Last week I delivered a keynote talk at Synechron’s InSync Charlotte event around developing the data driven organization. The talk showed the audience the alignment between digital transformation and data driven practices and how evolving data practices should address small data (through citizen data science programs) and big data governance that can enable organizations to be organizationally ready to experiment with machine learning and artificial intelligence. Later that evening, I shared the stage with Imam Hoque, COO of Quantexa and Sandeep Kumar, Head of Capital Market Solutions at Synechron where we answered questions about AI in the organization.

McKinsey Global Institute – Artificial Intelligence The Next Digital Frontier

Let me share a punchline from a recent WSJ article, Artificial Intelligence Is Ready for Business, Are Businesses Ready for AI?

The report is based on a survey of over 3,000 AI-aware C-level executives across 10 countries and 14 sectors. Only 20% of respondents had adopted AI at scale in a core part of their business. 40% were partial adopters or experimenters, while another 40 percent were essentially contemplators.

So unless you work for a tech company, it’s very likely that your AI journey is just beginning if it has begun at all. The main barriers fall into three categories (i) lack of talent, (ii) data isn’t ready for AI, and (iii) unclear to business leaders what problems and opportunities where AI can drive value.

Preparing for AI

1. 1. Define data governance – This may seem like an odd place to start, but without a data governance policy stating who has access to data and on permissible uses it can lead to organization dysfunction. Some will interpret the lack of a data policy to imply that they can do anything they want with the available data, others will be paralyzed and may prevent basic data from being used in analysis. Data governance teams should also be developing data catalogs, dictionaries, and reviews of data quality that are all important for using data in AI experiments.

1. 1. Establish your data lake – AI and ML algorithms work best when lots of raw data is applied to them. Centralizing primary data sources (including external ones) makes it easier for AI developers and data scientists to tap in and leverage them. In Chapter 3 of my recently published book, Driving Digital: The Leader’s Guide to Digital Transformation Through Technology I cover the data technologies many organizations use to develop nosql data stores and other data lakes.

1. 1. Find industry partners – AI talent is still scarce, and leveraging AI to develop insights and applications may be a long journey for many organizations that cannot attract sufficient talent. In addition, AI applications can be developed to leverage industry knowledge and plug in common data sources. For these reasons, many organizations will benefit by partnering with AI experts in their industry rather than building on their own.

1. 1. Identify examples out of your industry – Many business leaders are still learning about what types of problems can be solved with AI. While you might find some examples in your own industry, you’ll get a better sense of the opportunities by looking at successes in other industries that are strong in the AI areas of interest. Looking for AI in customer experience then maybe review retail experiences. Have a lot of unstructured content, then look at news media for examples of natural language processing and natural language generation. Want to use AI to identify fraud then look for Banking and Fintech examples.

1. Backlog POCs that drive value – Once you have your data policies in order, a starting data lake, an understanding of where industry partners can help you accelerate, and some basic learning in your organization about where AI is successful then you’re ready to explore options in your organizations. Start by asking big, bold questions, consider the value in the result rather than the path to get there and then plan to experiment.

These steps shouldn’t be done in sequence. Data governance is likely to be an ongoing program while building the data lake, identifying partners, and seeking examples should be parallel activities. Once milestones are achieved in these areas, then you can consider brainstorming POCs.

What are your AI plans for 2018?