Leading organizations across sectors are heavily experimenting with “deep learning”, the ability to extract features, sentiment, intent from observed data to evaluate what lift it can bring to their businesses. The pay-off is massive: these techniques represent almost half of the data analytics potential value or between $3 and $6 Trillion in annual value, and that’s not including new industries such as autonomous cars. These claims are independently supported by leading research, such as that of Harvard Business School, and experts, such as Google’s Jeff Dean who believes that If you’re not using deep-learning already you should be.
Although deep-learning’s maturity means it is no longer just for experts, one thing is experimenting with it; another is creating value beyond today’s baselines. Most organizations struggle with the paradigm change that deep-learning requires in terms of data architecture, technology stack, and data science expertise. With the right enablers deep-learning capabilities can be within reach.
The first step is enabling right-time data, a real-time ingestion, management and action of user data and metadata. The organization needs to work across silos to formulate data models, event-driven architectures and contextual states, all from a customer or business user point of view. The second step is enabling deeper models and techniques. The organization requires a prioritized roadmap and the expertise to design, test and scale new cognitive models. Last, identifying, evaluating and selecting scalable solutions across the technology stack, such as data processing tools, data platforms, data infrastructure, visualization tools, and a future-proof make vs buy approach. The organization needs to work with vendors to validate performance claims.