At Rise of AI Summit 2020, Dilek Sezgün, leader of IBM's Infrastructure and AI Ecosystem for AI, big data and open source, argues that AI is not magic but a practical journey any company can start, built on data. She cites a figure that AI will contribute 16 trillion dollars to the global economy by 2030, and says most companies use only about 15 percent of their data, with 80 percent left locked and unused. She presents IBM's 'AI ladder' framework, collect, organize, analyze and infuse data, built on a hybrid cloud foundation using Red Hat OpenShift and Cloud Pak for Data, with IBM Garage as a startup-style collaboration model for developing use cases. Her core recommendation is that companies should not start with a broad AI strategy but with concrete pilot projects, building or assigning a small in-house AI team, providing broad AI training, and only then formalizing a strategy once they know what delivers value. In the closing Q&A, she says roughly 80 percent of AI effort goes into data preparation, describes tools such as automated image augmentation, addresses fear of cloud adoption among German SMEs, and stresses that ethical, transparent AI is a prerequisite for broader adoption.
Key takeaways
- Sezgün cites that most companies use only about 15 percent of their data, leaving roughly 80 percent unused.
- She presents IBM's 'AI ladder' framework: collect, organize, analyze and infuse data, on a hybrid cloud foundation.
- IBM's stack centers on Red Hat OpenShift and Cloud Pak for Data as an open, container-based, multi-cloud base.
- Her main recommendation is to start with concrete AI projects rather than a broad strategy document.
- She says roughly 80 percent of AI effort in practice goes into data preparation, citing an IBM tool that augments small image sets.
Detailed notes
AI is a journey, not magic
Sezgün opens by addressing the common perception that AI is a 'black box'. She frames data as the fuel for digital transformation and cites that AI will add 16 trillion dollars to the global economy by 2030. She says most companies use only about 15 percent of their data, with 80 percent locked and unused, and places most companies in a middle stage focused on cost reduction and modernization rather than insight-driven analytics or new AI-based business models.
The AI ladder and IBM's technology stack
She presents IBM's 'AI ladder': collect, organize, analyze and infuse data, framed as the information architecture for AI, requiring decisions on data location (cloud or on-premise), governance, ethics, trust and transparency. IBM's technology stack for this includes Red Hat OpenShift as an open, container-based foundation that works across different clouds and infrastructure providers, Cloud Pak for Data built on top of it, data-scientist tooling for preparing, building, running and managing models, and prebuilt Watson applications and chatbots so companies do not have to start from scratch.
IBM Garage and the call to action: start with projects, not strategy
Sezgün describes IBM Garage as a startup-like collaboration model bringing together client and IBM expertise, whether a client already has use cases or needs help identifying them. Her central recommendation is not to begin with an AI strategy but with concrete projects, since starting small reveals what is actually effective for a given company and industry. She advises building or designating a small in-house AI team that interacts with existing data center teams, providing broad AI training (mentioning Coursera and YouTube resources), and only then formalizing a strategy, followed by internal and external communication about resulting use cases.
Q&A highlights
Asked about companies unsure of their data quality, Sezgün says roughly 80 percent of AI effort goes into data preparation and cites an IBM visual-recognition tool that can generate additional images from as few as five input pictures. Asked about cloud adoption fear, particularly among German SMEs, she notes IBM's own base is Red Hat OpenShift as cloud infrastructure. Asked whether AI is for everyone, she says that even if AI will not suit every business model, most people will encounter it, drawing a comparison to early skepticism about smartphones, and stresses that ethical AI must underpin broader adoption or people will remain afraid to participate.
Quotes
AI isn't magic. AI you could use in everywhere, and the base of everything is data.
Dilek Sezgün
Questions answered in this talk
What is IBM's AI ladder?
According to Dilek Sezgün at Rise of AI Summit 2020, the AI ladder is IBM's framework for the information architecture needed for AI: collecting, organizing, analyzing and infusing data, built on a hybrid cloud foundation such as Red Hat OpenShift and Cloud Pak for Data.
Should a company start AI adoption with a strategy document?
Dilek Sezgün recommends at Rise of AI Summit 2020 that companies start with concrete pilot projects rather than an AI strategy, because running projects reveals what is actually effective for the business before formalizing a broader strategy.
How much of AI project effort goes into data preparation?
Dilek Sezgün states at Rise of AI Summit 2020 that around 80 percent of the effort in AI projects goes into data preparation, and mentions IBM tools, such as one that generates additional training images from as few as five input pictures, to help with this.