3 Ideas on Data and AI From IBM's Think 2019 Conference

I had the opportunity to attend IBM's THINK technology conference in San Francisco. After submitting a short film to a competition organised by IBM Australia and New Zealand's Managing Director, David La Rose, I was invited to THINK with four other IBMers. The conference spanned topics such as AI, Cloud, Research, Marketing, Automation and more. I want to share three themes I observed in the sessions I attended, namely in the Data and AI space.

1. Organisations are looking at open-source tools for data science.

Many of IBM's clients shared the lessons they learnt from their data science projects. A common lesson was that organisations recognised the utility of open-source software in building highly complex data science pipelines. Tools such as Python, Jupyter Notebooks and TensorFlow were frequently mentioned. One particular story focused on the challenges of using very large datasets containing proprietary data formats. A client described how they would develop an application that can work with proprietary data A, then another application for proprietary data B and so on. Often they would have disparate teams working on each of these applications making interoperability a challenge. Combined with the sheer scale of their data (in the order of 100s of petabytes), the client concluded that it was almost impossible to get complete coverage of their data and, in turn, build scalable models to generate insights. The solution they adopted was to migrate their data into a cloud environment, cluster their data assets into logical groupings and then build APIs to perform transformations and provide access. All of these tasks were performed using open-source software. Rather than build independent applications based on proprietary formats, they could simply invoke their APIs to retrieve data in a suitable format and then carry out later stages of their data science pipeline like exploratory analysis, feature engineering and modelling.

2. AI projects usually start inside or outside an organisation.

During the Chairman's Address, IBM's Chair, President and CEO, Ginni Rometty, described the approaches that many of IBM's clients have taken to adopt enterprise AI. There were two approaches that stood out in many of the THINK sessions: the outside-in and inside-out paths. The former involves starting with the customer and then working inwards. For example, an organisation could implement a customer-facing chatbot app that answers questions about particular products. The next iteration of the app could involve integration with customer data which provides a more personalised experience for the user. Further iterations may introduce prescriptive models that recommend new products to the customer. The pattern is that the organisation started outside with the customer and then worked inwards to infuse AI into their business. Conversely, the inside-out approach often starts with the organisation's data and uses AI to improve internal decision-making and processes. An example of this is leveraging internal HR employee data to predict attrition or using time-series information to improve workflows. A common characteristic of the inside-out approach that clients mentioned is that the data they use has historically been contained in silos. I made a short video on AI adoption here which goes into this topic further.

3. Data is going to underpin organisations' business platforms.

Mark Foster, the Senior Vice President of IBM's Global Business Services, presented a session on the "blueprint" for smart business. Related to the AI adoption approaches described above, Mark argued that a business platform is needed to connect the outside-in and inside-out approaches. The business platform that he described combines data, processes and people to create value for an organisation. For example, an organisation may hold information about customer transactions, which is used within a workflow to predict fraud, which is subsequently used to inform an employee of suspicious activity that needs to be reviewed. The business platform encapsulates all of these components and it starts with data. A major theme at THINK, which was also reflected in IBM's C-Suite study, was that incumbent data; that is data that has been historically contained within an organisation, is going to be a competitive advantage for businesses. Organisations that can fuel their business platform with incumbent data have the potential to leverage greater insight than those whose data is shared. You can watch Mark's presentation here.