Once considered complex concepts best left to researchers in labs, now technologies like Artificial Intelligence, Deep Learning, and Neural Networks are regularly used for bold new business applications. There is no doubt that AI is here to stay and that AI applications will continue to disrupt every major industry. Financial institutions are building AI-backed chatbots that draw on years of data from customer site visits. In the same vein of “customer intent,” call center recordings are also being mined for answers to FAQs.
Data Lakes Weren’t Enough
Back in 2014, everyone talked about Big Data, but nobody knew what that really meant for large enterprises. The world has since moved toward cheaper storage, large-scale cloud adoption, and citizen data scientists. This shift allowed companies of all sizes to seriously consider AI, ML, and Deep Learning.
Enterprises have spent millions of dollars creating processed data lakes, which were intended to be a company’s single source of truth, but data science teams still complain about dirty data and the need for analytics-focused master data management (MDM) tools for AI solutions to have a workable knowledgebase. Traditionally, 80% of time is spent on data preparation, while only 20% is spent on the actual data science. Although many have spent years touting the importance of inverting that ratio, most companies continue to overlook solutions within that 80%, instead focusing only on what sits in the 20% simply because it is “jazzy” or “valuation worthy.”
The Inherent Un-Sexiness of Structured Data
Although organizing structured data is a huge challenge, the focus has been increasingly on unstructured data like social, image, IOT, audio, or video data. We have seen interesting applications of these unstructured datasets, such as fast-changing investment signals in capital markets or cyber threat use cases for governments. However, for most client-facing industries, the discussions around structured datasets remains sequestered in board rooms. Companies ignore data problems that could overhaul day-to-day operations but that simply aren’t “cool” enough.
Many companies do not use their first-party datasets optimally because they have not focused on the more “basic” issue of MDM or budgeted properly for these problems. Organizations struggle with denormalized data lakes residing in distributed storage like HDFS, S3, or Azure Blobs. None of these datasets is leading to an explosion of value-generating solutions or applications, which was the original promise of centralized data lakes and the money spent on technology consultants—what we call “Pipe Dream 1.0.”
Now, we see a rush to build AI solutions. Since it is far easier to get budgets approved for “exciting” AI projects than for “boring” MDM projects, AI projects are often approved without assessing an organization’s readiness to embark on this journey. We call this “Pipe Dream 2.0.” Considerations for AI projects should include things like the state of internal data assets, data partnerships, and the readiness of internal teams and capabilities.
Ironically, lost in the AI hype are the reasons AI was needed in the first place: to help organizations monetize their data assets and compete effectively in the marketplace. An AI system requires a knowledgebase stored in the proper context for model training (Information Architecture), but most AI vendors overlook the issue. Problems often arise in these hastily thought-out projects, often reflected in the form of project extensions and scope changes. Although the underlying data structure itself is to blame, the scapegoat often becomes “data science” and “AI.” To succeed, organizations must develop a sound data strategy and ability to capture and curate knowledge at scale for AI to process and mine that knowledge, learn, and take actions.
The First Step in a Proper Data Strategy: Start by Asking the Right Questions
To build a strong foundational data strategy, we suggest answering the following questions:
- Who is driving this roadmap? What kind of internal backing does this initiative have?
- Is the organization aware of this program or not?
- Is the organization ready for this transformation or not, and what makes you think so?
- What is your data strategy, and could it support the program? Do you have a Data Strategy Owner or Sponsor?
- Which business functions are most advanced in terms of data readiness?
- To identify the low-hanging fruit, which functional areas are less sensitive to change? What problems do you intend to solve and why?
- What does the desired end state look like for each stage of this solution?
- How do you prefer to arrive at this desired end state—build, buy, partner?
- Do you prefer internal teams supporting these solutions, or would you rather rely on transformation partners? Do you want to keep certain parts of the solution in-house? Are you prepared to support potential partners internally?
- Where do you stand in terms of current data partnerships?
- Are you currently capturing the right data sets to support the journey?
- Do you have platforms/tools of preference? Why or why not? Are those platforms/tools future-proof or are they third-party tools that fail when put in production?
- Are you working with the right set of vendors/partners? What does your partner ecosystem look like? Is this ecosystem set up for success?
As you begin to ask the right questions, the truth emerges: Data strategy is the weak link, and when it is ignored, the AI journey goes sideways. That’s where we come in. Opera defines the journey from data structure/strategy to AI use cases. We join forces with client teams to drive their goals toward the desired outcome, and we don’t shy away from asking these tough questions. We come in with pre-built and reusable data assets, data and MDM solutions, data partnerships, and data governance frameworks that can accelerate this data-to-value journey.
Opera believes in practical, scalable, and transformative AI, and getting raw datasets to the model-ready stage is an important part of this journey. We lay the right foundation with our MDM solutions and help you build and roll out a series of use cases that generate measurable business outcomes. Through flexible technology frameworks that support SQL, NoSQL, distributed storage/computing, and real time processing capabilities, we meet our customers where they are currently struggling. Finally, all of our solutions are built on open architecture, enabling easy integration with our clients’ existing infrastructure and applications.
Contact us today and find out where your AI journey can take you.