBM Cloud Pak for Data can help backbone of this culture. Data models need to be unbiased and help the business avoid any regulatory concerns. Security needs to be foundational by design, enforcing process controls so that data science models are continuously tested for resiliency. While flexibility and speed are important, but they are not excuses to engage in a random walk through the “AI wilds.
2. Architecture of AI @ Scale
AI begins with IA. Information architecture is the foundational data system that enable users across the company to easily build and operate machine learning and AI systems at scale. A well-defined and structured information architecture that accommodates hybrid cloud, big data, and AI while complying with all applicable regulations like GDPR is essential. The right data governance mechanism is a must-have for AI success.
Hybrid cloud capabilities makes sure that the architecture is flexible and will require less hardware for excess compute needs and storage capacity—and ultimately that it can scale to meet the real time needs of a competitive, growing digital business. If information architecture is robust enough, AI can have the potential to spread both within and outside the pilot use case. Even a “citizen” data scientist would be able to conduct self-service analytics at the point where their piece of the business is happening.
3. Technology of AI @ Scale
The AI tech stack has to be open, resilient, integrated but also modular. An open and robust integrated tech stack can deliver huge improvements in data scientist and AI team productivity – as well as business agility, scalability, availability, and utilization – ultimately leading to potential impact on revenues and cost structures.
Your AI tech stack should support both build and buy personas. More monotonous AI tasks like natural language processing, image recognition, speech-to-text and text-to-speech can be delivered using world-class AI engines like Watson APIs. For more custom use cases like a recommendation engine, personalized marketing or risk modeling, the platform should empower the data science teams to develop AI models using all kind of open frameworks and languages.
Data needs to be the backbone of the AI tech stack—that is what fuels AI potential. In a unified, coherent manner, your platform should be able to help you manage, describe, combine and universally access data. The AI tech stack has to be built on container and microservices architecture to streamline and simplify model development, testing and deployment. Containers and microservices technology create an abstraction layer between AI models and the underlying infrastructure. This lets data science and AI teams focus on model development, business logic and not the infrastructure.
Cloud Pak for Data is the complete, data and AI platform to empower enterprise business with AI @ Scale. See how easy Cloud Pak for Data allows you to easily develop and deploy innovative applications by checking out the video above or trying our no-cost, 7-day trial.
And join the Asia Pacific Data and AI Forum to get regular updates on the innovation IBM is driving around AI.